1 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
5 gen_llsym, otherwise call maybe_gen_llsym.
8 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
9 has ',' character inside of it.
11 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
15 as clobbering CC_REGNUM.
17 2018-03-08 Richard Biener <rguenther@suse.de>
20 * tree-scalar-evolution.c: Include tree-into-ssa.h.
21 (follow_copies_to_constant): Do not follow SSA names registered
24 2018-03-08 Richard Biener <rguenther@suse.de>
26 PR tree-optimization/84178
27 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
29 (version_loop_for_if_conversion): Delay update_ssa call.
30 (tree_if_conversion): Delay update_ssa until after predicate
33 2018-03-08 David Malcolm <dmalcolm@redhat.com>
35 PR tree-optimization/84178
36 * tree-if-conv.c (release_bb_predicate): Remove the
37 the assertion that the stmts have NULL use_ops.
38 Discard the statements, asserting that they haven't
39 yet been added to a BB.
41 2018-03-08 Richard Biener <rguenther@suse.de>
43 PR tree-optimization/84746
44 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
45 (phi_translate): Pass in destination ANTIC_OUT set.
46 (phi_translate_1): Likewise. For a simplified result lookup
47 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
48 (phi_translate_set): Adjust.
49 (do_pre_regular_insertion): Likewise.
50 (do_pre_partial_partial_insertion): Likewise.
52 2018-03-08 Martin Liska <mliska@suse.cz>
55 * doc/gcov.texi: Document usage of profile files.
56 * gcov-io.h: Document changes in the format.
58 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
62 * dwarf2out.c (struct dw_line_info_table): Update comments for
64 (FORCE_RESET_NEXT_VIEW): New.
65 (FORCE_RESETTING_VIEW_P): New.
66 (RESETTING_VIEW_P): Check for -1 too.
67 (ZERO_VIEW_P): Likewise.
68 (new_line_info_table): Force-reset next view.
69 (dwarf2out_begin_function): Likewise.
70 (dwarf2out_source_line): Simplify zero_view_p initialization.
71 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
72 view directly. Omit view when omitting .loc at line 0.
74 2018-03-08 Jakub Jelinek <jakub@redhat.com>
76 PR tree-optimization/84740
77 * tree-switch-conversion.c (process_switch): Call build_constructors
78 only if info.phi_count is non-zero.
80 PR tree-optimization/84739
81 * tree-tailcall.c (find_tail_calls): Check call arguments against
82 DECL_ARGUMENTS (current_function_decl) rather than
83 DECL_ARGUMENTS (func) when checking for tail recursion.
85 2018-03-07 Jakub Jelinek <jakub@redhat.com>
87 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
88 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
89 Volker Reichelt's entry and add entries for people that perform
90 GCC fuzzy testing and report numerous bugs.
92 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
95 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
96 readonly data in sdata, if that is disabled.
97 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
98 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
99 -mreadonly-in-sdata option.
101 2018-03-07 Martin Sebor <msebor@redhat.com>
103 PR tree-optimization/84468
104 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
105 basic block when looking for nul assignment.
107 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
110 * except.h (output_function_exception_table): Adjust prototype.
111 * except.c (output_function_exception_table): Remove FNNAME parameter
112 and add SECTION parameter. Ouput one part of the table at a time.
113 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
114 the first part of the exception table and emit unwind directives.
115 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
116 (i386_pe_seh_cold_init): Likewise.
117 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
118 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
119 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
120 (ix86_output_call_insn): Emit a nop in one more case for SEH.
121 * config/i386/winnt.c: Include except.h.
122 (struct seh_frame_state): Add reg_offset, after_prologue and
123 in_cold_section fields.
124 (i386_pe_seh_end_prologue): Set seh->after_prologue.
125 (i386_pe_seh_cold_init): New function.
126 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
127 to seh->in_cold_section.
128 (seh_emit_push): Record the offset of the push.
129 (seh_emit_save): Record the offet of the save.
130 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
131 Test seh->after_prologue to disregard the epilogue.
132 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
133 (i386_pe_end_cold_function): New function.
135 2018-03-07 Jakub Jelinek <jakub@redhat.com>
138 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
139 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
142 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
144 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
145 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
148 * multiple_target.c: Include tree-inline.h and intl.h.
149 (expand_target_clones): Diagnose and fail if node->definition and
150 !tree_versionable_function_p (node->decl).
152 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
154 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
156 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
157 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
158 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
160 2018-03-06 Jakub Jelinek <jakub@redhat.com>
163 * combine.c (try_combine): Use reg_or_subregno instead of handling
164 just paradoxical SUBREGs and REGs.
166 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
168 * config/arc/arc.c (arc_finalize_pic): Remove function.
169 (arc_must_save_register): We use single base PIC register, remove
170 checks to save/restore the PIC register.
171 (arc_expand_prologue): Likewise.
172 * config/arc/arc-protos.h (arc_set_default_type_attributes):
174 (arc_verify_short): Likewise.
175 (arc_attr_type): Likewise.
176 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
177 (walk_stores): Likewise.
178 (arc_address_cost): Make it static.
179 (arc_verify_short): Likewise.
180 (branch_dest): Likewise.
181 (arc_attr_type): Likewise.
182 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
183 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
184 (arc_final_prescan_insn): Remove inserting the nops due to
185 hardware hazards. It is done in reorg step.
186 (insn_length_variant_t): Remove.
187 (insn_length_parameters_t): Likewise.
188 (arc_insn_length_parameters): Likewise.
189 (arc_get_insn_variants): Likewise.
190 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
192 2018-03-06 Jakub Jelinek <jakub@redhat.com>
195 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
198 PR tree-optimization/84687
199 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
201 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
203 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
205 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
206 Rename to ppc_speculation_barrier.
207 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
208 __builtin_ppc_speculation_barrier.
210 2018-03-05 Jakub Jelinek <jakub@redhat.com>
213 * combine.c (combine_simplify_rtx): Don't try to simplify if
214 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
217 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
219 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
220 to 32 bytes when compiling for POWER9.
222 2018-03-05 Jakub Jelinek <jakub@redhat.com>
225 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
226 regparm >= 3 with no arg reg available also for calls with
227 flag_force_indirect_call. Pass decl to ix86_function_regparm.
230 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
232 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
234 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
237 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
239 2018-03-05 Richard Biener <rguenther@suse.de>
241 PR tree-optimization/84486
242 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
243 When inserting a __builtin_assume_aligned call set the LHS
244 SSA name alignment info accordingly.
246 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
248 PR tree-optimization/84114
249 * config/aarch64/aarch64.c (aarch64_reassociation_width)
250 Avoid reassociation of FLOAT_MODE addition.
252 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
254 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
255 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
256 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
257 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
258 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
259 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
260 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
262 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
263 __builtin_ia32_wbinvd): New builtins.
264 (SPECIAL_ARGS2): New.
265 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
266 (SPECIAL_ARGS2): New.
267 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
268 (ix86_valid_target_attribute_inner_p): Ditto.
269 (ix86_init_mmx_sse_builtins): Add special_args2.
270 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
271 TARGET_WBNOINVD_P): New.
272 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
273 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
274 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
275 * config/i386/immintrin.h (_wbinvd): New intrinsic.
276 * config/i386/pconfigintrin.h: New file.
277 * config/i386/wbnoinvdintrin.h: Ditto.
278 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
279 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
281 2018-03-05 Richard Biener <rguenther@suse.de>
283 PR tree-optimization/84670
284 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
286 (BB_VISITED_WITH_VISITED_SUCCS): New define.
287 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
288 (compute_antic_aux): Only assert the number of values in ANTIC_IN
289 doesn't grow if all successors (recursively) were visited at least
292 2018-03-05 Richard Biener <rguenther@suse.de>
294 PR tree-optimization/84650
295 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
296 if executed in the loop pipeline.
298 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
300 * doc/configfiles.texi (Configuration Files): Move info about
301 conditionalizing $target-protos.h to...
302 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
303 differs from $target-protos.h.
305 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
306 Chung-Ju Wu <jasonwucj@gmail.com>
308 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
309 * config/nds32/nds32-multiple.md (setmemsi): Define.
310 * config/nds32/nds32-memory-manipulation.c
311 (nds32_gen_dup_4_byte_to_word_value): New.
312 (emit_setmem_word_loop): New.
313 (emit_setmem_byte_loop): New.
314 (nds32_expand_setmem_loop): New.
315 (nds32_expand_setmem_loop_v3m): New.
316 (nds32_expand_setmem_unroll): New.
317 (nds32_expand_setmem): New.
319 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
320 Chung-Ju Wu <jasonwucj@gmail.com>
322 * config/nds32/nds32-memory-manipulation.c
323 (nds32_emit_load_store): New.
324 (nds32_emit_post_inc_load_store): New.
325 (nds32_emit_mem_move): New.
326 (nds32_emit_mem_move_block): New.
327 (nds32_expand_movmemsi_loop_unknown_size): New.
328 (nds32_expand_movmemsi_loop_known_size): New.
329 (nds32_expand_movmemsi_loop): New.
330 (nds32_expand_movmemsi_unroll): New.
331 (nds32_expand_movmemqi): Rename ...
332 (nds32_expand_movmemsi): ... to this.
333 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
334 (movmemsi): ... to this.
335 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
336 (nds32_expand_movmemsi): ... to this.
338 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
339 Monk Chiang <sh.chiang04@gmail.com>
340 Chung-Ju Wu <jasonwucj@gmail.com>
342 * config/nds32/nds32-protos.h
343 (nds32_expand_load_multiple): New arguments.
344 (nds32_expand_store_multiple): Ditto.
345 (nds32_valid_multiple_load_store): Rename ...
346 (nds32_valid_multiple_load_store_p): ... to this.
347 * config/nds32/nds32-memory-manipulation.c
348 (nds32_expand_load_multiple): Refine implementation.
349 (nds32_expand_store_multiple): Ditto.
350 * config/nds32/nds32-multiple.md
351 (load_multiple): Update nds32_expand_load_multiple interface.
352 (store_multiple): Update nds32_expand_store_multiple interface.
353 * config/nds32/nds32-predicates.c
354 (nds32_valid_multiple_load_store): Rename ...
355 (nds32_valid_multiple_load_store_p): ... to this and refine
357 * config/nds32/predicates.md
358 (nds32_load_multiple_and_update_address_operation): New predicate.
359 (nds32_store_multiple_and_update_address_operation): New predicate.
361 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
362 Chung-Ju Wu <jasonwucj@gmail.com>
364 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
365 (combo): New attribute.
366 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
368 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
370 * config/nds32/nds32.opt: Change -mcmodel= default value.
372 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
373 Monk Chiang <sh.chiang04@gmail.com>
374 Chung-Ju Wu <jasonwucj@gmail.com>
376 * config/nds32/constants.md (unspec_element): New enum.
377 * config/nds32/constraints.md (Umw): New constraint.
378 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
379 * config/nds32/nds32-intrinsic.md: Likewise.
380 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
381 (nds32_valid_smw_lwm_base_p): New.
382 (nds32_output_smw_single_word): New.
383 (nds32_output_lmw_single_word): New.
384 (nds32_expand_unaligned_load): New.
385 (nds32_expand_unaligned_store): New.
386 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
387 (nds32_output_smw_single_word): Declare.
388 (nds32_output_lmw_single_word): Declare.
389 (nds32_expand_unaligned_load): Declare.
390 (nds32_expand_unaligned_store): Declare.
391 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
392 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
393 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
394 NDS32_BUILTIN_UASTORE_DW.
395 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
398 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
399 Kito Cheng <kito.cheng@gmail.com>
400 Chung-Ju Wu <jasonwucj@gmail.com>
402 * config/nds32/nds32-intrinsic.c
403 (nds32_expand_builtin_null_ftype_reg): Delete.
404 (nds32_expand_builtin_reg_ftype_imm): Ditto.
405 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
406 (nds32_read_argument): New.
407 (nds32_legitimize_target): Ditto.
408 (nds32_legitimize_argument): Ditto.
409 (nds32_check_constant_argument): Ditto.
410 (nds32_expand_unop_builtin): Ditto.
411 (nds32_expand_unopimm_builtin): Ditto.
412 (nds32_expand_binop_builtin): Ditto.
413 (nds32_builtin_decl_impl): Ditto.
414 (builtin_description): Ditto.
415 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
416 (nds32_init_builtins_impl): Ditto.
417 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
418 (nds32_builtin_decl): New.
419 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
420 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
422 2018-03-02 Jeff Law <law@redhat.com>
424 * reorg.c (stop_search_p): Handle DEBUG_INSN.
425 (redundant_insn, fill_simple_delay_slots): Likewise.
426 (fill_slots_from_thread): Likewise.
427 * resource.c (mark_referenced_resources): Likewise.
428 (mark_set_resources, find_dead_or_set_registers): Likewise.
430 2018-03-02 Jakub Jelinek <jakub@redhat.com>
432 * substring-locations.h (format_warning_va): Formatting fix for
434 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
436 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
437 * substring-locations.c: Include intl.h.
438 (format_warning_va): Turned into small wrapper around
439 format_warning_n_va, renamed to ...
440 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
441 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
443 (format_warning_at_substring_n): New function.
444 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
445 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
446 format_warning_at_substring with just a shorter name instead of
447 const function pointer.
448 (fmtwarn_n): New function.
449 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
450 appropriate, get rid of all the fmtstr temporaries, move conditionals
451 with G_() wrapped string literals directly into fmtwarn arguments,
452 cast dir.len to (int), formatting fixes.
454 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
456 * doc/invoke.texi: Remove "Cilk Plus" references.
458 2018-03-02 Jakub Jelinek <jakub@redhat.com>
459 Richard Biener <rguenther@suse.de>
462 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
463 for error or warning attributes if CALL_FROM_THUNK_P is set.
466 2018-03-02 Jakub Jelinek <jakub@redhat.com>
469 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
470 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
473 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
474 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
476 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
477 instead of -1U in last predictors element's probability member.
479 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
482 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
483 arguments if they are comparable.
485 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
487 PR tree-optimization/84634
488 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
489 masks and masked_loop_p with a single loop_masks, making sure it's
490 null for bb vectorization.
492 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
494 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
495 (vect_analyze_data_ref_access): Use loop->safe_len rather than
496 loop->force_vectorize to check whether there is no alias.
498 2018-03-02 Jakub Jelinek <jakub@redhat.com>
501 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
503 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
505 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
506 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
507 instead of a loop around prev_real_insn.
508 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
512 * config/i386/i386.c (ix86_print_operand): Use conditional
513 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
516 2018-03-02 Richard Biener <rguenther@suse.de>
518 PR tree-optimization/84427
519 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
520 (bitmap_set_subtract_values): Rewrite to handle multiple
523 (prune_clobbered_mems): Likewise.
524 (phi_translate): Take edge instead of pred/phiblock.
525 (phi_translate_1): Likewise.
526 (phi_translate_set): Likewise. Insert all translated
527 exprs for a value into the set, keeping possibly multiple
528 expressions per value.
529 (compute_antic_aux): Adjust for phi_translate changes.
530 When intersecting union the expressions and prune those
531 not in the final value set, keeping possibly multiple
532 expressions per value. Do not use value-insertion
533 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
534 all expressions. Add verification that the value-sets
535 only shrink during iteration.
536 (compute_partial_antic_aux): Adjust for the phi_translate changes.
537 (do_pre_regular_insertion): Likewise.
538 (do_pre_partial_partial_insertion): Likewise.
540 2018-03-02 Richard Biener <rguenther@suse.de>
543 * config/darwin.c (saved_debug_info_level): New static global.
544 (darwin_asm_lto_start): Disable debug info generation for LTO out.
545 (darwin_asm_lto_end): Restore debug info generation settings.
547 2018-03-01 Martin Liska <mliska@suse.cz>
550 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
553 2018-03-01 Richard Biener <rguenther@suse.de>
556 * dwarf2out.c (gen_variable_die): Properly handle late VLA
557 type annotation with LTO when debug was disabled at compile-time.
559 2018-03-01 Matthew Fortune <mfortune@gmail.com>
561 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
563 (mips_final_postscan_insn): Likewise.
565 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
567 PR rtl-optimization/84528
568 * alias.c (init_alias_target): Add commentary.
569 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
570 a unique base value if the frame pointer is not eliminated
571 to the stack pointer.
573 2018-03-01 Tom de Vries <tom@codesourcery.com>
575 PR rtl-optimization/83327
576 * lra-int.h (hard_regs_spilled_into): Declare.
577 * lra.c (hard_regs_spilled_into): Define.
578 (init_reg_info): Init hard_regs_spilled_into.
579 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
580 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
581 (process_bb_lives): Handle hard_regs_spilled_into.
582 (lra_create_live_ranges_1): Before doing liveness propagation, clear
583 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
585 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
587 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
588 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
589 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
590 * config/rs6000/aix72.h: New file.
592 2018-02-28 Jakub Jelinek <jakub@redhat.com>
594 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
595 instead of warning_at with conditional singular and plural messages
599 * stor-layout.c (update_alignment_for_field): For
600 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
601 && !DECL_PACKED (field), do the alignment update, just use
602 only desired_align instead of MAX (type_align, desired_align)
604 (place_field): Don't do known_align < desired_align handling
605 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
606 is non-NULL, instead do it after rli->prev_field handling and
607 only if not within a bitfield word. For DECL_PACKED (field)
608 use type_align of BITS_PER_UNIT.
610 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
612 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
613 superfluous parentheses and trailing spaces.
615 2018-02-28 Richard Biener <rguenther@suse.de>
617 PR tree-optimization/84584
618 * graphite-scop-detection.c (scop_detection::add_scop): Discard
619 SCoPs with fake exit edge.
621 2018-02-28 Martin Liska <mliska@suse.cz>
624 * timevar.c (timer::print): Fix format to properly print 100%
627 2018-02-28 Richard Biener <rguenther@suse.de>
630 * genmatch.c (capture_info::walk_match): Do not mark
631 captured expressions without operands as expr_p given
632 they act more like predicates and should be subject to
633 "lost tail" side-effect preserving.
635 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
637 PR rtl-optimization/81611
638 * auto-inc-dec.c (attempt_change): Move dead note from
639 mem_insn if it's the next use of regno
640 (find_address): Take address use of reg holding
641 non-incremented value. Add parm to limit search to the named
643 (merge_in_block): Attempt to use a mem insn that is the next
644 use of the original regno.
646 2018-02-27 Martin Sebor <msebor@redhat.com>
649 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
650 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
652 2018-02-27 Martin Sebor <msebor@redhat.com>
655 * diagnostic-core.h (warning_n, error_n, inform_n): Change
656 n argument to unsigned HOST_WIDE_INT.
657 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
658 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
659 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
660 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
662 2018-02-27 Richard Biener <rguenther@suse.de>
664 PR tree-optimization/84512
665 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
666 Do not use the estimate returned from record_stmt_cost for
667 the scalar iteration cost but sum properly using add_stmt_cost.
669 2018-02-27 Richard Biener <rguenther@suse.de>
671 PR tree-optimization/84466
672 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
673 Adjust last change to less strictly validate use operands.
675 2018-02-27 Martin Liska <mliska@suse.cz>
677 PR gcov-profile/84548
678 * gcov.c (process_file): Allow partial overlap and consider it
679 also as group functions.
680 (output_lines): Properly calculate range of lines for a group.
682 2018-02-27 Martin Liska <mliska@suse.cz>
684 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
685 'ggc' suffixes. Change first column width.
686 (timer::print): Fix formatting of the column.
688 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
690 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
691 preserve inline entry blocks for the sake of debug inline
692 entry point markers alone.
693 (remove_unused_locals): Suggest in comments a better place to
694 force the preservation of inline entry blocks that are
695 otherwise unused, but do not preserve them.
697 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
699 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
701 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
704 * config/i386/constraints.md (Bs): Replace
705 ix86_indirect_branch_register with
706 TARGET_INDIRECT_BRANCH_REGISTER.
708 * config/i386/i386.md (indirect_jump): Likewise.
709 (tablejump): Likewise.
710 (*sibcall_memory): Likewise.
711 (*sibcall_value_memory): Likewise.
712 Peepholes of indirect call and jump via memory: Likewise.
713 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
714 (*sibcall_value_GOT_32): Likewise.
715 * config/i386/predicates.md (indirect_branch_operand): Likewise.
716 (GOT_memory_operand): Likewise.
717 (call_insn_operand): Likewise.
718 (sibcall_insn_operand): Likewise.
719 (GOT32_symbol_operand): Likewise.
720 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
722 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
724 PR rtl-optimization/83496
725 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
726 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
727 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
728 redundant insn, if any.
729 (relax_delay_slots): Likewise.
730 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
732 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
734 PR tree-optimization/83965
735 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
736 that grouped statements are part of a reduction chain. Return
737 true if the statement is not marked as a reduction itself but
739 (vect_recog_dot_prod_pattern): Don't check whether the statement
740 is part of a group here.
741 (vect_recog_sad_pattern): Likewise.
742 (vect_recog_widen_sum_pattern): Likewise.
744 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
747 * final.c (rest_of_clean_state): Also look for calls inside sequences.
749 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
752 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
754 (ix86_output_indirect_function_return): New prototype.
755 (ix86_split_simple_return_pop_internal): Likewise.
756 * config/i386/i386.c (indirect_return_via_cx): New.
757 (indirect_return_via_cx_bnd): Likewise.
758 (indirect_thunk_name): Handle return va CX_REG.
759 (output_indirect_thunk_function): Create alias for
760 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
761 (ix86_output_indirect_jmp): Remove the bool argument.
762 (ix86_output_indirect_function_return): New function.
763 (ix86_split_simple_return_pop_internal): Likewise.
764 * config/i386/i386.md (*indirect_jump): Don't pass false
765 to ix86_output_indirect_jmp.
766 (*tablejump_1): Likewise.
767 (simple_return_pop_internal): Change it to define_insn_and_split.
768 Call ix86_split_simple_return_pop_internal to split it for
770 (simple_return_indirect_internal): Call
771 ix86_output_indirect_function_return instead of
772 ix86_output_indirect_jmp.
774 2018-02-26 Jakub Jelinek <jakub@redhat.com>
777 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
778 memset and value initialization afterwards.
780 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
782 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
784 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
787 * common/config/aarch64/aarch64-common.c
788 (aarch_option_optimization_table[]): Switch
789 off fomit-frame-pointer
791 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
792 Chung-Ju Wu <jasonwucj@gmail.com>
794 * config/nds32/nds32-multiple.md (load_multiple): Disallow
796 (store_multiple): Ditto.
798 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
800 * config.gcc: Add --with-cpu support for nds32 target.
801 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
802 * config/nds32/nds32.opt: Add -mcpu= option.
804 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
806 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
807 isel=yes): Warn for these deprecated options.
809 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
811 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
812 ISA_2_5_MASKS_EMBEDDED.
814 2018-02-23 Jakub Jelinek <jakub@redhat.com>
816 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
817 p->max as pointers rather than using iterative_hash_expr.
819 2018-02-23 Carl Love <cel@us.ibm.com>
821 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
822 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
824 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
825 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
826 P8V_BUILTIN_VEC_VUNSIGNED2.
828 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
831 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
832 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
834 * lra-constraints.c (curr_insn_transform): Set up
835 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
837 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
838 * lra-eliminations.c (spill_pseudos): Ditto.
839 (process_insn_for_elimination): Ditto.
840 * lra-lives.c (reg_early_clobber_p): Use the new macros.
841 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
842 LRA_NON_CLOBBERED_ALT.
844 2018-02-22 Martin Sebor <msebor@redhat.com>
846 PR tree-optimization/84480
847 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
848 to maybe_diag_stxncpy_trunc. Call it.
849 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
850 from gimple_fold_builtin_strcpy. Print inlining stack.
851 (handle_builtin_stxncpy): Print inlining stack.
852 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
854 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
857 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
858 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
859 and -fcheck-pointer-bounds are used together.
860 (indirect_thunk_prefix): New enum.
861 (indirect_thunk_need_prefix): New function.
862 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
863 "_nt" instead of "_bnd" for NOTRACK prefix.
864 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
865 (output_indirect_thunk_function): Likewise.
867 (ix86_code_end): Update output_indirect_thunk_function calls.
868 (ix86_output_indirect_branch_via_reg): Replace
869 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
870 (ix86_output_indirect_branch_via_push): Likewise.
871 (ix86_output_function_return): Likewise.
872 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
873 incompatible with -fcf-protection=branch and
874 -fcheck-pointer-bounds.
876 2018-02-22 Steve Ellcey <sellcey@cavium.com>
879 * config/aarch64/aarch64.c (aarch64_print_address_internal):
880 Change gcc_assert call to output_operand_lossage.
882 2018-02-22 Steve Ellcey <sellcey@cavium.com>
884 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
886 2018-02-22 DJ Delorie <dj@redhat.com>
887 Sebastian Perta <sebastian.perta@renesas.com>
888 Oleg Endo <olegendo@gcc.gnu.org>
890 * config/rx/rx.c (rx_rtx_costs): New function.
891 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
893 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
895 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
897 2018-02-22 Martin Liska <mliska@suse.cz>
900 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
901 Add "native" as a possible value.
903 2018-02-22 Martin Liska <mliska@suse.cz>
906 * config/i386/i386.c (ix86_option_override_internal):
907 Add "native" as a possible value for -march and -mtune.
909 2018-02-22 Jakub Jelinek <jakub@redhat.com>
912 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
913 to all type variants.
915 PR tree-optimization/84503
916 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
917 width as info->bitpos + info->bitsize - start.
918 (merged_store_group::merge_overlapping): Simplify width computation.
919 (check_no_overlap): New function.
920 (imm_store_chain_info::try_coalesce_bswap): Compute expected
921 start + width and last_order of the group, fail if check_no_overlap
923 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
924 to group if check_no_overlap fails.
926 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
928 * config/rs6000/altivec.md: Delete contraint arguments to
929 define_expand, define_split, and define_peephole2, and in
930 define_insn_and_split if always unused.
931 * config/rs6000/darwin.md: Ditto.
932 * config/rs6000/dfp.md: Ditto.
933 * config/rs6000/rs6000.md: Ditto.
934 * config/rs6000/sync.md: Ditto.
935 * config/rs6000/vector.md: Ditto.
936 * config/rs6000/vsx.md: Ditto.
938 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
940 * config/rs6000/altivec.md: Write output control strings as braced
941 blocks instead of double-quoted strings.
942 * config/rs6000/darwin.md: Ditto.
943 * config/rs6000/rs6000.md: Ditto.
944 * config/rs6000/vector.md: Ditto.
945 * config/rs6000/vsx.md: Ditto.
947 2018-02-21 Jason Merrill <jason@redhat.com>
949 PR c++/84314 - ICE with templates and fastcall attribute.
950 * attribs.c (build_type_attribute_qual_variant): Remove assert.
952 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
954 * ipa-cp.c (determine_versionability): Fix comment typos.
956 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
959 * ipa-cp.c (determine_versionability): Do not version functions caling
962 2018-02-21 Martin Liska <mliska@suse.cz>
965 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
966 Add "native" as a possible value.
967 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
968 the macro when native cpu detection is available.
970 2018-02-21 Martin Liska <mliska@suse.cz>
973 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
974 Add "native" as a possible value.
975 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
976 when native cpu detection is available.
978 2018-02-21 Jakub Jelinek <jakub@redhat.com>
979 Martin Sebor <msebor@redhat.com>
981 PR tree-optimization/84478
982 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
984 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
985 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
986 support which is conservatively correct, for 2 only stay conservative
987 for maxlen. Formatting and comment capitalization fixes. Add STRICT
988 argument to the 2 argument get_range_strlen, adjust 6 arg
989 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
991 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
992 (gimple_fold_builtin_strlen): Pass true as last argument to
995 2018-02-20 Martin Sebor <msebor@redhat.com>
998 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
999 (builtin_memref::set_base_and_offset): Same. Handle inner references.
1000 (builtin_memref::builtin_memref): Factor out parts into
1001 set_base_and_offset and call it.
1003 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1006 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
1007 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
1008 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
1009 search at the associated MODE_INT.
1011 2018-02-20 Jeff Law <law@redhat.com>
1014 PR tree-optimization/81592
1016 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
1017 for range data rather than using global data.
1018 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
1019 range data rather than using global data.
1020 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
1021 pass it to children as needed.
1022 (struct directive::fmtresult): Similarly.
1023 (struct directive::set_width): Similarly.
1024 (struct directive::set_precision): Similarly.
1025 (format_integer, format_directive, parse_directive): Similarly.
1026 (format_none): Accept unnamed vr_values parameter.
1027 (format_percent, format_floating, format_character): Similarly.
1028 (format_string, format_plain): Similarly.
1029 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
1030 the EVRP range analyzer for range data rather than using global data.
1031 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
1032 gimple-ssa-evrp-analyze.h
1033 (class sprintf_dom_walker): Add after_dom_children member function.
1034 Add evrp_range_analyzer member.
1035 (sprintf_dom_walker::before_dom_children): Call into the EVRP
1036 range analyzer as needed.
1037 (sprintf_dom_walker::after_dom_children): New member function.
1038 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
1040 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
1041 (evrp_range_analyzer::pop_to_marker): Likewise.
1043 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1045 PR tree-optimization/84419
1046 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
1047 with the required type if its current type is compatible but
1050 2018-02-20 Jakub Jelinek <jakub@redhat.com>
1053 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
1054 after vectorization.
1056 2018-02-20 Martin Liska <mliska@suse.cz>
1059 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
1060 possible values if we don't have a hint.
1062 2018-02-20 Martin Liska <mliska@suse.cz>
1066 * final.c (shorten_branches): Build align_tab array with one
1068 * opts.c (finish_options): Add alignment option limit check.
1069 (MAX_CODE_ALIGN): Likewise.
1070 (MAX_CODE_ALIGN_VALUE): Likewise.
1071 * doc/invoke.texi: Document maximum allowed option value for
1072 all -falign-* options.
1074 2018-02-19 Jakub Jelinek <jakub@redhat.com>
1077 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
1078 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
1079 * var-tracking.c (emit_note_insn_var_location): Remove all references
1080 to NOTE_INSN_CALL_ARG_LOCATION.
1081 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
1082 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
1083 Use copy_rtx_if_shared.
1084 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
1085 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
1086 (dwarf2out_var_location): Remove handling of
1087 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
1089 * final.c (final_scan_insn): Remove all references to
1090 NOTE_INSN_CALL_ARG_LOCATION.
1091 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
1092 before dumping final insns.
1093 * except.c (emit_note_eh_region_end): Remove all references to
1094 NOTE_INSN_CALL_ARG_LOCATION.
1095 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
1096 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
1097 * config/arc/arc.c (hwloop_optimize): Likewise.
1098 * config/arm/arm.c (create_fix_barrier): Likewise.
1099 * config/s390/s390.c (s390_chunkify_start): Likewise.
1100 * config/sh/sh.c (find_barrier): Likewise.
1101 * config/i386/i386.c (rest_of_insert_endbranch,
1102 ix86_seh_fixup_eh_fallthru): Likewise.
1103 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
1104 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
1105 * config/frv/frv.c (frv_function_prologue): Likewise.
1106 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
1108 (note_outside_basic_block_p): Remove all references to
1109 NOTE_INSN_CALL_ARG_LOCATION.
1110 * gengtype.c (adjust_field_rtx_def): Likewise.
1111 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
1113 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
1114 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
1117 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
1120 PR tree-optimization/84452
1121 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
1122 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
1125 2018-02-19 Martin Liska <mliska@suse.cz>
1128 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
1130 2018-02-19 Martin Liska <mliska@suse.cz>
1131 Richard Sandiford <richard.sandiford@linaro.org>
1133 PR tree-optimization/82491
1134 * gimple-fold.c (get_base_constructor): Make earlier bail out
1137 2018-02-19 Carl Love <cel@us.ibm.com>
1139 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
1140 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
1142 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
1143 P8V_BUILTIN_VEC_NEG.
1145 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
1147 * config/rl78/rl78.md (movdf): New define expand.
1149 2018-02-19 Martin Liska <mliska@suse.cz>
1152 * doc/invoke.texi: Fix typo.
1153 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
1155 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
1157 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
1158 handle rs6000_single_float and rs6000_double_float specially for
1161 2018-02-16 Jeff Law <law@redhat.com>
1163 * config/rx/rx.c (add_pop_cfi_notes): New function.;
1166 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1169 * ipa-inline.c (inline_small_functions): Fix a typo.
1171 2018-02-16 Nathan Sidwell <nathan@acm.org>
1173 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
1175 2018-02-16 Carl Love <cel@us.ibm.com>
1177 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
1178 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
1179 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
1180 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
1181 expansion to P8V_BUILTIN_VEC_FLOAT2.
1183 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
1185 PR rtl-optimization/70023
1186 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
1187 src_regno into account.
1189 2018-02-16 Carl Love <cel@us.ibm.com>
1191 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
1192 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
1193 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
1194 * config/rs6000/rs6000.c: Remove case statements for
1195 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
1196 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
1197 and P9V_BUILTIN_VEC_VINSERT4B.
1198 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
1199 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
1200 * config/rs6000/vsx.md:
1201 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
1204 2018-02-16 Carl Love <cel@us.ibm.com>
1206 * config/rs6000/altivec.h: Add builtin names vec_extract4b
1208 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
1210 * config/rs6000/rs6000-c.c: Add the definitions for
1211 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
1212 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
1213 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
1214 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
1215 definition for insert4b and define insn *insert3b_internal.
1216 * doc/extend.texi: Add documentation for vec_extract4b.
1218 2018-02-16 Nathan Sidwell <nathan@acm.org>
1220 * doc/extend.texi (Backwards Compatibility): Mention friend
1221 injection. Note for-scope is deprecated.
1222 * doc/invoke.texi (-ffriend-injection): Deprecate.
1224 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
1226 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
1227 that moved to I2, also allow destinations that are a paradoxical
1228 subreg (instead of a normal reg).
1230 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
1233 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
1236 2018-02-16 Richard Biener <rguenther@suse.de>
1238 PR tree-optimization/84037
1239 PR tree-optimization/84016
1241 * config/i386/i386.c (ix86_builtin_vectorization_cost):
1242 Adjust vec_construct for the fact we need additional higher latency
1243 128bit inserts for AVX256 and AVX512 vector builds.
1244 (ix86_add_stmt_cost): Scale vector construction cost for
1247 2018-02-16 Richard Biener <rguenther@suse.de>
1249 PR tree-optimization/84417
1250 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
1251 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
1252 (non_rewritable_lvalue_p): Likewise, use poly-ints.
1254 2018-02-16 Martin Liska <mliska@suse.cz>
1257 * internal-fn.def (ASAN_CHECK): Set proper flags.
1258 (ASAN_MARK): Likewise.
1260 2018-02-16 Julia Koval <julia.koval@intel.com>
1262 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
1263 from PTA_CANNONLAKE.
1265 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1268 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
1269 Use ++iter rather than iter++ for std::list iterators.
1270 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
1271 defer deleting them until all nodes in the forest are processed. Do
1272 free even leaf nodes. Change to_process into auto_vec.
1275 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
1276 * vec.h (vec_default_construct): Use memset instead of placement new
1277 if BROKEN_VALUE_INITIALIZATION is defined.
1278 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
1279 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
1282 PR rtl-optimization/83723
1283 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
1284 * lra.c (lra_substitute_pseudo): Likewise. If true, use
1285 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
1287 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
1289 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
1291 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
1293 PR rtl-optimization/81443
1294 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
1295 from inner REGs to paradoxical SUBREGs.
1297 2018-02-16 Richard Biener <rguenther@suse.de>
1299 PR tree-optimization/84399
1300 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1301 For operands we can analyze at their definition make sure we can
1302 analyze them at each use as well.
1304 2018-02-16 Richard Biener <rguenther@suse.de>
1306 PR tree-optimization/84190
1307 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
1308 volatile accesses if the decl isn't volatile.
1310 2018-02-15 Jason Merrill <jason@redhat.com>
1312 PR c++/84314 - ICE with templates and fastcall attribute.
1313 * attribs.c (build_type_attribute_qual_variant): Don't clobber
1314 TYPE_CANONICAL on an existing type.
1316 2018-02-15 Jakub Jelinek <jakub@redhat.com>
1318 PR tree-optimization/84383
1319 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
1320 dstoff nor call operand_equal_p if dstbase is NULL.
1322 PR tree-optimization/84334
1323 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
1324 also a CONSTANT_CLASS_P, punt.
1326 2018-02-14 Jim Wilson <jimw@sifive.com>
1328 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
1329 first SMALL_OPERAND check. New local min_second_step. Move assert
1330 to where locals are set. Add TARGET_RVC support.
1331 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
1333 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
1335 * doc/invoke.texi: Correct -Wformat-overflow code sample.
1337 2018-02-14 Martin Sebor <msebor@redhat.com>
1339 PR tree-optimization/83698
1340 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
1341 arrays constrain the offset range to their bounds.
1342 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
1343 (builtin_access::overlap): Avoid setting the size of overlap if it's
1345 (maybe_diag_overlap): Also consider arrays when deciding what values
1346 of offsets to include in diagnostics.
1348 2018-02-14 Martin Sebor <msebor@redhat.com>
1351 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
1352 that correspond to the kind of a declaration.
1354 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
1357 * config/pa/pa.md: Load address of PIC label using the linkage table
1358 if the label is nonlocal.
1360 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
1362 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
1363 warning message if user requests -maltivec=be.
1364 * doc/invoke.texi: Document deprecation of -maltivec=be.
1366 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
1369 * config/rs6000/rs6000-c.c: Update definitions for
1370 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
1371 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
1373 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1376 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
1377 add _get_ssp intrinsics. Remove argument from
1378 __builtin_ia32_rdssp[d|q].
1379 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
1380 * config/i386/i386-builtin.def: Remove argument from
1381 __builtin_ia32_rdssp[d|q].
1382 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
1383 ix86_expand_special_args_builtin for _rdssp[d|q].
1384 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
1385 Clear register before usage.
1386 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
1387 Add documentation for new _get_ssp and _inc_ssp intrinsics.
1389 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
1391 PR tree-optimization/84357
1392 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
1393 operand 1 of an ARRAY_REF too.
1395 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
1398 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
1399 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1401 (set_of_reg): New struct.
1402 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
1403 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
1404 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1406 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
1407 Split into bitclr, bitset, bitinvert patterns if appropriate.
1408 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
1409 use rx_fuse_in_memory_bitop.
1410 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
1411 to named insn, correct maximum insn length.
1413 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
1416 * machmode.def: Define a complex mode for PARTIAL_INT.
1417 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
1419 * doc/rtl.texi: Document CSPImode.
1420 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
1422 (msp430_hard_regno_nregs_with_padding): Likewise.
1424 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
1427 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
1429 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
1431 PR rtl-optimization/84169
1432 * combine.c (try_combine): New variable split_i2i3. Set it to true if
1433 we generated a parallel as new i3 and we split that to new i2 and i3
1434 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
1435 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
1436 those to i2, not i1. Partially rewrite this scan code.
1438 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1441 * stor-layout.c (place_field): For variable length fields, adjust
1442 offset_align afterwards not just based on the field's alignment,
1443 but also on the size.
1446 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
1447 of exps and logs in the use_exp2 case.
1449 2018-02-13 Jeff Law <law@redhat.com>
1451 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
1454 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
1457 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
1461 * common.opt (gas-loc-support, gas-locview-support): New.
1462 (ginline-points, ginternal-reset-location-views): New.
1463 * doc/invoke.texi: Document them. Use @itemx where intended.
1464 (gvariable-location-views): Adjust.
1465 * target.def (reset_location_view): New.
1466 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
1467 (TARGET_RESET_LOCATION_VIEW): New.
1468 * doc/tm.texi: Rebuilt.
1469 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
1470 (dwarf2out_default_as_locview_support): New.
1471 (output_asm_line_debug_info): Use option variables.
1472 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
1473 (output_loc_list): Likewise.
1474 (add_high_low_attributes): Check option variables.
1475 Don't output entry view attribute in strict mode.
1476 (gen_inlined_subroutine_die): Check option variables.
1477 (dwarf2out_inline_entry): Likewise.
1478 (init_sections_and_labels): Likewise.
1479 (dwarf2out_early_finish): Likewise.
1480 (maybe_reset_location_view): New, from...
1481 (dwarf2out_var_location): ... here. Call it.
1482 * debug.h (dwarf2out_default_as_loc_support): Declare.
1483 (dwarf2out_default_as_locview_support): Declare.
1484 * hooks.c (hook_int_rtx_insn_0): New.
1485 * hooks.h (hook_int_rtx_insn_0): Declare.
1486 * toplev.c (process_options): Take -gas-loc-support and
1487 -gas-locview-support from dwarf2out. Enable
1488 -gvariable-location-views by default only with locview
1489 assembler support. Enable -ginternal-reset-location-views by
1490 default only if the target defines the corresponding hook.
1491 Enable -ginline-points by default if location views are
1492 enabled; force it disabled if statement frontiers are
1494 * tree-inline.c (expand_call_inline): Check option variables.
1495 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
1497 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
1499 PR tree-optimization/84321
1500 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
1501 handling. Also check whether the anti-range contains any values
1502 that satisfy the mask; switch to a VR_RANGE if not.
1504 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
1507 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
1509 2018-02-13 Martin Jambor <mjambor@suse.cz>
1512 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
1513 of call statements, also set location of a load to a temporary.
1515 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
1517 * config/rl78/rl78.c (add_vector_labels): New function.
1518 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
1519 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
1520 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
1521 which checks that no arguments are passed.
1522 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
1523 * doc/extend.texi: Documentation for the new attribute.
1525 2018-02-13 Andreas Schwab <schwab@suse.de>
1527 * config/riscv/linux.h (CPP_SPEC): Define.
1529 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1532 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
1533 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
1534 OPTION_MASK_ISA_AES as first argument to def_builtin_const
1535 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
1536 instead of OPTION_MASK_ISA_PCLMUL as first argument to
1537 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
1538 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
1539 temporarily for AES and PCLMUL builtins.
1541 PR tree-optimization/84339
1542 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
1543 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
1547 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
1548 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
1549 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
1551 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
1553 * omp-simd-clone.h: New file.
1554 * omp-simd-clone.c: Include omp-simd-clone.h.
1555 (expand_simd_clones): No longer static.
1556 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
1557 cgraph.h and omp-simd-clone.h.
1558 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
1559 (vect_recog_widen_shift_pattern): Formatting fix.
1560 (vect_pattern_recog_1): Don't check optab for calls.
1563 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
1564 operands[2] into a REG before using gen_lowpart on it.
1566 2018-02-12 Jeff Law <law@redhat.com>
1569 * config/sh/sh.c (find_barrier): Consider a sibling call
1572 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
1573 successfully back substituting a reg.
1575 2018-02-12 Richard Biener <rguenther@suse.de>
1577 PR tree-optimization/84037
1578 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
1579 parameter, move visited init to caller.
1580 (vect_slp_analyze_operations): Separate cost from validity
1581 check, initialize visited once for all instances.
1582 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
1584 * tree-vect-stmts.c (vect_model_simple_cost): Make early
1586 (vect_model_promotion_demotion_cost): Likewise.
1587 (vectorizable_bswap): Guard cost modeling with !slp_node
1588 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
1590 (vectorizable_call): Likewise.
1591 (vectorizable_conversion): Likewise.
1592 (vectorizable_assignment): Likewise.
1593 (vectorizable_shift): Likewise.
1594 (vectorizable_operation): Likewise.
1595 (vectorizable_store): Likewise.
1596 (vectorizable_load): Likewise.
1597 (vectorizable_condition): Likewise.
1598 (vectorizable_comparison): Likewise.
1600 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
1603 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
1604 (ASAN_MARK): Fix fnspec to account for return value, change pointer
1605 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
1607 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1610 * params.def (inline-min-speedup): Increase from 8 to 15.
1611 (max-inline-insns-auto): Decrease from 40 to 30.
1612 * ipa-split.c (consider_split): Add some buffer for function to
1613 be considered inlining candidate.
1614 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
1617 2018-02-12 Richard Biener <rguenther@suse.de>
1619 PR tree-optimization/84037
1620 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
1621 matched stmts if we cannot swap the non-matched ones.
1623 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1625 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
1626 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
1627 _mm_maskz_scalef_round_ss): New intrinsics.
1628 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
1629 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
1630 __builtin_ia32_scalefss_round): Remove.
1631 (__builtin_ia32_scalefsd_mask_round,
1632 __builtin_ia32_scalefss_mask_round): New intrinsics.
1633 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
1634 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
1635 ((match_operand:VF_128 2 "<round_nimm_predicate>"
1636 "<round_constraint>")): Changed to ...
1637 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
1638 "<round_scalar_constraint>")): ... this.
1639 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
1640 %0, %1, %2<round_op3>}"): Changed to ...
1641 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
1642 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
1643 %2<round_scalar_mask_op3>}"): ... this.
1644 * config/i386/subst.md (round_scalar_nimm_predicate): New.
1646 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1648 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
1649 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
1650 (_mm_maskz_sqrt_round_ss): New intrinsics.
1651 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
1652 (__builtin_ia32_sqrtsd_mask_round)
1653 (__builtin_ia32_sqrtss_mask_round): New builtins.
1654 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
1655 (__builtin_ia32_sqrtss_round): Remove.
1656 (__builtin_ia32_sqrtsd_mask_round)
1657 (__builtin_ia32_sqrtss_mask_round): New builtins.
1658 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
1659 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
1660 ((match_operand:VF_128 1 "vector_operand"
1661 "xBm,<round_constraint>")): Changed to ...
1662 ((match_operand:VF_128 1 "vector_operand"
1663 "xBm,<round_scalar_constraint>")): ... this.
1664 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
1665 %0, %2, %<iptr>1<round_op3>}): Changed to ...
1666 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
1667 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
1668 %<iptr>1<round_scalar_mask_op3>}): ... this.
1669 ((set_attr "prefix" "<round_prefix>")): Changed to ...
1670 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
1672 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
1675 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
1676 Cast vec_cmpeq result to correct type.
1677 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
1678 Cast vec_cmpgt result to correct type.
1680 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
1682 * final.c (final_scan_insn_1): Renamed from...
1683 (final_scan_insn): ... this. New wrapper, to recover
1684 seen from the outermost call in recursive ones.
1685 * config/sparc/sparc.c (output_return): Drop seen from call.
1686 (output_sibcall): Likewise.
1687 * config/visium/visium.c (output_branch): Likewise.
1689 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
1691 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
1694 2018-02-10 Alan Modra <amodra@gmail.com>
1697 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
1698 Specify LR as an input.
1700 2018-02-10 Jakub Jelinek <jakub@redhat.com>
1703 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
1704 remove_member_access_dummy_vars): New functions.
1705 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
1706 lower_omp_1, execute_lower_omp): Use them.
1708 PR rtl-optimization/84308
1709 * shrink-wrap.c (spread_components): Release todo vector.
1711 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
1713 PR rtl-optimization/57193
1714 * ira-color.c (struct allocno_color_data): Add member
1715 conflict_allocno_hard_prefs.
1716 (update_conflict_allocno_hard_prefs): New.
1717 (bucket_allocno_compare_func): Add a preference based on
1718 conflict_allocno_hard_prefs.
1719 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
1720 (color_allocnos): Remove a dead code. Initiate
1721 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
1723 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1726 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
1727 constraint from =wa to wa. Avoid a subreg on the output operand,
1728 instead use a pseudo and subreg it in a move.
1729 (p9_xxbrd_<mode>): Changed to ...
1730 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
1731 (p9_xxbrd_v2df): New expander.
1732 (p9_xxbrw_<mode>): Changed to ...
1733 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
1734 (p9_xxbrw_v4sf): New expander.
1736 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1738 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
1740 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
1743 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
1744 multiply in 32-bit mode.
1745 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
1746 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
1749 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1751 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
1752 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
1753 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
1754 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
1756 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
1759 * dwarf2out.c (is_trivial_indirect_ref): New function.
1760 (dwarf2out_late_global_decl): Do not generate a location
1761 attribute for variables that have a non-trivial DECL_VALUE_EXPR
1762 and that are not defined in the current unit.
1764 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1766 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
1767 instead of a libcall for UNORDERED.
1769 2018-02-09 Tamar Christina <tamar.christina@arm.com>
1772 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
1773 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
1775 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1778 * config/s390/s390.c (s390_set_current_function): Invoke
1779 s390_indirect_branch_settings also if fndecl didn't change.
1781 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1783 * config/rs6000/rs6000.md (blockage): Set length to zero.
1785 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1787 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
1789 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1792 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
1793 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
1797 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
1798 PARALLEL incoming that failed vt_get_decl_and_offset check.
1801 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
1802 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
1804 (get_variable_section): For decls in named .bss* sections pass true as
1805 second argument to bss_initializer_p.
1807 2018-02-09 Marek Polacek <polacek@redhat.com>
1808 Jakub Jelinek <jakub@redhat.com>
1811 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
1812 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
1813 Sync some changes from cxx_fold_indirect_ref.
1815 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1817 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
1819 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
1820 (BLOCK_INLINE_ENTRY_LABEL): New.
1821 (dwarf2out_var_location): Disregard inline entry markers.
1822 (inline_entry_data): New struct.
1823 (inline_entry_data_hasher): New hashtable type.
1824 (inline_entry_data_hasher::hash): New.
1825 (inline_entry_data_hasher::equal): New.
1826 (inline_entry_data_table): New variable.
1827 (add_high_low_attributes): Add DW_AT_entry_pc and
1828 DW_AT_GNU_entry_view attributes if a pending entry is found
1829 in inline_entry_data_table. Add old entry_pc attribute only
1830 if debug nonbinding markers are disabled.
1831 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
1832 markers are enabled.
1833 (block_within_block_p, dwarf2out_inline_entry): New.
1834 (dwarf2out_finish): Check that no entries remained in
1835 inline_entry_data_table.
1836 * final.c (reemit_insn_block_notes): Handle inline entry notes.
1837 (final_scan_insn, notice_source_line): Likewise.
1838 (rest_of_clean_state): Skip inline entry markers.
1839 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
1841 * gimple.c (gimple_build_debug_inline_entry): New.
1842 * gimple.h (enum gimple_debug_subcode): Add
1843 GIMPLE_DEBUG_INLINE_ENTRY.
1844 (gimple_build_debug_inline_entry): Declare.
1845 (gimple_debug_inline_entry_p): New.
1846 (gimple_debug_nonbind_marker_p): Adjust.
1847 * insn-notes.def (INLINE_ENTRY): New.
1848 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
1849 inline entry marker notes.
1850 (print_insn): Likewise.
1851 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
1852 (INSN_DEBUG_MARKER_KIND): Likewise.
1853 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
1854 * tree-inline.c (expand_call_inline): Build and insert
1855 debug_inline_entry stmt.
1856 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
1857 inline entry blocks early, if nonbind markers are enabled.
1858 (dump_scope_block): Dump fragment info.
1859 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
1860 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
1861 (gimple_build_debug_inline_entry): New.
1862 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
1863 Enable/disable inline entry points too.
1864 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
1865 (DEBUG_INSN): Describe inline entry markers.
1867 * common.opt (gvariable-location-views): New.
1868 (gvariable-location-views=incompat5): New.
1869 * config.in: Rebuilt.
1870 * configure: Rebuilt.
1871 * configure.ac: Test assembler for view support.
1872 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
1873 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
1874 * dwarf2out.c (var_loc_view): New typedef.
1875 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
1876 (dwarf2out_locviews_in_attribute): New.
1877 (dwarf2out_locviews_in_loclist): New.
1878 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
1879 (enum dw_line_info_opcode): Add LI_adv_address.
1880 (struct dw_line_info_table): Add view.
1881 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
1882 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
1883 (zero_view_p): New variable.
1884 (ZERO_VIEW_P): New macro.
1885 (output_asm_line_debug_info): New.
1886 (struct var_loc_node): Add view.
1887 (add_AT_view_list, AT_loc_list): New.
1888 (add_var_loc_to_decl): Add view param. Test it against last.
1889 (new_loc_list): Add view params. Record them.
1890 (AT_loc_list_ptr): Handle loc and view lists.
1891 (view_list_to_loc_list_val_node): New.
1892 (print_dw_val): Handle dw_val_class_view_list.
1893 (size_of_die): Likewise.
1894 (value_format): Likewise.
1895 (loc_list_has_views): New.
1896 (gen_llsym): Set vl_symbol too.
1897 (maybe_gen_llsym, skip_loc_list_entry): New.
1898 (dwarf2out_maybe_output_loclist_view_pair): New.
1899 (output_loc_list): Output view list or entries too.
1900 (output_view_list_offset): New.
1901 (output_die): Handle dw_val_class_view_list.
1902 (output_dwarf_version): New.
1903 (output_compilation_unit_header): Use it.
1904 (output_skeleton_debug_sections): Likewise.
1905 (output_rnglists, output_line_info): Likewise.
1906 (output_pubnames, output_aranges): Update version comments.
1907 (output_one_line_info_table): Output view numbers in asm comments.
1908 (dw_loc_list): Determine current endview, pass it to new_loc_list.
1909 Call maybe_gen_llsym.
1910 (loc_list_from_tree_1): Adjust.
1911 (add_AT_location_description): Create view list attribute if
1912 needed, check it's absent otherwise.
1913 (convert_cfa_to_fb_loc_list): Adjust.
1914 (maybe_emit_file): Call output_asm_line_debug_info for test.
1915 (dwarf2out_var_location): Reset views as needed. Precompute
1916 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
1917 attribute. Set view.
1918 (new_line_info_table): Reset next view.
1919 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
1920 (dwarf2out_source_line): Likewise. Output view resets and labels to
1921 the assembler, or select appropriate line info opcodes.
1922 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
1923 (optimize_string_length): Catch it. Adjust.
1924 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
1925 dw_val_class_view_list, and remove it if no longer needed.
1926 (hash_loc_list): Hash view numbers.
1927 (loc_list_hasher::equal): Compare them.
1928 (optimize_location_lists): Check whether a view list symbol is
1929 needed, and whether the locview attribute is present, and
1930 whether they match. Remove the locview attribute if no longer
1932 (index_location_lists): Call skip_loc_list_entry for test.
1933 (dwarf2out_finish): Call output_asm_line_debug_info for test.
1934 Use output_dwarf_version.
1935 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
1936 (struct dw_val_node): Add val_view_list.
1937 * final.c (SEEN_NEXT_VIEW): New.
1938 (set_next_view_needed): New.
1939 (clear_next_view_needed): New.
1940 (maybe_output_next_view): New.
1941 (final_start_function): Rename to...
1942 (final_start_function_1): ... this. Take pointer to FIRST,
1943 add SEEN parameter. Emit param bindings in the initial view.
1944 (final_start_function): Reintroduce SEEN-less interface.
1945 (final): Rename to...
1946 (final_1): ... this. Take SEEN parameter. Output final pending
1947 next view at the end.
1948 (final): Reintroduce seen-less interface.
1949 (final_scan_insn): Output pending next view before switching
1950 sections or ending a block. Mark the next view as needed when
1951 outputting variable locations. Notify debug backend of section
1952 changes, and of location view changes.
1953 (rest_of_handle_final): Adjust.
1954 * toplev.c (process_options): Autodetect value for debug variable
1955 location views option. Warn on incompat5 without -gdwarf-5.
1956 * doc/invoke.texi (gvariable-location-views): New.
1957 (gvariable-location-views=incompat5): New.
1958 (gno-variable-location-views): New.
1960 2018-02-08 David Malcolm <dmalcolm@redhat.com>
1962 PR tree-optimization/84136
1963 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
1964 that the result of find_edge is non-NULL.
1966 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1969 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
1970 storing integer register in SImode. Fix cost of 256 and 512
1971 byte aligned SSE register store.
1973 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1975 * config/i386/i386.c (ix86_multiplication_cost): Fix
1976 multiplication cost for TARGET_AVX512DQ.
1978 2018-02-08 Marek Polacek <polacek@redhat.com>
1980 PR tree-optimization/84238
1981 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
1984 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1986 PR tree-optimization/84265
1987 * tree-vect-stmts.c (vectorizable_store): Don't treat
1988 VMAT_CONTIGUOUS accesses as grouped.
1989 (vectorizable_load): Likewise.
1991 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1993 PR tree-optimization/81635
1994 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
1995 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
1996 (test_round_for_mask): New functions.
1997 (wide_int_cc_tests): Call test_round_for_mask.
1998 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
1999 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
2000 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
2001 range returned by get_range_info.
2003 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2006 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
2007 * symtab.c: Include builtins.h
2008 (symtab_node::output_to_lto_symbol_table_p): Move here
2009 from lto-streamer-out.c:output_symbol_p.
2010 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
2011 (output_symbol_p): Move all logic to symtab.c
2012 (produce_symtab): Update.
2014 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2016 * config/s390/s390-opts.h (enum indirect_branch): Define.
2017 * config/s390/s390-protos.h (s390_return_addr_from_memory)
2018 (s390_indirect_branch_via_thunk)
2019 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
2020 (enum s390_indirect_branch_type): Define.
2021 * config/s390/s390.c (struct s390_frame_layout, struct
2022 machine_function): Remove.
2023 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
2024 (indirect_branch_table_label_no, indirect_branch_table_name):
2026 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
2027 (enum s390_indirect_branch_option): Define.
2028 (s390_return_addr_from_memory): New function.
2029 (s390_handle_string_attribute): New function.
2030 (s390_attribute_table): Add new attribute handler.
2031 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
2032 (s390_indirect_branch_via_thunk): New function.
2033 (s390_indirect_branch_via_inline_thunk): New function.
2034 (s390_function_ok_for_sibcall): When jumping via thunk disallow
2035 sibling call optimization for non z10 compiles.
2036 (s390_emit_call): Force indirect branch target to be a single
2037 register. Add r1 clobber for non-z10 compiles.
2038 (s390_emit_epilogue): Emit return jump via return_use expander.
2039 (s390_reorg): Handle JUMP_INSNs as execute targets.
2040 (s390_option_override_internal): Perform validity checks for the
2041 new command line options.
2042 (s390_indirect_branch_attrvalue): New function.
2043 (s390_indirect_branch_settings): New function.
2044 (s390_set_current_function): Invoke s390_indirect_branch_settings.
2045 (s390_output_indirect_thunk_function): New function.
2046 (s390_code_end): Implement target hook.
2047 (s390_case_values_threshold): Implement target hook.
2048 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
2050 * config/s390/s390.h (struct s390_frame_layout)
2051 (struct machine_function): Move here from s390.c.
2052 (TARGET_INDIRECT_BRANCH_NOBP_RET)
2053 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
2054 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
2055 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
2056 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
2057 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
2058 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
2059 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
2060 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
2061 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
2062 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
2063 (mnemonic attribute): Add values which aren't recognized
2065 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
2066 pattern for branch conversion. Fix mnemonic attribute.
2067 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
2068 indirect branch via thunk if requested.
2069 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
2070 ("*indirect_jump"): Disable for branch conversion using out of
2072 ("indirect_jump_via_thunk<mode>_z10")
2073 ("indirect_jump_via_thunk<mode>")
2074 ("indirect_jump_via_inlinethunk<mode>_z10")
2075 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
2076 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
2077 ("casesi_jump_via_inlinethunk<mode>_z10")
2078 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
2079 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
2080 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
2081 ("*indirect2_jump"): Disable for branch conversion.
2082 ("casesi_jump"): Turn into expander and expand patterns for branch
2084 ("return_use"): New expander.
2085 ("*return"): Emit return via thunk and rename it to ...
2086 ("*return<mode>"): ... this one.
2087 * config/s390/s390.opt: Add new options and and enum for the
2090 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2092 * lra-constraints.c (match_reload): Unconditionally use
2093 gen_lowpart_SUBREG, rather than selecting between that
2094 and equivalent gen_rtx_SUBREG code.
2096 2018-02-08 Richard Biener <rguenther@suse.de>
2098 PR tree-optimization/84233
2099 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
2100 changed flag instead of boguously re-using phi_inserted.
2102 2018-02-08 Martin Jambor <mjambor@suse.cz>
2104 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
2105 static local variables.
2107 2018-02-08 Richard Biener <rguenther@suse.de>
2109 PR tree-optimization/84278
2110 * tree-vect-stmts.c (vectorizable_store): When looking for
2111 smaller vector types to perform grouped strided loads/stores
2112 make sure the mode is supported by the target.
2113 (vectorizable_load): Likewise.
2115 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2117 * config/aarch64/aarch64.c (aarch64_components_for_bb):
2118 Increase LDP/STP opportunities by adding adjacent callee-saves.
2120 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2122 PR rtl-optimization/84068
2123 PR rtl-optimization/83459
2124 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
2126 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
2128 PR tree-optimization/84224
2129 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
2130 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
2133 2018-02-07 Iain Sandoe <iain@codesourcery.com>
2136 * config/rs6000/altivec.md (*restore_world): Remove LR use.
2137 * config/rs6000/predicates.md (restore_world_operation): Adjust op
2138 count, remove one USE.
2140 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
2142 * doc/install.texi (Configuration): Document the
2143 --with-long-double-format={ibm,ieee} PowerPC configuration
2147 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
2148 Convert from define_expand to be define_insn_and_split. Rework
2149 float/double/_Float128 conversions to QI/HI/SImode to work with
2150 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
2151 conversions to QI/HImode types did a store and then a load to
2152 truncate the value. For conversions to VSX registers, don't split
2153 the insn, instead emit the code directly. Use the code iterator
2154 any_fix to combine signed and unsigned conversions.
2155 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
2156 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
2157 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2158 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
2159 (fix_<mode>di2_hw): Likewise.
2160 (fixuns_<mode>di2_hw): Likewise.
2161 (fix_<mode>si2_hw): Likewise.
2162 (fixuns_<mode>si2_hw): Likewise.
2163 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
2164 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2165 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
2166 fix<uns>_trunc<SFDF:mode>si2_p8.
2167 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
2169 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
2170 (fix<uns>_<mode>_mem): Likewise.
2171 (fctiw<u>z_<mode>_mem): Likewise.
2172 (fix<uns>_<mode>_mem): Likewise.
2173 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
2174 the register allocator from doing a direct move to the GPRs to do
2175 a store, and instead use the ISA 3.0 store byte/half-word from
2176 vector register instruction. For IEEE 128-bit floating point,
2177 also optimize stores of 32-bit ints.
2178 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
2180 2018-02-07 Alan Hayward <alan.hayward@arm.com>
2182 * genextract.c (push_pathstr_operand): New function to support
2184 (walk_rtx): Call push_pathstr_operand.
2185 (print_path): Support [a-zA-Z].
2187 2018-02-07 Richard Biener <rguenther@suse.de>
2189 PR tree-optimization/84037
2190 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
2191 (cse_and_gimplify_to_preheader): Declare.
2192 (vect_get_place_in_interleaving_chain): Likewise.
2193 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
2195 (_loop_vec_info::~_loop_vec_info): Delete it.
2196 (cse_and_gimplify_to_preheader): New function.
2197 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
2198 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
2199 (vectorizable_load): Likewise. For grouped stores always base
2200 the IV on the first element.
2201 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
2202 condition before gimplifying.
2204 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2206 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
2207 *DIV_EXPR and *MOD_EXPR.
2209 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
2212 * config/i386/i386.c (ix86_option_override_internal): Mask out
2213 the CF_SET bit when checking -fcf-protection.
2215 2018-02-07 Tom de Vries <tom@codesourcery.com>
2218 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
2221 2018-02-07 Richard Biener <rguenther@suse.de>
2223 PR tree-optimization/84204
2224 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
2227 PR tree-optimization/84205
2228 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
2229 special-case isl_ast_op_zdiv_r.
2231 PR tree-optimization/84223
2232 * graphite-scop-detection.c (gather_bbs::before_dom_children):
2233 Only add conditions from within the region.
2234 (gather_bbs::after_dom_children): Adjust.
2236 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
2239 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
2240 * config/avr/avr.md: Only post-reload split REG-REG moves if
2241 either register is GENERAL_REG_P.
2243 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2245 PR tree-optimization/84235
2246 * tree-ssa-scopedtables.c
2247 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
2248 if the subtraction is performed in floating point type where NaNs are
2249 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
2250 build 1. Formatting fix.
2252 2018-02-06 Jakub Jelinek <jakub@redhat.com>
2255 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
2256 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
2257 and skip it regardless of bb boundaries. Use CALL_P macro,
2258 don't test INSN_P (insn) together with CALL_P or JUMP_P check
2259 unnecessarily, formatting fix.
2261 2018-02-06 Michael Collison <michael.collison@arm.com>
2263 * config/arm/thumb2.md:
2264 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
2265 (*thumb_mov_notscc): Ditto.
2267 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
2270 * config/rs6000/rs6000.md (su code attribute): Use "u" for
2271 unsigned_fix, not "s".
2273 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2275 * configure.ac (gcc_fn_eh_frame_ro): New function.
2276 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
2277 correct .eh_frame permissions.
2278 * configure: Regenerate.
2280 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
2282 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
2285 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2287 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2288 Display warning message for -mno-speculate-indirect-jumps.
2290 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
2292 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
2294 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
2296 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
2298 PR tree-optimization/84225
2299 * tree-eh.c (find_trapping_overflow): Only call
2300 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
2302 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2305 * config/i386/i386.c: Reimplement the check of possible options
2306 -mibt/-mshstk conbination. Change error messages.
2307 * doc/invoke.texi: Fix a typo: remove extra '='.
2309 2018-02-06 Marek Polacek <polacek@redhat.com>
2311 PR tree-optimization/84228
2312 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
2314 2018-02-06 Tamar Christina <tamar.christina@arm.com>
2317 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
2318 emitted arch directives.
2319 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
2320 __ARM_FEATURE_COPROC before changing architectures.
2322 2018-02-06 Richard Biener <rguenther@suse.de>
2324 * config/i386/i386.c (print_reg): Fix typo.
2325 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
2327 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
2329 * configure: Regenerate.
2331 2018-02-05 Martin Sebor <msebor@redhat.com>
2333 PR tree-optimization/83369
2334 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
2337 2018-02-05 Martin Liska <mliska@suse.cz>
2339 * doc/invoke.texi: Cherry-pick upstream r323995.
2341 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
2343 * ira.c (ira_init_register_move_cost): Adjust comment.
2345 2018-02-05 Martin Liska <mliska@suse.cz>
2347 PR gcov-profile/84137
2348 * doc/gcov.texi: Fix typo in documentation.
2350 2018-02-05 Martin Liska <mliska@suse.cz>
2352 PR gcov-profile/83879
2353 * doc/gcov.texi: Document necessity of --dynamic-list-data when
2354 using dlopen functionality.
2356 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
2358 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
2359 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
2360 _mm_maskz_range_ss, _mm_mask_range_round_ss,
2361 _mm_maskz_range_round_ss): New intrinsics.
2362 (__builtin_ia32_rangesd128_round)
2363 (__builtin_ia32_rangess128_round): Remove.
2364 (__builtin_ia32_rangesd128_mask_round,
2365 __builtin_ia32_rangess128_mask_round): New builtins.
2366 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
2367 __builtin_ia32_rangess128_round): Remove.
2368 (__builtin_ia32_rangesd128_mask_round,
2369 __builtin_ia32_rangess128_mask_round): New builtins.
2370 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
2371 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
2372 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
2373 "<round_saeonly_constraint>")): Changed to ...
2374 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
2375 "<round_saeonly_scalar_constraint>")): ... this.
2376 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
2377 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
2378 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
2379 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
2380 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
2382 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
2384 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
2386 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
2387 Remove all values except native, 8540 and 8548.
2389 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
2391 * config/i386/i386.c (ix86_output_function_return): Pass
2392 INVALID_REGNUM, instead of -1, as invalid register number to
2393 indirect_thunk_name and output_indirect_thunk.
2395 2018-02-02 Julia Koval <julia.koval@intel.com>
2397 * config.gcc: Add -march=icelake.
2398 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
2399 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
2400 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
2401 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
2402 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
2403 (processor_target_table): Add icelake.
2404 (ix86_option_override_internal): Handle new PTAs.
2405 (get_builtin_code_for_version): Handle icelake.
2406 (M_INTEL_COREI7_ICELAKE): New.
2407 (fold_builtin_cpu): Handle icelake.
2408 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
2409 * doc/invoke.texi: Add -march=icelake.
2411 2018-02-02 Julia Koval <julia.koval@intel.com>
2413 * config/i386/i386.c (ix86_option_override_internal): Change flags type
2414 to wide_int_bitmask.
2415 * wide-int-bitmask.h: New.
2417 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2420 * config/i386/i386.md: Replace Pmode with word_mode in
2421 builtin_setjmp_setup and builtin_longjmp to support x32.
2423 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
2427 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
2429 (rs6000_supported_cpu_names): New static variable.
2430 (linux_cpu_translation_table): Likewise.
2431 (elf_platform) <cpu>: Define new static variable and use it.
2432 Translate kernel AT_PLATFORM name to canonical name if needed.
2433 Error if platform name is unknown.
2435 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
2438 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
2440 2018-02-01 Jeff Law <law@redhat.com>
2443 * config/i386/i386.c (release_scratch_register_on_entry): Add new
2444 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
2445 the scratch if RELEASE_VIA_POP is false.
2446 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
2447 If we have to save a temporary register, decrement SIZE appropriately.
2448 Pass new arguments to release_scratch_register_on_entry.
2449 (ix86_adjust_stack_and_probe): Likewise.
2450 (ix86_emit_probe_stack_range): Pass new arguments to
2451 release_scratch_register_on_entry.
2453 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
2455 PR rtl-optimization/84157
2456 * combine.c (change_zero_ext): Use REG_P predicate in
2457 front of HARD_REGISTER_P predicate.
2459 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
2461 * config/avr/avr.c (avr_option_override): Move disabling of
2462 -fdelete-null-pointer-checks to...
2463 * common/config/avr/avr-common.c (avr_option_optimization_table):
2466 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2468 PR tree-optimization/81635
2469 * tree-data-ref.c (split_constant_offset_1): For types that
2470 wrap on overflow, try to use range info to prove that wrapping
2473 2018-02-01 Renlin Li <renlin.li@arm.com>
2476 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
2478 (aarch64_register_move_cost): Likewise.
2479 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
2481 (REG_CLASS_NAMES): Likewise.
2482 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
2483 TAILCALL_ADDR_REGS. Remove IP registers.
2484 * config/aarch64/aarch64.md (Ucs): Update register constraint.
2486 2018-02-01 Richard Biener <rguenther@suse.de>
2488 * domwalk.h (dom_walker::dom_walker): Add additional constructor
2489 for specifying RPO order and allow NULL for that.
2490 * domwalk.c (dom_walker::dom_walker): Likewise.
2491 (dom_walker::walk): Handle NULL RPO order.
2492 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
2494 (rewrite_update_dom_walker): Likewise.
2495 (mark_def_dom_walker): Likewise.
2497 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2499 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
2500 (aarch64_maybe_expand_sve_subreg_move): Declare.
2501 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
2502 * config/aarch64/predicates.md (aarch64_any_register_operand): New
2504 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
2505 that are semantically a reverse operation.
2506 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
2507 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
2508 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
2510 (aarch64_can_change_mode_class): For big-endian, forbid changes
2511 between two SVE modes if they have different element sizes.
2513 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2515 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
2516 the TImode handling for big-endian targets.
2518 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2520 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
2521 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
2523 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
2524 Remove BSWAP handing for big-endian targets and use the form of
2525 LD1RQ appropariate for the mode.
2527 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2529 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
2530 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
2533 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2536 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
2537 check for operands that need to go through aarch64_sve_reload_be.
2539 2018-02-01 Jakub Jelinek <jakub@redhat.com>
2541 PR tree-optimization/81661
2542 PR tree-optimization/84117
2543 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
2544 * tree-eh.c: Include gimplify.h.
2545 (find_trapping_overflow, replace_trapping_overflow,
2546 rewrite_to_non_trapping_overflow): New functions.
2547 * tree-vect-loop.c: Include tree-eh.h.
2548 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
2549 * tree-data-ref.c: Include tree-eh.h.
2550 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
2552 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
2554 PR rtl-optimization/84123
2555 * combine.c (change_zero_ext): Check if hard register satisfies
2556 can_change_dest_mode before calling gen_lowpart_SUBREG.
2558 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
2561 * ira.c (ira_init_register_move_cost): Remove assert.
2563 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2565 PR rtl-optimization/84071
2566 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
2567 * doc/tm.texi: Regenerate.
2569 2018-01-31 Richard Biener <rguenther@suse.de>
2571 PR tree-optimization/84132
2572 * tree-data-ref.c (analyze_miv_subscript): Properly
2573 check whether evolution_function_is_affine_multivariate_p
2574 before calling gcd_of_steps_may_divide_p.
2576 2018-01-31 Julia Koval <julia.koval@intel.com>
2579 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
2580 * config/i386/i386.md (rdpid_rex64) New.
2581 (rdpid): Make 32bit only.
2583 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
2586 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
2587 an IDENTIFIER_NODE for FUNCTION_TYPE's.
2589 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2592 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
2594 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
2596 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2598 PR rtl-optimization/84071
2599 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
2600 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
2602 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2604 * config/arc/arc.c (arc_handle_aux_attribute): New function.
2605 (arc_attribute_table): Add 'aux' attribute.
2606 (arc_in_small_data_p): Consider aux like variables.
2607 (arc_is_aux_reg_p): New function.
2608 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
2609 (arc_get_aux_arg): New function.
2610 (prepare_move_operands): Handle aux-register access.
2611 (arc_handle_aux_attribute): New function.
2612 * doc/extend.texi (ARC Variable attributes): Add subsection.
2614 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2616 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
2617 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
2618 (arc_attribute_table): Add 'uncached' attribute.
2619 (arc_print_operand): Print '.di' flag for uncached memory
2621 (arc_in_small_data_p): Do not consider for small data the uncached
2623 (arc_is_uncached_mem_p): New function.
2624 * config/arc/predicates.md (compact_store_memory_operand): Check
2625 for uncached memory accesses.
2626 (nonvol_nonimm_operand): Likewise.
2627 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
2629 2018-01-31 Jakub Jelinek <jakub@redhat.com>
2632 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
2633 falign-loops=): Add Optimization flag.
2635 2018-01-30 Jeff Law <law@redhat.com>
2638 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
2639 INT_REGISTERS_SAVED. Check it prior to calling
2640 get_scratch_register_on_entry.
2641 (ix86_adjust_stack_and_probe): Similarly.
2642 (ix86_emit_probe_stack_range): Similarly.
2643 (ix86_expand_prologue): Corresponding changes.
2645 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2648 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
2649 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
2651 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
2654 * lra-constraints.c (curr_insn_transform): Process AND in the
2657 2018-01-30 Jakub Jelinek <jakub@redhat.com>
2659 PR rtl-optimization/83986
2660 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
2661 dependence against last_pending_memory_flush in addition to
2664 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
2666 PR tree-optimization/81611
2667 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
2670 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2673 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
2676 2018-01-30 Richard Biener <rguenther@suse.de>
2677 Jakub Jelinek <jakub@redhat.com>
2679 PR tree-optimization/84111
2680 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
2681 inner loops added during recursion, as they don't have up-to-date
2684 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2687 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
2688 (can_inline_edge_by_limits_p): ... here.
2689 (can_early_inline_edge_p, check_callers,
2690 update_caller_keys, update_callee_keys, recursive_inlining,
2691 add_new_edges_to_heap, speculation_useful_p,
2692 inline_small_functions,
2693 inline_small_functions, flatten_function,
2694 inline_to_all_callers_1): Update.
2696 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2698 * profile-count.c (profile_count::combine_with_ipa_count): Handle
2701 2018-01-30 Richard Biener <rguenther@suse.de>
2703 PR tree-optimization/83008
2704 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
2705 invariant and constant vector uses in stmts when they need
2708 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2711 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
2712 * configure: Regenerate.
2714 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2716 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
2718 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
2719 Use gen_rtx_REG rather than gen_lowpart.
2721 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2723 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
2724 rather than 0 when creating partial subregs.
2726 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2728 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
2731 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
2734 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
2735 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
2736 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
2737 flags. This restores the settings used before the 2017-07-24.
2738 Turning off pre increment/decrement/modify allows IVOPTS to
2739 optimize DF/SF loops where the index is an int.
2741 2018-01-29 Richard Biener <rguenther@suse.de>
2742 Kelvin Nilsen <kelvin@gcc.gnu.org>
2745 * tree-vect-stmts.c (vectorizable_call): Don't call
2746 targetm.vectorize_builtin_md_vectorized_function if callee is
2749 2018-01-22 Carl Love <cel@us.ibm.com>
2751 * doc/extend.tex: Fix typo in second arg in
2752 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
2754 2018-01-29 Richard Biener <rguenther@suse.de>
2756 PR tree-optimization/84086
2757 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
2758 (flush_ssaname_freelist): When SSA names were released reset
2759 the SCEV hash table.
2761 2018-01-29 Richard Biener <rguenther@suse.de>
2763 PR tree-optimization/84057
2764 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
2765 removed paths when removing edges.
2767 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
2769 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
2770 -mfunction-return=@var{choice}.
2772 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
2775 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
2777 (layout::print_source_line): Likewise.
2778 (test_get_line_width_without_trailing_whitespace): Add test cases.
2780 2018-01-27 Jakub Jelinek <jakub@redhat.com>
2783 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
2786 2018-01-26 Jim Wilson <jimw@sifive.com>
2788 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
2790 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
2793 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2795 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
2796 and CMP + SUB-immediate -> SUBS.
2798 2018-01-26 Martin Sebor <msebor@redhat.com>
2800 PR tree-optimization/83896
2801 * tree-ssa-strlen.c (get_string_len): Rename...
2802 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
2803 Avoid assuming length is constant.
2804 (handle_char_store): Use HOST_WIDE_INT for string length.
2806 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
2809 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
2810 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
2812 2018-01-26 Richard Biener <rguenther@suse.de>
2814 PR rtl-optimization/84003
2815 * dse.c (record_store): Only record redundant stores when
2816 the earlier store aliases at least all accesses the later one does.
2818 2018-01-26 Jakub Jelinek <jakub@redhat.com>
2820 PR rtl-optimization/83985
2821 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
2822 REG_CFA_RESTORE insns.
2823 (delete_unmarked_insns): Don't ignore separate shrink wrapping
2824 REG_CFA_RESTORE insns here.
2827 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
2828 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
2830 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2832 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
2833 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
2834 (arc_init): Likewise.
2835 (arc_override_options): Likewise.
2836 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
2838 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
2840 * config/arc/arc.h (TARGET_DBNZ): Define.
2841 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
2842 properly set the tune attribute.
2843 (dbnz): Use TARGET_DBNZ guard.
2844 * config/arc/arc.opt (mtune): Add core3 option.
2846 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2848 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
2849 recognize new pic like addresses.
2850 (arc_delegitimize_address): Clean up.
2852 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2854 * config/arc/arc-arches.def: Option mrf16 valid for all
2856 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
2857 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
2858 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
2859 * config/arc/arc-tables.opt: Regenerate.
2860 * config/arc/arc.c (arc_conditional_register_usage): Handle
2861 reduced register file case.
2862 (arc_file_start): Set must have build attributes.
2863 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
2865 * config/arc/arc.opt (mrf16): Add new option.
2866 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
2867 * config/arc/genmultilib.awk: Handle new mrf16 option.
2868 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
2869 * config/arc/t-multilib: Regenerate.
2870 * doc/invoke.texi (ARC Options): Document mrf16 option.
2872 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2874 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
2875 * config/arc/arc.c (arc_handle_secure_attribute): New function.
2876 (arc_attribute_table): Add 'secure_call' attribute.
2877 (arc_print_operand): Print secure call operand.
2878 (arc_function_ok_for_sibcall): Don't optimize tail calls when
2880 (arc_is_secure_call_p): New function. * config/arc/arc.md
2881 (call_i): Add support for sjli instruction.
2882 (call_value_i): Likewise.
2883 * config/arc/constraints.md (Csc): New constraint.
2885 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2886 John Eric Martin <John.Martin@emmicro-us.com>
2888 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
2889 * config/arc/arc.c (_arc_jli_section): New struct.
2890 (arc_jli_section): New type.
2891 (rc_jli_sections): New static variable.
2892 (arc_handle_jli_attribute): New function.
2893 (arc_attribute_table): Add jli_always and jli_fixed attribute.
2894 (arc_file_end): New function.
2895 (TARGET_ASM_FILE_END): Define.
2896 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
2897 (arc_add_jli_section): New function.
2898 (jli_call_scan): Likewise.
2899 (arc_reorg): Call jli_call_scan.
2900 (arc_output_addsi): Remove 'S' from printing asm operand.
2901 (arc_is_jli_call_p): New function.
2902 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
2904 (movhi_insn): Likewise.
2905 (movsi_insn): Likewise.
2906 (movsi_set_cc_insn): Likewise.
2907 (loadqi_update): Likewise.
2908 (load_zeroextendqisi_update): Likewise.
2909 (load_signextendqisi_update): Likewise.
2910 (loadhi_update): Likewise.
2911 (load_zeroextendhisi_update): Likewise.
2912 (load_signextendhisi_update): Likewise.
2913 (loadsi_update): Likewise.
2914 (loadsf_update): Likewise.
2915 (movsicc_insn): Likewise.
2916 (bset_insn): Likewise.
2917 (bxor_insn): Likewise.
2918 (bclr_insn): Likewise.
2919 (bmsk_insn): Likewise.
2920 (bicsi3_insn): Likewise.
2921 (cmpsi_cc_c_insn): Likewise.
2922 (movsi_ne): Likewise.
2923 (movsi_cond_exec): Likewise.
2924 (clrsbsi2): Likewise.
2935 (call_i): Remove 'S' asm letter, add jli instruction.
2936 (call_value_i): Likewise.
2937 * config/arc/arc.op (mjli-always): New option.
2938 * config/arc/constraints.md (Cji): New constraint.
2939 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
2941 (subsf3_fpx): Likewise.
2942 (mulsf3_fpx): Likewise.
2943 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
2945 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
2947 * doc/invoke.texi (ARC): Document mjli-always option.
2949 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
2951 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
2952 avoid addition with 0 and use incw and decw where possible.
2954 2018-01-26 Richard Biener <rguenther@suse.de>
2956 PR tree-optimization/81082
2957 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
2958 association if it requires casting to unsigned.
2959 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
2960 from fold_plusminus_mult_expr to catch important cases late when
2961 range info is available.
2963 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2965 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
2966 * configure.ac (hidden_linkonce): New test.
2967 * configure: Regenerate.
2968 * config.in: Regenerate.
2970 2018-01-26 Julia Koval <julia.koval@intel.com>
2972 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
2973 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
2974 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
2975 _mm_mask_bitshuffle_epi64_mask): Fix type.
2976 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
2977 USI_FTYPE_V4DI_V4DI_USI): Remove.
2978 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
2979 __builtin_ia32_vpshufbitqmb256_mask,
2980 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
2981 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
2982 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
2984 2018-01-26 Alan Modra <amodra@gmail.com>
2987 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
2988 UNSPEC_VBPERMQ. Sort other unspecs.
2990 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
2992 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
2994 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
2997 * predict.c (drop_profile): Do not push/pop cfun; update also
2999 (handle_missing_profiles): Fix logic looking for zero profiles.
3001 2018-01-25 Jakub Jelinek <jakub@redhat.com>
3004 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
3005 on functions with #pragma omp declare simd or functions with simd
3007 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
3008 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
3009 Remove trailing \n from warning_at calls.
3011 2018-01-25 Tom de Vries <tom@codesourcery.com>
3014 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3015 for neutered workers.
3017 2018-01-24 Joseph Myers <joseph@codesourcery.com>
3020 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
3021 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
3023 2018-01-24 Jeff Law <law@redhat.com>
3026 * i386.c (get_probe_interval): Move to earlier point.
3027 (ix86_compute_frame_layout): If -fstack-clash-protection and
3028 the frame is larger than the probe interval, then use pushes
3029 to save registers rather than reg->mem moves.
3030 (ix86_expand_prologue): Remove conditional for int_registers_saved
3033 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
3036 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
3037 min/max for never referenced object.
3039 2018-01-24 Jakub Jelinek <jakub@redhat.com>
3042 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
3044 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
3045 attributes from DECL_ATTRIBUTES (decl) without affecting
3046 DECL_ATTRIBUTES (current_function_decl).
3047 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
3048 functions with non-NULL DECL_ABSTRACT_ORIGIN.
3050 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
3052 PR tree-optimization/83979
3053 * fold-const.c (fold_comparison): Use constant_boolean_node
3054 instead of boolean_{true,false}_node.
3056 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
3058 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
3061 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3063 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3064 Simplify the clause that sets the length attribute.
3065 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3066 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
3067 clause that sets the length attribute.
3068 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3070 2018-01-24 Tom de Vries <tom@codesourcery.com>
3073 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
3074 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
3075 Add strict parameter.
3076 (prevent_branch_around_nothing): Insert dummy insn between branch to
3077 label and label with no ptx insn inbetween.
3078 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
3080 2018-01-24 Tom de Vries <tom@codesourcery.com>
3083 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3084 for neutered threads in warp.
3085 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
3087 2018-01-24 Richard Biener <rguenther@suse.de>
3089 PR tree-optimization/83176
3090 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
3093 2018-01-24 Richard Biener <rguenther@suse.de>
3095 PR tree-optimization/82819
3096 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
3097 code generating pluses that are no-ops in the target precision.
3099 2018-01-24 Richard Biener <rguenther@suse.de>
3102 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
3104 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3106 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
3107 to merge probabilities.
3108 * predict.c (probably_never_executed): Also mark as cold functions
3109 with global 0 profile and guessed local profile.
3110 * profile-count.c (profile_probability::combine_with_count): New
3112 * profile-count.h (profile_probability::operator*,
3113 profile_probability::operator*=, profile_probability::operator/,
3114 profile_probability::operator/=): Reduce precision to adjusted
3115 and set value to guessed on contradictory divisions.
3116 (profile_probability::combine_with_freq): Remove.
3117 (profile_probability::combine_wiht_count): Declare.
3118 (profile_count::force_nonzero):: Set to adjusted.
3119 (profile_count::probability_in):: Set quality to adjusted.
3120 * tree-ssa-tail-merge.c (replace_block_by): Use
3123 2018-01-23 Andrew Waterman <andrew@sifive.com>
3124 Jim Wilson <jimw@sifive.com>
3126 * config/riscv/riscv.c (riscv_stack_boundary): New.
3127 (riscv_option_override): Set riscv_stack_boundary. Handle
3128 riscv_preferred_stack_boundary_arg.
3129 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
3130 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
3131 (STACK_BOUNDARY): Set to riscv_stack_boundary.
3132 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
3133 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
3134 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
3136 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
3139 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
3140 of struct ix86_frame.
3141 (ix86_expand_epilogue): Likewise. Add a local variable for
3142 the reg_save_offset field in struct ix86_frame.
3144 2018-01-23 Bin Cheng <bin.cheng@arm.com>
3146 PR tree-optimization/82604
3147 * tree-loop-distribution.c (enum partition_kind): New enum item
3148 PKIND_PARTIAL_MEMSET.
3149 (partition_builtin_p): Support above new enum item.
3150 (generate_code_for_partition): Ditto.
3151 (compute_access_range): Differentiate cases that equality can be
3152 proven at all loops, the innermost loops or no loops.
3153 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
3154 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
3155 (finalize_partitions, distribute_loop): Don't fuse partition of
3156 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
3157 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
3160 2018-01-23 Martin Liska <mliska@suse.cz>
3162 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
3163 order to ignore the predictor.
3164 (PRED_POLYMORPHIC_CALL): Likewise.
3165 (PRED_RECURSIVE_CALL): Likewise.
3167 2018-01-23 Martin Liska <mliska@suse.cz>
3169 * tree-profile.c (tree_profiling): Print function header to
3170 aware reader which function we are working on.
3171 * value-prof.c (gimple_find_values_to_profile): Do not print
3172 not interesting value histograms.
3174 2018-01-23 Martin Liska <mliska@suse.cz>
3176 * profile-count.h (enum profile_quality): Add
3177 profile_uninitialized as the first value. Do not number values
3178 as they are zero based.
3179 (profile_count::verify): Update sanity check.
3180 (profile_probability::verify): Likewise.
3182 2018-01-23 Nathan Sidwell <nathan@acm.org>
3184 * doc/invoke.texi (ffor-scope): Deprecate.
3186 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3188 PR tree-optimization/83510
3189 * domwalk.c (set_all_edges_as_executable): New function.
3190 (dom_walker::dom_walker): Convert bool param
3191 "skip_unreachable_blocks" to enum reachability. Move setup of
3192 edge flags to set_all_edges_as_executable and only do it when
3193 reachability is REACHABLE_BLOCKS.
3194 * domwalk.h (enum dom_walker::reachability): New enum.
3195 (dom_walker::dom_walker): Convert bool param
3196 "skip_unreachable_blocks" to enum reachability.
3197 (set_all_edges_as_executable): New decl.
3198 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
3199 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
3201 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
3202 but converting true to REACHABLE_BLOCKS.
3203 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
3205 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
3206 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
3207 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
3209 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
3210 if check_all_array_refs will be called.
3212 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3214 * tree.c (selftest::test_location_wrappers): Add more test
3217 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3219 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
3220 (selftest::test_bit_in_range): Likewise.
3222 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3225 * doc/sourcebuild.texi (vect_float): Say that the selector
3226 only describes the situation when -funsafe-math-optimizations is on.
3227 (vect_float_strict): Document.
3229 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3231 PR tree-optimization/83965
3232 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
3233 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
3234 instead of checking only for a reduction.
3235 (vect_recog_widen_sum_pattern): Likewise.
3237 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3239 * predict.c (probably_never_executed): Only use precise profile info.
3240 (compute_function_frequency): Skip after inlining hack since we now
3241 have quality checking.
3243 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3245 * profile-count.h (profile_probability::very_unlikely,
3246 profile_probability::unlikely, profile_probability::even): Set
3247 precision to guessed.
3249 2018-01-23 Richard Biener <rguenther@suse.de>
3251 PR tree-optimization/83963
3252 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
3253 Properly terminate dominator walk when crossing the exit edge not
3254 when visiting its source block.
3256 2018-01-23 Jakub Jelinek <jakub@redhat.com>
3259 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
3260 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
3262 2018-01-22 Jakub Jelinek <jakub@redhat.com>
3264 PR tree-optimization/83957
3265 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
3266 semicolon after for body surrounded by braces.
3268 PR tree-optimization/83081
3269 * profile-count.h (profile_probability::split): New method.
3270 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
3271 Use profile_probability::split.
3272 (do_compare_rtx_and_jump): Fix adjustment of probabilities
3273 when splitting a single conditional jump into 2.
3275 2018-01-22 David Malcolm <dmalcolm@redhat.com>
3277 PR tree-optimization/69452
3278 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
3281 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3283 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
3284 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
3285 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
3287 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3289 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
3290 * config/rl78/rl78.md (movdi): New define_expand.
3291 * config/rl78/rl78.c (rl78_split_movdi): New function.
3293 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
3296 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
3298 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
3299 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
3300 128-bit to produce an UNSPEC move to get the double word with the
3301 signbit and then a shift directly to do signbit.
3302 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
3303 implementation with a new version that just does either a direct
3304 move or a regular move. Move memory interface to separate insns.
3305 Move insns so they are next to the expander.
3306 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
3307 with signbit move. Split big and little endian case.
3308 (signbit<mode>2_dm_mem_le): Likewise.
3309 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
3310 (signbit<mode>2_dm2): Likewise.
3312 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3314 * config/rl78/rl78.md (anddi3): New define_expand.
3316 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3318 * config/rl78/rl78.md (umindi3): New define_expand.
3320 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3322 * config/rl78/rl78.md (smindi3): New define_expand.
3324 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3326 * config/rl78/rl78.md (smaxdi3): New define_expand.
3328 2018-01-22 Carl Love <cel@us.ibm.com>
3330 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
3331 LVX_V1TI): Add macro expansion.
3332 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
3333 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
3334 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
3335 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
3336 Change check to determine if the instruction is a byte reversing
3337 entry. Fix typo in comment.
3338 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
3339 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
3340 Add def_builtin calls for new builtins.
3341 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
3342 Add define_insn expansion.
3344 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3346 * config/rl78/rl78.md (umaxdi3): New define_expand.
3348 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3350 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
3351 for non-QImode registers.
3353 2018-01-22 Richard Biener <rguenther@suse.de>
3355 PR tree-optimization/83963
3356 * graphite-scop-detection.c (scop_detection::get_sese): Delay
3357 including the loop exit block.
3358 (scop_detection::merge_sese): Likewise.
3359 (scop_detection::add_scop): Do it here instead.
3361 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3363 * doc/sourcebuild.texi (arm_softfloat): Document.
3365 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
3368 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
3369 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
3370 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
3372 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3373 David Edelsohn <dje.gcc@gmail.com>
3376 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3377 Change "crset eq" to "crset 2".
3378 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3379 (*call_indirect_aix<mode>_nospec): Likewise.
3380 (*call_value_indirect_aix<mode>_nospec): Likewise.
3381 (*call_indirect_elfv2<mode>_nospec): Likewise.
3382 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
3383 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
3384 change assembly output from . to $.
3385 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3386 (indirect_jump<mode>_nospec): Change assembly output from . to $.
3387 (*tablejump<mode>_internal1_nospec): Likewise.
3389 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
3392 * config/sh/sh_optimize_sett_clrt.cc:
3393 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
3395 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
3397 PR tree-optimization/83940
3398 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
3399 offset_dt to vect_constant_def rather than vect_unknown_def_type.
3400 (vect_check_load_store_mask): Add a mask_dt_out parameter and
3401 use it to pass back the definition type.
3402 (vect_check_store_rhs): Likewise rhs_dt_out.
3403 (vect_build_gather_load_calls): Add a mask_dt argument and use
3404 it instead of a call to vect_is_simple_use.
3405 (vectorizable_store): Update calls to vect_check_load_store_mask
3406 and vect_check_store_rhs. Use the dt returned by the latter instead
3407 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
3408 instead of calls to vect_is_simple_use. Pass the scalar rather
3409 than the vector operand to vect_is_simple_use when handling
3410 second and subsequent copies of an rhs value.
3411 (vectorizable_load): Update calls to vect_check_load_store_mask
3412 and vect_build_gather_load_calls. Use the cached mask_dt and
3413 gs_info.offset_dt instead of calls to vect_is_simple_use.
3415 2018-01-20 Jakub Jelinek <jakub@redhat.com>
3418 * tree-emutls.c: Include gimplify.h.
3419 (lower_emutls_2): New function.
3420 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
3421 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
3422 it before further processing.
3425 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
3426 UINTVAL (trueop1) instead of INTVAL (op1).
3428 2018-01-19 Jakub Jelinek <jakub@redhat.com>
3432 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
3433 INCOMING_FRAME_SP_OFFSET if not defined.
3434 (scan_trace): Add ENTRY argument. If true and
3435 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
3436 emit a note to adjust the CFA offset.
3437 (create_cfi_notes): Adjust scan_trace callers.
3438 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
3439 INCOMING_FRAME_SP_OFFSET in the CIE.
3440 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
3441 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
3443 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
3444 * doc/tm.texi: Regenerated.
3446 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3448 PR rtl-optimization/83147
3449 * lra-constraints.c (remove_inheritance_pseudos): Use
3450 lra_substitute_pseudo_within_insn.
3452 2018-01-19 Tom de Vries <tom@codesourcery.com>
3453 Cesar Philippidis <cesar@codesourcery.com>
3456 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
3458 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
3461 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
3462 spaces for function labels.
3464 2018-01-19 Martin Liska <mliska@suse.cz>
3466 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
3467 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
3468 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
3469 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
3470 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
3471 (PRED_CONST_RETURN): Change from 69 to 65.
3472 (PRED_NULL_RETURN): Change from 91 to 71.
3473 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
3474 (PRED_LOOP_GUARD): Change from 66 to 73.
3476 2018-01-19 Martin Liska <mliska@suse.cz>
3478 * predict.c (predict_insn_def): Add new assert.
3479 (struct branch_predictor): Change type to signed integer.
3480 (test_prediction_value_range): Amend test to cover
3482 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
3483 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
3484 (PRED_LOOP_ITERATIONS_MAX): Likewise.
3485 (PRED_LOOP_IV_COMPARE): Likewise.
3486 * predict.h (PROB_UNINITIALIZED): Define new constant.
3488 2018-01-19 Martin Liska <mliska@suse.cz>
3490 * predict.c (dump_prediction): Add new format for
3491 analyze_brprob.py script which is enabled with -details
3493 * profile-count.h (precise_p): New function.
3495 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3497 PR tree-optimization/83922
3498 * tree-vect-loop.c (vect_verify_full_masking): Return false if
3499 there are no statements that need masking.
3500 (vect_active_double_reduction_p): New function.
3501 (vect_analyze_loop_operations): Use it when handling phis that
3502 are not in the loop header.
3504 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3506 PR tree-optimization/83914
3507 * tree-vect-loop.c (vectorizable_induction): Don't convert
3508 init_expr or apply the peeling adjustment for inductions
3509 that are nested within the vectorized loop.
3511 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3513 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
3516 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3520 * function.h (gimplify_parameters): Add gimple_seq * argument.
3521 * function.c: Include gimple.h and options.h.
3522 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
3523 for the added local temporaries if needed.
3524 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
3525 if there are any parameter cleanups, wrap whole body into a
3526 try/finally with the cleanups.
3528 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
3531 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
3532 Use GET_MODE_CLASS for scalar floating point.
3534 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3538 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
3539 Fix call of call_cgraph_insertion_hooks.
3541 2018-01-18 Martin Sebor <msebor@redhat.com>
3543 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
3545 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3548 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
3551 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
3554 * common.opt: (-ffile-prefix-map): New option.
3555 * opts.c (common_handle_option): Defer it.
3556 * opts-global.c (handle_common_deferred_options): Handle it.
3557 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
3558 * file-prefix-map.h: New file.
3559 (remap_debug_filename, add_debug_prefix_map): ...here.
3560 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
3561 * final.c (debug_prefix_map, add_debug_prefix_map
3562 remap_debug_filename): Move to...
3563 * file-prefix-map.c: New file.
3564 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
3565 generalize, get rid of alloca(), use strrchr() instead of strchr().
3566 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
3567 Implement in terms of add_prefix_map().
3568 (remap_macro_filename, remap_debug_filename): Implement in term of
3570 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
3571 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
3572 * dbxout.c: Include file-prefix-map.h.
3573 * varasm.c: Likewise.
3574 * vmsdbgout.c: Likewise.
3575 * xcoffout.c: Likewise.
3576 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
3577 * doc/cppopts.texi (-fmacro-prefix-map): Document.
3578 * doc/invoke.texi (-ffile-prefix-map): Document.
3579 (-fdebug-prefix-map): Update description.
3581 2018-01-18 Martin Liska <mliska@suse.cz>
3583 * config/i386/i386.c (indirect_thunk_name): Document that also
3585 (output_indirect_thunk): Document why both instructions
3586 (pause and lfence) are generated.
3588 2018-01-18 Richard Biener <rguenther@suse.de>
3590 PR tree-optimization/83887
3591 * graphite-scop-detection.c
3592 (scop_detection::get_nearest_dom_with_single_entry): Remove.
3593 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
3594 (scop_detection::merge_sese): Re-implement with a flood-fill
3595 algorithm that properly finds a SESE region if it exists.
3597 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3600 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
3601 pointer_diff optimizations use view_convert instead of convert.
3603 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3605 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3606 Generate different code for -mno-speculate-indirect-jumps.
3607 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3608 (*call_indirect_aix<mode>): Disable for
3609 -mno-speculate-indirect-jumps.
3610 (*call_indirect_aix<mode>_nospec): New define_insn.
3611 (*call_value_indirect_aix<mode>): Disable for
3612 -mno-speculate-indirect-jumps.
3613 (*call_value_indirect_aix<mode>_nospec): New define_insn.
3614 (*sibcall_nonlocal_sysv<mode>): Generate different code for
3615 -mno-speculate-indirect-jumps.
3616 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3618 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3620 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
3621 long double type, set the flags for noting the default long double
3622 type, even if we don't pass or return a long double type.
3624 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
3627 * ipa-inline.c (flatten_function): Do not overwrite final inlining
3630 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
3632 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
3633 support for merge[hl].
3634 (fold_mergehl_helper): New helper function.
3635 (tree-vector-builder.h): New #include for tree_vector_builder usage.
3636 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
3637 (altivec_vmrglw_direct): Add xxmrglw insn.
3639 2018-01-17 Andrew Waterman <andrew@sifive.com>
3641 * config/riscv/riscv.c (riscv_conditional_register_usage): If
3642 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
3644 2018-01-17 David Malcolm <dmalcolm@redhat.com>
3647 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
3648 call the lto_location_cache before reading the
3649 DECL_SOURCE_LOCATION of the types.
3651 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3652 Richard Sandiford <richard.sandiford@linaro.org>
3654 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
3655 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
3656 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
3657 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
3658 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
3660 * config/aarch64/constraints.md (aarch64_movti_operand):
3662 * config/aarch64/predicates.md (Uti): Add new constraint.
3664 2018-01-17 Carl Love <cel@us.ibm.com>
3666 * config/rs6000/vsx.md (define_expand xl_len_r,
3667 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
3668 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
3670 (define_expand, define_insn): Move the shift left from the
3671 define_insn to the define_expand for lxvl and stxvl instructions.
3672 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
3673 and XL_LEN_R definitions to PURE.
3675 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
3677 * config/i386/i386.c (indirect_thunk_name): Declare regno
3678 as unsigned int. Compare regno with INVALID_REGNUM.
3679 (output_indirect_thunk): Ditto.
3680 (output_indirect_thunk_function): Ditto.
3681 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
3682 in the call to output_indirect_thunk_function.
3684 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
3687 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
3688 rather than the size of inner_type to determine the stack slot size
3689 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
3691 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
3694 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
3697 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
3699 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
3700 endian Linux systems to optionally enable multilibs for selecting
3701 the long double type if the user configured an explicit type.
3702 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
3703 have no long double multilibs if not defined.
3704 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
3705 warn if the user used -mabi={ieee,ibm}longdouble and we built
3706 multilibs for long double.
3707 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
3708 appropriate multilib option.
3709 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
3711 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
3712 for building long double multilibs.
3713 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
3715 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
3717 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
3720 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
3722 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
3725 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
3728 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
3731 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3733 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
3734 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
3736 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
3738 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
3739 different rtl trees depending on TARGET_64BIT.
3740 (rs6000_gen_lvx): Likewise.
3742 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3744 * config/visium/visium.md (nop): Tweak comment.
3745 (hazard_nop): Likewise.
3747 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3749 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
3750 -mspeculate-indirect-jumps.
3751 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
3752 for -mno-speculate-indirect-jumps.
3753 (*call_indirect_elfv2<mode>_nospec): New define_insn.
3754 (*call_value_indirect_elfv2<mode>): Disable for
3755 -mno-speculate-indirect-jumps.
3756 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
3757 (indirect_jump): Emit different RTL for
3758 -mno-speculate-indirect-jumps.
3759 (*indirect_jump<mode>): Disable for
3760 -mno-speculate-indirect-jumps.
3761 (*indirect_jump<mode>_nospec): New define_insn.
3762 (tablejump): Emit different RTL for
3763 -mno-speculate-indirect-jumps.
3764 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
3765 (tablejumpsi_nospec): New define_expand.
3766 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
3767 (tablejumpdi_nospec): New define_expand.
3768 (*tablejump<mode>_internal1): Disable for
3769 -mno-speculate-indirect-jumps.
3770 (*tablejump<mode>_internal1_nospec): New define_insn.
3771 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
3774 2018-01-16 Artyom Skrobov tyomitch@gmail.com
3776 * caller-save.c (insert_save): Drop unnecessary parameter. All
3779 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3780 Richard Biener <rguenth@suse.de>
3783 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
3784 return early, inline manually is_gimple_sizepos. Make sure if we
3785 call gimplify_expr we don't end up with a gimple constant.
3786 * tree.c (variably_modified_type_p): Don't return true for
3787 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
3788 * gimplify.h (is_gimple_sizepos): Remove.
3790 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3792 PR tree-optimization/83857
3793 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
3794 vectorizable_live_operation for pure SLP statements.
3795 (vectorizable_live_operation): Handle PHIs.
3797 2018-01-16 Richard Biener <rguenther@suse.de>
3799 PR tree-optimization/83867
3800 * tree-vect-stmts.c (vect_transform_stmt): Precompute
3801 nested_in_vect_loop_p since the scalar stmt may get invalidated.
3803 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3806 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
3807 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
3808 If off is not INTEGER_CST, issue a may not be aligned warning
3809 rather than isn't aligned. Use isn%'t rather than isn't.
3810 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
3812 <case MULT_EXPR>: Improve the case when bottom and one of the
3813 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
3814 operand, in that case check if the other operand is multiple of
3815 bottom divided by the INTEGER_CST operand.
3817 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3820 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
3821 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
3822 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
3823 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
3824 * config/pa/pa.c (pa_function_arg_advance): Likewise.
3825 (pa_function_arg, pa_arg_partial_bytes): Likewise.
3826 (pa_function_arg_size): New function.
3828 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3830 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
3831 in a separate statement.
3833 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3835 PR tree-optimization/83847
3836 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
3837 group gathers and scatters.
3839 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3841 PR rtl-optimization/86620
3842 * params.def (max-sched-ready-insns): Bump minimum value to 1.
3844 PR rtl-optimization/83213
3845 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
3846 to last if both are JUMP_INSNs.
3848 PR tree-optimization/83843
3849 * gimple-ssa-store-merging.c
3850 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
3851 store_immediate_info for bswap/nop orig_stores.
3853 2018-01-15 Andrew Waterman <andrew@sifive.com>
3855 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
3857 <UDIV>: Increase cost if !TARGET_DIV.
3859 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
3861 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
3862 (define_attr "cr_logical_3op"): New.
3863 (cceq_ior_compare): Adjust.
3864 (cceq_ior_compare_complement): Adjust.
3865 (*cceq_rev_compare): Adjust.
3866 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
3867 (is_cracked_insn): Adjust.
3868 (insn_must_be_first_in_group): Adjust.
3869 * config/rs6000/40x.md: Adjust.
3870 * config/rs6000/440.md: Adjust.
3871 * config/rs6000/476.md: Adjust.
3872 * config/rs6000/601.md: Adjust.
3873 * config/rs6000/603.md: Adjust.
3874 * config/rs6000/6xx.md: Adjust.
3875 * config/rs6000/7450.md: Adjust.
3876 * config/rs6000/7xx.md: Adjust.
3877 * config/rs6000/8540.md: Adjust.
3878 * config/rs6000/cell.md: Adjust.
3879 * config/rs6000/e300c2c3.md: Adjust.
3880 * config/rs6000/e500mc.md: Adjust.
3881 * config/rs6000/e500mc64.md: Adjust.
3882 * config/rs6000/e5500.md: Adjust.
3883 * config/rs6000/e6500.md: Adjust.
3884 * config/rs6000/mpc.md: Adjust.
3885 * config/rs6000/power4.md: Adjust.
3886 * config/rs6000/power5.md: Adjust.
3887 * config/rs6000/power6.md: Adjust.
3888 * config/rs6000/power7.md: Adjust.
3889 * config/rs6000/power8.md: Adjust.
3890 * config/rs6000/power9.md: Adjust.
3891 * config/rs6000/rs64.md: Adjust.
3892 * config/rs6000/titan.md: Adjust.
3894 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3896 * config/i386/predicates.md (indirect_branch_operand): Rewrite
3897 ix86_indirect_branch_register logic.
3899 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3901 * config/i386/constraints.md (Bs): Update
3902 ix86_indirect_branch_register check. Don't check
3903 ix86_indirect_branch_register with GOT_memory_operand.
3905 * config/i386/predicates.md (GOT_memory_operand): Don't check
3906 ix86_indirect_branch_register here.
3907 (GOT32_symbol_operand): Likewise.
3909 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3911 * config/i386/predicates.md (constant_call_address_operand):
3912 Rewrite ix86_indirect_branch_register logic.
3913 (sibcall_insn_operand): Likewise.
3915 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3917 * config/i386/constraints.md (Bs): Replace
3918 ix86_indirect_branch_thunk_register with
3919 ix86_indirect_branch_register.
3921 * config/i386/i386.md (indirect_jump): Likewise.
3922 (tablejump): Likewise.
3923 (*sibcall_memory): Likewise.
3924 (*sibcall_value_memory): Likewise.
3925 Peepholes of indirect call and jump via memory: Likewise.
3926 * config/i386/i386.opt: Likewise.
3927 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3928 (GOT_memory_operand): Likewise.
3929 (call_insn_operand): Likewise.
3930 (sibcall_insn_operand): Likewise.
3931 (GOT32_symbol_operand): Likewise.
3933 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3936 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
3937 type rather than type addr's type points to.
3938 (expand_omp_atomic_mutex): Likewise.
3939 (expand_omp_atomic): Likewise.
3941 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3944 * config/i386/i386.c (output_indirect_thunk_function): Use
3945 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
3946 for __x86_return_thunk.
3948 2018-01-15 Richard Biener <rguenther@suse.de>
3951 * expmed.c (extract_bit_field_1): Fix typo.
3953 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3956 * config/arm/iterators.md (VF): New mode iterator.
3957 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
3958 Remove integer-related logic from pattern.
3959 (neon_vabd<mode>_3): Likewise.
3961 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3964 * common.opt (fstrict-overflow): No longer an alias.
3965 (fwrapv-pointer): New option.
3966 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
3967 also for pointer types based on flag_wrapv_pointer.
3968 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
3969 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
3970 opts->x_flag_wrapv got set.
3971 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
3972 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
3973 POINTER_TYPE_OVERFLOW_UNDEFINED.
3974 * match.pd: Likewise in address comparison pattern.
3975 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
3977 2018-01-15 Richard Biener <rguenther@suse.de>
3980 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
3981 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
3982 Reset type names to their identifier if their TYPE_DECL doesn't
3983 have linkage (and thus is used for ODR and devirt).
3984 (save_debug_info_for_decl): Remove.
3985 (save_debug_info_for_type): Likewise.
3986 (add_tree_to_fld_list): Adjust.
3987 * tree-pretty-print.c (dump_generic_node): Make dumping of
3988 type names more robust.
3990 2018-01-15 Richard Biener <rguenther@suse.de>
3992 * BASE-VER: Bump to 8.0.1.
3994 2018-01-14 Martin Sebor <msebor@redhat.com>
3997 * builtins.c (check_access): Avoid warning when the no-warning bit
4000 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
4002 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
4003 * ira-color (allocno_hard_regs_compare): Likewise.
4005 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
4008 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
4009 Use .pushsection/.popsection.
4011 2018-01-14 Martin Sebor <msebor@redhat.com>
4014 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
4016 2018-01-14 Jakub Jelinek <jakub@redhat.com>
4018 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
4019 entry from extra_headers.
4020 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
4021 extra_headers, make the list bitwise identical to the i?86-*-* one.
4023 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4025 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
4026 -mcmodel=large with -mindirect-branch=thunk,
4027 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
4028 -mfunction-return=thunk-extern.
4029 * doc/invoke.texi: Document -mcmodel=large is incompatible with
4030 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
4031 -mfunction-return=thunk and -mfunction-return=thunk-extern.
4033 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4035 * config/i386/i386.c (print_reg): Print the name of the full
4036 integer register without '%'.
4037 (ix86_print_operand): Handle 'V'.
4038 * doc/extend.texi: Document 'V' modifier.
4040 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4042 * config/i386/constraints.md (Bs): Disallow memory operand for
4043 -mindirect-branch-register.
4045 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4046 (GOT_memory_operand): Likewise.
4047 (call_insn_operand): Likewise.
4048 (sibcall_insn_operand): Likewise.
4049 (GOT32_symbol_operand): Likewise.
4050 * config/i386/i386.md (indirect_jump): Call convert_memory_address
4051 for -mindirect-branch-register.
4052 (tablejump): Likewise.
4053 (*sibcall_memory): Likewise.
4054 (*sibcall_value_memory): Likewise.
4055 Disallow peepholes of indirect call and jump via memory for
4056 -mindirect-branch-register.
4057 (*call_pop): Replace m with Bw.
4058 (*call_value_pop): Likewise.
4059 (*sibcall_pop_memory): Replace m with Bs.
4060 * config/i386/i386.opt (mindirect-branch-register): New option.
4061 * doc/invoke.texi: Document -mindirect-branch-register option.
4063 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4065 * config/i386/i386-protos.h (ix86_output_function_return): New.
4066 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
4067 set function_return_type.
4068 (indirect_thunk_name): Add ret_p to indicate thunk for function
4070 (output_indirect_thunk_function): Pass false to
4071 indirect_thunk_name.
4072 (ix86_output_indirect_branch_via_reg): Likewise.
4073 (ix86_output_indirect_branch_via_push): Likewise.
4074 (output_indirect_thunk_function): Create alias for function
4075 return thunk if regno < 0.
4076 (ix86_output_function_return): New function.
4077 (ix86_handle_fndecl_attribute): Handle function_return.
4078 (ix86_attribute_table): Add function_return.
4079 * config/i386/i386.h (machine_function): Add
4080 function_return_type.
4081 * config/i386/i386.md (simple_return_internal): Use
4082 ix86_output_function_return.
4083 (simple_return_internal_long): Likewise.
4084 * config/i386/i386.opt (mfunction-return=): New option.
4085 (indirect_branch): Mention -mfunction-return=.
4086 * doc/extend.texi: Document function_return function attribute.
4087 * doc/invoke.texi: Document -mfunction-return= option.
4089 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4091 * config/i386/i386-opts.h (indirect_branch): New.
4092 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
4093 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
4094 with local indirect jump when converting indirect call and jump.
4095 (ix86_set_indirect_branch_type): New.
4096 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
4097 (indirectlabelno): New.
4098 (indirect_thunk_needed): Likewise.
4099 (indirect_thunk_bnd_needed): Likewise.
4100 (indirect_thunks_used): Likewise.
4101 (indirect_thunks_bnd_used): Likewise.
4102 (INDIRECT_LABEL): Likewise.
4103 (indirect_thunk_name): Likewise.
4104 (output_indirect_thunk): Likewise.
4105 (output_indirect_thunk_function): Likewise.
4106 (ix86_output_indirect_branch_via_reg): Likewise.
4107 (ix86_output_indirect_branch_via_push): Likewise.
4108 (ix86_output_indirect_branch): Likewise.
4109 (ix86_output_indirect_jmp): Likewise.
4110 (ix86_code_end): Call output_indirect_thunk_function if needed.
4111 (ix86_output_call_insn): Call ix86_output_indirect_branch if
4113 (ix86_handle_fndecl_attribute): Handle indirect_branch.
4114 (ix86_attribute_table): Add indirect_branch.
4115 * config/i386/i386.h (machine_function): Add indirect_branch_type
4116 and has_local_indirect_jump.
4117 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
4119 (tablejump): Likewise.
4120 (*indirect_jump): Use ix86_output_indirect_jmp.
4121 (*tablejump_1): Likewise.
4122 (simple_return_indirect_internal): Likewise.
4123 * config/i386/i386.opt (mindirect-branch=): New option.
4124 (indirect_branch): New.
4127 (thunk-inline): Likewise.
4128 (thunk-extern): Likewise.
4129 * doc/extend.texi: Document indirect_branch function attribute.
4130 * doc/invoke.texi: Document -mindirect-branch= option.
4132 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
4135 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
4137 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
4139 * ipa-inline.c (want_inline_small_function_p): Return false if
4140 inlining has already failed with CIF_FINAL_ERROR.
4141 (update_caller_keys): Call want_inline_small_function_p before
4143 (update_callee_keys): Likewise.
4145 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4147 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
4149 (rs6000_quadword_masked_address_p): Likewise.
4150 (quad_aligned_load_p): Likewise.
4151 (quad_aligned_store_p): Likewise.
4152 (const_load_sequence_p): Add comment to describe the outer-most loop.
4153 (mimic_memory_attributes_and_flags): New function.
4154 (rs6000_gen_stvx): Likewise.
4155 (replace_swapped_aligned_store): Likewise.
4156 (rs6000_gen_lvx): Likewise.
4157 (replace_swapped_aligned_load): Likewise.
4158 (replace_swapped_load_constant): Capitalize argument name in
4159 comment describing this function.
4160 (rs6000_analyze_swaps): Add a third pass to search for vector loads
4161 and stores that access quad-word aligned addresses and replace
4162 with stvx or lvx instructions when appropriate.
4163 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
4164 New function prototype.
4165 (rs6000_quadword_masked_address_p): Likewise.
4166 (rs6000_gen_lvx): Likewise.
4167 (rs6000_gen_stvx): Likewise.
4168 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
4169 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
4170 when memory address is aligned.
4171 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
4172 this split to select lvx instruction when memory address is aligned.
4173 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
4174 instruction when memory address is aligned.
4175 (*vsx_le_perm_load_v16qi): Likewise.
4176 (four unnamed splitters): Modify to select the stvx instruction
4177 when memory is aligned.
4179 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
4181 * predict.c (determine_unlikely_bbs): Handle correctly BBs
4182 which appears in the queue multiple times.
4184 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4185 Alan Hayward <alan.hayward@arm.com>
4186 David Sherwood <david.sherwood@arm.com>
4188 * tree-vectorizer.h (vec_lower_bound): New structure.
4189 (_loop_vec_info): Add check_nonzero and lower_bounds.
4190 (LOOP_VINFO_CHECK_NONZERO): New macro.
4191 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
4192 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
4193 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
4194 fields. Make seg_len the distance travelled, not including the
4196 (dr_direction_indicator): Declare.
4197 (dr_zero_step_indicator): Likewise.
4198 (dr_known_forward_stride_p): Likewise.
4199 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
4201 (runtime_alias_check_p): Allow runtime alias checks with
4203 (operator ==): Compare access_size and align.
4204 (prune_runtime_alias_test_list): Rework for new distinction between
4205 the access_size and seg_len.
4206 (create_intersect_range_checks_index): Likewise. Cope with polynomial
4208 (get_segment_min_max): New function.
4209 (create_intersect_range_checks): Use it.
4210 (dr_step_indicator): New function.
4211 (dr_direction_indicator): Likewise.
4212 (dr_zero_step_indicator): Likewise.
4213 (dr_known_forward_stride_p): Likewise.
4214 * tree-loop-distribution.c (data_ref_segment_size): Return
4215 DR_STEP * (niters - 1).
4216 (compute_alias_check_pairs): Update call to the dr_with_seg_len
4218 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
4219 (vect_preserves_scalar_order_p): New function, split out from...
4220 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
4221 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
4222 (vect_vfa_access_size): New function.
4223 (vect_vfa_align): Likewise.
4224 (vect_compile_time_alias): Take access_size_a and access_b arguments.
4225 (dump_lower_bound): New function.
4226 (vect_check_lower_bound): Likewise.
4227 (vect_small_gap_p): Likewise.
4228 (vectorizable_with_step_bound_p): Likewise.
4229 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
4230 depencies if the vectorization factor is 1. Convert the checks
4231 for nonzero steps into checks on the bounds of DR_STEP. Try using
4232 a bunds check for variable steps if the minimum required step is
4233 relatively small. Update calls to the dr_with_seg_len
4234 constructor and to vect_compile_time_alias.
4235 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
4237 (vect_loop_versioning): Call it.
4238 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
4240 (vect_estimate_min_profitable_iters): Account for any bounds checks.
4242 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4243 Alan Hayward <alan.hayward@arm.com>
4244 David Sherwood <david.sherwood@arm.com>
4246 * doc/sourcebuild.texi (vect_scatter_store): Document.
4247 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
4249 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
4251 * genopinit.c (main): Add supports_vec_scatter_store and
4252 supports_vec_scatter_store_cached to target_optabs.
4253 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
4254 IFN_MASK_SCATTER_STORE.
4255 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
4257 * internal-fn.h (internal_store_fn_p): Declare.
4258 (internal_fn_stored_value_index): Likewise.
4259 * internal-fn.c (scatter_store_direct): New macro.
4260 (expand_scatter_store_optab_fn): New function.
4261 (direct_scatter_store_optab_supported_p): New macro.
4262 (internal_store_fn_p): New function.
4263 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
4264 IFN_MASK_SCATTER_STORE.
4265 (internal_fn_mask_index): Likewise.
4266 (internal_fn_stored_value_index): New function.
4267 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
4269 * optabs-query.h (supports_vec_scatter_store_p): Declare.
4270 * optabs-query.c (supports_vec_scatter_store_p): New function.
4271 * tree-vectorizer.h (vect_get_store_rhs): Declare.
4272 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
4273 true for scatter stores.
4274 (vect_gather_scatter_fn_p): Handle scatter stores too.
4275 (vect_check_gather_scatter): Consider using scatter stores if
4276 supports_vec_scatter_store_p.
4277 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
4279 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4280 internal_fn_stored_value_index.
4281 (check_load_store_masking): Handle scatter stores too.
4282 (vect_get_store_rhs): Make public.
4283 (vectorizable_call): Use internal_store_fn_p.
4284 (vectorizable_store): Handle scatter store internal functions.
4285 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
4286 when deciding whether the end of the group has been reached.
4287 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
4288 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
4289 (mask_scatter_store<mode>): New insns.
4291 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4292 Alan Hayward <alan.hayward@arm.com>
4293 David Sherwood <david.sherwood@arm.com>
4295 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
4296 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
4297 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
4299 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
4300 Use vect_truncate_gather_scatter_offset if we can't treat the
4301 operation as a normal gather load or scatter store.
4302 (get_group_load_store_type): Take the gather_scatter_info
4303 as argument. Try using a gather load or scatter store for
4304 single-element groups.
4305 (get_load_store_type): Update calls to get_group_load_store_type
4306 and vect_use_strided_gather_scatters_p.
4308 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4309 Alan Hayward <alan.hayward@arm.com>
4310 David Sherwood <david.sherwood@arm.com>
4312 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
4313 optional tree argument.
4314 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
4316 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
4317 but continue to use the current value as a fallback.
4318 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
4319 to compare the updates.
4320 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
4321 (get_load_store_type): Use it when handling a strided access.
4322 (vect_get_strided_load_store_ops): New function.
4323 (vect_get_data_ptr_increment): Likewise.
4324 (vectorizable_load): Handle strided gather loads. Always pass
4325 a step to vect_create_data_ref_ptr and bump_vector_ptr.
4327 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4328 Alan Hayward <alan.hayward@arm.com>
4329 David Sherwood <david.sherwood@arm.com>
4331 * doc/md.texi (gather_load@var{m}): Document.
4332 (mask_gather_load@var{m}): Likewise.
4333 * genopinit.c (main): Add supports_vec_gather_load and
4334 supports_vec_gather_load_cached to target_optabs.
4335 * optabs-tree.c (init_tree_optimization_optabs): Use
4336 ggc_cleared_alloc to allocate target_optabs.
4337 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
4338 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
4340 * internal-fn.h (internal_load_fn_p): Declare.
4341 (internal_gather_scatter_fn_p): Likewise.
4342 (internal_fn_mask_index): Likewise.
4343 (internal_gather_scatter_fn_supported_p): Likewise.
4344 * internal-fn.c (gather_load_direct): New macro.
4345 (expand_gather_load_optab_fn): New function.
4346 (direct_gather_load_optab_supported_p): New macro.
4347 (direct_internal_fn_optab): New function.
4348 (internal_load_fn_p): Likewise.
4349 (internal_gather_scatter_fn_p): Likewise.
4350 (internal_fn_mask_index): Likewise.
4351 (internal_gather_scatter_fn_supported_p): Likewise.
4352 * optabs-query.c (supports_at_least_one_mode_p): New function.
4353 (supports_vec_gather_load_p): Likewise.
4354 * optabs-query.h (supports_vec_gather_load_p): Declare.
4355 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
4356 and memory_type field.
4357 (NUM_PATTERNS): Bump to 15.
4358 * tree-vect-data-refs.c: Include internal-fn.h.
4359 (vect_gather_scatter_fn_p): New function.
4360 (vect_describe_gather_scatter_call): Likewise.
4361 (vect_check_gather_scatter): Try using internal functions for
4362 gather loads. Recognize existing calls to a gather load function.
4363 (vect_analyze_data_refs): Consider using gather loads if
4364 supports_vec_gather_load_p.
4365 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
4366 (vect_get_gather_scatter_offset_type): Likewise.
4367 (vect_convert_mask_for_vectype): Likewise.
4368 (vect_add_conversion_to_patterm): Likewise.
4369 (vect_try_gather_scatter_pattern): Likewise.
4370 (vect_recog_gather_scatter_pattern): New pattern recognizer.
4371 (vect_vect_recog_func_ptrs): Add it.
4372 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4373 internal_fn_mask_index and internal_gather_scatter_fn_p.
4374 (check_load_store_masking): Take the gather_scatter_info as an
4375 argument and handle gather loads.
4376 (vect_get_gather_scatter_ops): New function.
4377 (vectorizable_call): Check internal_load_fn_p.
4378 (vectorizable_load): Likewise. Handle gather load internal
4380 (vectorizable_store): Update call to check_load_store_masking.
4381 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
4382 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
4383 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
4384 (aarch64_gather_scale_operand_d): New predicates.
4385 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
4386 (mask_gather_load<mode>): New insns.
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 * optabs.def (fold_left_plus_optab): New optab.
4393 * doc/md.texi (fold_left_plus_@var{m}): Document.
4394 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
4395 * internal-fn.c (fold_left_direct): Define.
4396 (expand_fold_left_optab_fn): Likewise.
4397 (direct_fold_left_optab_supported_p): Likewise.
4398 * fold-const-call.c (fold_const_fold_left): New function.
4399 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
4400 * tree-parloops.c (valid_reduction_p): New function.
4401 (gather_scalar_reductions): Use it.
4402 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
4403 (vect_finish_replace_stmt): Declare.
4404 * tree-vect-loop.c (fold_left_reduction_fn): New function.
4405 (needs_fold_left_reduction_p): New function, split out from...
4406 (vect_is_simple_reduction): ...here. Accept reductions that
4407 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
4408 (vect_force_simple_reduction): Also store the reduction type in
4409 the assignment's STMT_VINFO_REDUC_TYPE.
4410 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
4411 (merge_with_identity): New function.
4412 (vect_expand_fold_left): Likewise.
4413 (vectorize_fold_left_reduction): Likewise.
4414 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
4415 scalar phi in place for it. Check for target support and reject
4416 cases that would reassociate the operation. Defer the transform
4417 phase to vectorize_fold_left_reduction.
4418 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
4419 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
4420 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
4422 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4424 * tree-if-conv.c (predicate_mem_writes): Remove redundant
4425 call to ifc_temp_var.
4427 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4428 Alan Hayward <alan.hayward@arm.com>
4429 David Sherwood <david.sherwood@arm.com>
4431 * target.def (legitimize_address_displacement): Take the original
4432 offset as a poly_int.
4433 * targhooks.h (default_legitimize_address_displacement): Update
4435 * targhooks.c (default_legitimize_address_displacement): Likewise.
4436 * doc/tm.texi: Regenerate.
4437 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
4438 as an argument, moving assert of ad->disp == ad->disp_term to...
4439 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
4440 Try calling targetm.legitimize_address_displacement before expanding
4441 the address rather than afterwards, and adjust for the new interface.
4442 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
4443 Match the new hook interface. Handle SVE addresses.
4444 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
4447 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4449 * Makefile.in (OBJS): Add early-remat.o.
4450 * target.def (select_early_remat_modes): New hook.
4451 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
4452 * doc/tm.texi: Regenerate.
4453 * targhooks.h (default_select_early_remat_modes): Declare.
4454 * targhooks.c (default_select_early_remat_modes): New function.
4455 * timevar.def (TV_EARLY_REMAT): New timevar.
4456 * passes.def (pass_early_remat): New pass.
4457 * tree-pass.h (make_pass_early_remat): Declare.
4458 * early-remat.c: New file.
4459 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
4461 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
4463 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4464 Alan Hayward <alan.hayward@arm.com>
4465 David Sherwood <david.sherwood@arm.com>
4467 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
4468 vfm1 with a bound_epilog parameter.
4469 (vect_do_peeling): Update calls accordingly, and move the prologue
4470 call earlier in the function. Treat the base bound_epilog as 0 for
4471 fully-masked loops and retain vf - 1 for other loops. Add 1 to
4472 this base when peeling for gaps.
4473 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
4474 with fully-masked loops.
4475 (vect_estimate_min_profitable_iters): Handle the single peeled
4476 iteration in that case.
4478 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4479 Alan Hayward <alan.hayward@arm.com>
4480 David Sherwood <david.sherwood@arm.com>
4482 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
4483 single-element interleaving even if the size is not a power of 2.
4484 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
4485 accesses for single-element interleaving if the group size is
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 * doc/md.texi (fold_extract_last_@var{m}): Document.
4493 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
4494 * optabs.def (fold_extract_last_optab): New optab.
4495 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
4496 * internal-fn.c (fold_extract_direct): New macro.
4497 (expand_fold_extract_optab_fn): Likewise.
4498 (direct_fold_extract_optab_supported_p): Likewise.
4499 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
4500 * tree-vect-loop.c (vect_model_reduction_cost): Handle
4501 EXTRACT_LAST_REDUCTION.
4502 (get_initial_def_for_reduction): Do not create an initial vector
4503 for EXTRACT_LAST_REDUCTION reductions.
4504 (vectorizable_reduction): Leave the scalar phi in place for
4505 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
4506 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
4507 epilogue code for EXTRACT_LAST_REDUCTION and defer the
4508 transform phase to vectorizable_condition.
4509 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
4511 (vect_finish_stmt_generation): ...here.
4512 (vect_finish_replace_stmt): New function.
4513 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
4514 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
4516 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
4518 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4519 Alan Hayward <alan.hayward@arm.com>
4520 David Sherwood <david.sherwood@arm.com>
4522 * doc/md.texi (extract_last_@var{m}): Document.
4523 * optabs.def (extract_last_optab): New optab.
4524 * internal-fn.def (EXTRACT_LAST): New internal function.
4525 * internal-fn.c (cond_unary_direct): New macro.
4526 (expand_cond_unary_optab_fn): Likewise.
4527 (direct_cond_unary_optab_supported_p): Likewise.
4528 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
4529 loops using EXTRACT_LAST.
4530 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
4531 (extract_last_<mode>): ...this optab.
4532 (vec_extract<mode><Vel>): Update accordingly.
4534 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4535 Alan Hayward <alan.hayward@arm.com>
4536 David Sherwood <david.sherwood@arm.com>
4538 * target.def (empty_mask_is_expensive): New hook.
4539 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
4540 * doc/tm.texi: Regenerate.
4541 * targhooks.h (default_empty_mask_is_expensive): Declare.
4542 * targhooks.c (default_empty_mask_is_expensive): New function.
4543 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
4544 if the target says that empty masks are expensive.
4545 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
4547 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
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 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
4554 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
4555 (vect_use_loop_mask_for_alignment_p): New function.
4556 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
4557 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
4558 niters_skip argument. Make sure that the first niters_skip elements
4559 of the first iteration are inactive.
4560 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
4561 Update call to vect_set_loop_masks_directly.
4562 (get_misalign_in_elems): New function, split out from...
4563 (vect_gen_prolog_loop_niters): ...here.
4564 (vect_update_init_of_dr): Take a code argument that specifies whether
4565 the adjustment should be added or subtracted.
4566 (vect_update_init_of_drs): Likewise.
4567 (vect_prepare_for_masked_peels): New function.
4568 (vect_do_peeling): Skip prologue peeling if we're using a mask
4569 instead. Update call to vect_update_inits_of_drs.
4570 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4572 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
4573 alignment. Do not include the number of peeled iterations in
4574 the minimum threshold in that case.
4575 (vectorizable_induction): Adjust the start value down by
4576 LOOP_VINFO_MASK_SKIP_NITERS iterations.
4577 (vect_transform_loop): Call vect_prepare_for_masked_peels.
4578 Take the number of skipped iterations into account when calculating
4580 * tree-vect-stmts.c (vect_gen_while_not): New function.
4582 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4583 Alan Hayward <alan.hayward@arm.com>
4584 David Sherwood <david.sherwood@arm.com>
4586 * doc/sourcebuild.texi (vect_fully_masked): Document.
4587 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
4589 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
4591 (vect_analyze_loop_2): ...here. Don't check the vectorization
4592 factor against the number of loop iterations if the loop is
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 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
4600 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
4601 (dump_groups): Update accordingly.
4602 (iv_use::mem_type): New member variable.
4603 (address_p): New function.
4604 (record_use): Add a mem_type argument and initialize the new
4606 (record_group_use): Add a mem_type argument. Use address_p.
4607 Remove obsolete null checks of base_object. Update call to record_use.
4608 (find_interesting_uses_op): Update call to record_group_use.
4609 (find_interesting_uses_cond): Likewise.
4610 (find_interesting_uses_address): Likewise.
4611 (get_mem_type_for_internal_fn): New function.
4612 (find_address_like_use): Likewise.
4613 (find_interesting_uses_stmt): Try find_address_like_use before
4614 calling find_interesting_uses_op.
4615 (addr_offset_valid_p): Use the iv mem_type field as the type
4616 of the addressed memory.
4617 (add_autoinc_candidates): Likewise.
4618 (get_address_cost): Likewise.
4619 (split_small_address_groups_p): Use address_p.
4620 (split_address_groups): Likewise.
4621 (add_iv_candidate_for_use): Likewise.
4622 (autoinc_possible_for_pair): Likewise.
4623 (rewrite_groups): Likewise.
4624 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
4625 (determine_group_iv_cost): Update after split of USE_ADDRESS.
4626 (get_alias_ptr_type_for_ptr_address): New function.
4627 (rewrite_use_address): Rewrite address uses in calls that were
4628 identified by find_address_like_use.
4630 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4631 Alan Hayward <alan.hayward@arm.com>
4632 David Sherwood <david.sherwood@arm.com>
4634 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
4636 * gimple-expr.h (is_gimple_addressable: Likewise.
4637 * gimple-expr.c (is_gimple_address): Likewise.
4638 * internal-fn.c (expand_call_mem_ref): New function.
4639 (expand_mask_load_optab_fn): Use it.
4640 (expand_mask_store_optab_fn): Likewise.
4642 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4643 Alan Hayward <alan.hayward@arm.com>
4644 David Sherwood <david.sherwood@arm.com>
4646 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
4647 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
4648 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
4649 (cond_umax@var{mode}): Document.
4650 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
4651 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
4652 (cond_umin_optab, cond_umax_optab): New optabs.
4653 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
4654 (COND_IOR, COND_XOR): New internal functions.
4655 * internal-fn.h (get_conditional_internal_fn): Declare.
4656 * internal-fn.c (cond_binary_direct): New macro.
4657 (expand_cond_binary_optab_fn): Likewise.
4658 (direct_cond_binary_optab_supported_p): Likewise.
4659 (get_conditional_internal_fn): New function.
4660 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
4661 Cope with reduction statements that are vectorized as calls rather
4663 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
4664 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
4665 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
4666 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4667 (UNSPEC_COND_EOR): New unspecs.
4668 (optab): Add mappings for them.
4669 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
4670 (sve_int_op, sve_fp_op): New int attributes.
4672 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4673 Alan Hayward <alan.hayward@arm.com>
4674 David Sherwood <david.sherwood@arm.com>
4676 * optabs.def (while_ult_optab): New optab.
4677 * doc/md.texi (while_ult@var{m}@var{n}): Document.
4678 * internal-fn.def (WHILE_ULT): New internal function.
4679 * internal-fn.h (direct_internal_fn_supported_p): New override
4680 that takes two types as argument.
4681 * internal-fn.c (while_direct): New macro.
4682 (expand_while_optab_fn): New function.
4683 (convert_optab_supported_p): Likewise.
4684 (direct_while_optab_supported_p): New macro.
4685 * wide-int.h (wi::udiv_ceil): New function.
4686 * tree-vectorizer.h (rgroup_masks): New structure.
4687 (vec_loop_masks): New typedef.
4688 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
4690 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
4691 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
4692 (vect_max_vf): New function.
4693 (slpeel_make_loop_iterate_ntimes): Delete.
4694 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
4695 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
4696 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
4697 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
4698 internal-fn.h, stor-layout.h and optabs-query.h.
4699 (vect_set_loop_mask): New function.
4700 (add_preheader_seq): Likewise.
4701 (add_header_seq): Likewise.
4702 (interleave_supported_p): Likewise.
4703 (vect_maybe_permute_loop_masks): Likewise.
4704 (vect_set_loop_masks_directly): Likewise.
4705 (vect_set_loop_condition_masked): Likewise.
4706 (vect_set_loop_condition_unmasked): New function, split out from
4707 slpeel_make_loop_iterate_ntimes.
4708 (slpeel_make_loop_iterate_ntimes): Rename to..
4709 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
4710 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
4711 (vect_do_peeling): Update call accordingly.
4712 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
4714 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4715 mask_compare_type, can_fully_mask_p and fully_masked_p.
4716 (release_vec_loop_masks): New function.
4717 (_loop_vec_info): Use it to free the loop masks.
4718 (can_produce_all_loop_masks_p): New function.
4719 (vect_get_max_nscalars_per_iter): Likewise.
4720 (vect_verify_full_masking): Likewise.
4721 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
4722 retries, and free the mask rgroups before retrying. Check loop-wide
4723 reasons for disallowing fully-masked loops. Make the final decision
4724 about whether use a fully-masked loop or not.
4725 (vect_estimate_min_profitable_iters): Do not assume that peeling
4726 for the number of iterations will be needed for fully-masked loops.
4727 (vectorizable_reduction): Disable fully-masked loops.
4728 (vectorizable_live_operation): Likewise.
4729 (vect_halve_mask_nunits): New function.
4730 (vect_double_mask_nunits): Likewise.
4731 (vect_record_loop_mask): Likewise.
4732 (vect_get_loop_mask): Likewise.
4733 (vect_transform_loop): Handle the case in which the final loop
4734 iteration might handle a partial vector. Call vect_set_loop_condition
4735 instead of slpeel_make_loop_iterate_ntimes.
4736 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
4737 (check_load_store_masking): New function.
4738 (prepare_load_store_mask): Likewise.
4739 (vectorizable_store): Handle fully-masked loops.
4740 (vectorizable_load): Likewise.
4741 (supportable_widening_operation): Use vect_halve_mask_nunits for
4743 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
4744 (vect_gen_while): New function.
4745 * config/aarch64/aarch64.md (umax<mode>3): New expander.
4746 (aarch64_uqdec<mode>): New insn.
4748 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4749 Alan Hayward <alan.hayward@arm.com>
4750 David Sherwood <david.sherwood@arm.com>
4752 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
4753 (reduc_xor_scal_optab): New optabs.
4754 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
4755 (reduc_xor_scal_@var{m}): Document.
4756 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
4757 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
4759 * fold-const-call.c (fold_const_call): Handle them.
4760 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
4761 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
4762 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
4763 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
4764 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
4765 (UNSPEC_XORV): New unspecs.
4766 (optab): Add entries for them.
4767 (BITWISEV): New int iterator.
4768 (bit_reduc_op): New int attributes.
4770 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4771 Alan Hayward <alan.hayward@arm.com>
4772 David Sherwood <david.sherwood@arm.com>
4774 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
4775 * internal-fn.def (VEC_SHL_INSERT): New internal function.
4776 * optabs.def (vec_shl_insert_optab): New optab.
4777 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
4778 (duplicate_and_interleave): Likewise.
4779 * tree-vect-loop.c: Include internal-fn.h.
4780 (neutral_op_for_slp_reduction): New function, split out from
4781 get_initial_defs_for_reduction.
4782 (get_initial_def_for_reduction): Handle option 2 for variable-length
4783 vectors by loading the neutral value into a vector and then shifting
4784 the initial value into element 0.
4785 (get_initial_defs_for_reduction): Replace the code argument with
4786 the neutral value calculated by neutral_op_for_slp_reduction.
4787 Use gimple_build_vector for constant-length vectors.
4788 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
4789 but the first group_size elements have a neutral value.
4790 Use duplicate_and_interleave otherwise.
4791 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
4792 Update call to get_initial_defs_for_reduction. Handle SLP
4793 reductions for variable-length vectors by creating one vector
4794 result for each scalar result, with the elements associated
4795 with other scalar results stubbed out with the neutral value.
4796 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
4797 Require IFN_VEC_SHL_INSERT for double reductions on
4798 variable-length vectors, or SLP reductions that have
4799 a neutral value. Require can_duplicate_and_interleave_p
4800 support for variable-length unchained SLP reductions if there
4801 is no neutral value, such as for MIN/MAX reductions. Also require
4802 the number of vector elements to be a multiple of the number of
4803 SLP statements when doing variable-length unchained SLP reductions.
4804 Update call to vect_create_epilog_for_reduction.
4805 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
4806 and remove initial values.
4807 (duplicate_and_interleave): Make public.
4808 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
4809 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
4811 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4812 Alan Hayward <alan.hayward@arm.com>
4813 David Sherwood <david.sherwood@arm.com>
4815 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
4816 (can_duplicate_and_interleave_p): New function.
4817 (vect_get_and_check_slp_defs): Take the vector of statements
4818 rather than just the current one. Remove excess parentheses.
4819 Restriction rejectinon of vect_constant_def and vect_external_def
4820 for variable-length vectors to boolean types, or types for which
4821 can_duplicate_and_interleave_p is false.
4822 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
4823 (duplicate_and_interleave): New function.
4824 (vect_get_constant_vectors): Use gimple_build_vector for
4825 constant-length vectors and suitable variable-length constant
4826 vectors. Use duplicate_and_interleave for other variable-length
4827 vectors. Don't defer the update when inserting new statements.
4829 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4830 Alan Hayward <alan.hayward@arm.com>
4831 David Sherwood <david.sherwood@arm.com>
4833 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
4834 min_profitable_iters doesn't go negative.
4836 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4837 Alan Hayward <alan.hayward@arm.com>
4838 David Sherwood <david.sherwood@arm.com>
4840 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
4841 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
4842 * optabs.def (vec_mask_load_lanes_optab): New optab.
4843 (vec_mask_store_lanes_optab): Likewise.
4844 * internal-fn.def (MASK_LOAD_LANES): New internal function.
4845 (MASK_STORE_LANES): Likewise.
4846 * internal-fn.c (mask_load_lanes_direct): New macro.
4847 (mask_store_lanes_direct): Likewise.
4848 (expand_mask_load_optab_fn): Handle masked operations.
4849 (expand_mask_load_lanes_optab_fn): New macro.
4850 (expand_mask_store_optab_fn): Handle masked operations.
4851 (expand_mask_store_lanes_optab_fn): New macro.
4852 (direct_mask_load_lanes_optab_supported_p): Likewise.
4853 (direct_mask_store_lanes_optab_supported_p): Likewise.
4854 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
4856 (vect_load_lanes_supported): Likewise.
4857 * tree-vect-data-refs.c (strip_conversion): New function.
4858 (can_group_stmts_p): Likewise.
4859 (vect_analyze_data_ref_accesses): Use it instead of checking
4860 for a pair of assignments.
4861 (vect_store_lanes_supported): Take a masked_p parameter.
4862 (vect_load_lanes_supported): Likewise.
4863 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
4864 vect_store_lanes_supported and vect_load_lanes_supported.
4865 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4866 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
4867 parameter. Don't allow gaps for masked accesses.
4868 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
4869 and vect_load_lanes_supported.
4870 (get_load_store_type): Take a masked_p parameter and update
4871 call to get_group_load_store_type.
4872 (vectorizable_store): Update call to get_load_store_type.
4873 Handle IFN_MASK_STORE_LANES.
4874 (vectorizable_load): Update call to get_load_store_type.
4875 Handle IFN_MASK_LOAD_LANES.
4877 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4878 Alan Hayward <alan.hayward@arm.com>
4879 David Sherwood <david.sherwood@arm.com>
4881 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
4883 * config/aarch64/aarch64-protos.h
4884 (aarch64_sve_struct_memory_operand_p): Declare.
4885 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
4886 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
4887 (VPRED, vpred): Handle SVE structure modes.
4888 * config/aarch64/constraints.md (Utx): New constraint.
4889 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
4890 (aarch64_sve_struct_nonimmediate_operand): New predicates.
4891 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
4892 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
4893 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
4894 structure modes. Split into pieces after RA.
4895 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
4896 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
4898 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
4899 SVE structure modes.
4900 (aarch64_classify_address): Likewise.
4901 (sizetochar): Move earlier in file.
4902 (aarch64_print_operand): Handle SVE register lists.
4903 (aarch64_array_mode): New function.
4904 (aarch64_sve_struct_memory_operand_p): Likewise.
4905 (TARGET_ARRAY_MODE): Redefine.
4907 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4908 Alan Hayward <alan.hayward@arm.com>
4909 David Sherwood <david.sherwood@arm.com>
4911 * target.def (array_mode): New target hook.
4912 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
4913 * doc/tm.texi: Regenerate.
4914 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
4915 * hooks.c (hook_optmode_mode_uhwi_none): New function.
4916 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
4918 * stor-layout.c (mode_for_array): Likewise. Support polynomial
4921 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4922 Alan Hayward <alan.hayward@arm.com>
4923 David Sherwood <david.sherwood@arm.com>
4925 * fold-const.c (fold_binary_loc): Check the argument types
4926 rather than the result type when testing for a vector operation.
4928 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4930 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
4931 * doc/tm.texi: Regenerate.
4933 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4934 Alan Hayward <alan.hayward@arm.com>
4935 David Sherwood <david.sherwood@arm.com>
4937 * doc/invoke.texi (-msve-vector-bits=): Document new option.
4938 (sve): Document new AArch64 extension.
4939 * doc/md.texi (w): Extend the description of the AArch64
4940 constraint to include SVE vectors.
4941 (Upl, Upa): Document new AArch64 predicate constraints.
4942 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
4944 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
4945 (msve-vector-bits=): New option.
4946 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
4947 SVE when these are disabled.
4948 (sve): New extension.
4949 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
4950 modes. Adjust their number of units based on aarch64_sve_vg.
4951 (MAX_BITSIZE_MODE_ANY_MODE): Define.
4952 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
4953 aarch64_addr_query_type.
4954 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
4955 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
4956 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
4957 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
4958 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
4959 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
4960 (aarch64_simd_imm_zero_p): Delete.
4961 (aarch64_check_zero_based_sve_index_immediate): Declare.
4962 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4963 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4964 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4965 (aarch64_sve_float_mul_immediate_p): Likewise.
4966 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4968 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
4969 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
4970 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
4971 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
4972 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
4973 (aarch64_regmode_natural_size): Likewise.
4974 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
4975 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
4977 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
4978 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
4979 for VG and the SVE predicate registers.
4980 (V_ALIASES): Add a "z"-prefixed alias.
4981 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
4982 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
4983 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
4984 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
4985 (REG_CLASS_NAMES): Add entries for them.
4986 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
4987 and the predicate registers.
4988 (aarch64_sve_vg): Declare.
4989 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
4990 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
4991 (REGMODE_NATURAL_SIZE): Define.
4992 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
4994 * config/aarch64/aarch64.c: Include cfgrtl.h.
4995 (simd_immediate_info): Add a constructor for series vectors,
4996 and an associated step field.
4997 (aarch64_sve_vg): New variable.
4998 (aarch64_dbx_register_number): Handle VG and the predicate registers.
4999 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
5000 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
5001 (VEC_ANY_DATA, VEC_STRUCT): New constants.
5002 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
5003 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
5004 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
5005 (aarch64_get_mask_mode): New functions.
5006 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
5007 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
5008 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
5009 predicate modes and predicate registers. Explicitly restrict
5010 GPRs to modes of 16 bytes or smaller. Only allow FP registers
5011 to store a vector mode if it is recognized by
5012 aarch64_classify_vector_mode.
5013 (aarch64_regmode_natural_size): New function.
5014 (aarch64_hard_regno_caller_save_mode): Return the original mode
5016 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
5017 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
5018 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
5019 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
5021 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
5022 does not overlap dest if the function is frame-related. Handle
5024 (aarch64_split_add_offset): New function.
5025 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
5026 them aarch64_add_offset.
5027 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
5028 and update call to aarch64_sub_sp.
5029 (aarch64_add_cfa_expression): New function.
5030 (aarch64_expand_prologue): Pass extra temporary registers to the
5031 functions above. Handle the case in which we need to emit new
5032 DW_CFA_expressions for registers that were originally saved
5033 relative to the stack pointer, but now have to be expressed
5034 relative to the frame pointer.
5035 (aarch64_output_mi_thunk): Pass extra temporary registers to the
5037 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
5038 IP0 and IP1 values for SVE frames.
5039 (aarch64_expand_vec_series): New function.
5040 (aarch64_expand_sve_widened_duplicate): Likewise.
5041 (aarch64_expand_sve_const_vector): Likewise.
5042 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
5043 Handle SVE constants. Use emit_move_insn to move a force_const_mem
5044 into the register, rather than emitting a SET directly.
5045 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
5046 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
5047 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
5048 (offset_9bit_signed_scaled_p): New functions.
5049 (aarch64_replicate_bitmask_imm): New function.
5050 (aarch64_bitmask_imm): Use it.
5051 (aarch64_cannot_force_const_mem): Reject expressions involving
5052 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
5053 (aarch64_classify_index): Handle SVE indices, by requiring
5054 a plain register index with a scale that matches the element size.
5055 (aarch64_classify_address): Handle SVE addresses. Assert that
5056 the mode of the address is VOIDmode or an integer mode.
5057 Update call to aarch64_classify_symbol.
5058 (aarch64_classify_symbolic_expression): Update call to
5059 aarch64_classify_symbol.
5060 (aarch64_const_vec_all_in_range_p): New function.
5061 (aarch64_print_vector_float_operand): Likewise.
5062 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
5063 "vN" for FP registers with SVE modes. Handle (const ...) vectors
5064 and the FP immediates 1.0 and 0.5.
5065 (aarch64_print_address_internal): Handle SVE addresses.
5066 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
5067 (aarch64_regno_regclass): Handle predicate registers.
5068 (aarch64_secondary_reload): Handle big-endian reloads of SVE
5070 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
5071 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
5072 (aarch64_convert_sve_vector_bits): New function.
5073 (aarch64_override_options): Use it to handle -msve-vector-bits=.
5074 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5076 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
5077 Handle SVE vector and predicate modes. Accept VL-based constants
5078 that need only one temporary register, and VL offsets that require
5079 no temporary registers.
5080 (aarch64_conditional_register_usage): Mark the predicate registers
5081 as fixed if SVE isn't available.
5082 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
5083 Return true for SVE vector and predicate modes.
5084 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
5085 rather than an unsigned int. Handle SVE modes.
5086 (aarch64_preferred_simd_mode): Update call accordingly. Handle
5088 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
5090 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5091 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5092 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5093 (aarch64_sve_float_mul_immediate_p): New functions.
5094 (aarch64_sve_valid_immediate): New function.
5095 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
5096 Explicitly reject structure modes. Check for INDEX constants.
5097 Handle PTRUE and PFALSE constants.
5098 (aarch64_check_zero_based_sve_index_immediate): New function.
5099 (aarch64_simd_imm_zero_p): Delete.
5100 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
5101 vector modes. Accept constants in the range of CNT[BHWD].
5102 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
5103 ask for an Advanced SIMD mode.
5104 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
5105 (aarch64_simd_vector_alignment): Handle SVE predicates.
5106 (aarch64_vectorize_preferred_vector_alignment): New function.
5107 (aarch64_simd_vector_alignment_reachable): Use it instead of
5109 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
5110 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
5112 (MAX_VECT_LEN): Delete.
5113 (expand_vec_perm_d): Add a vec_flags field.
5114 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
5115 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
5116 (aarch64_evpc_ext): Don't apply a big-endian lane correction
5118 (aarch64_evpc_rev): Rename to...
5119 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
5120 (aarch64_evpc_rev_global): New function.
5121 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
5122 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
5124 (aarch64_evpc_sve_tbl): New function.
5125 (aarch64_expand_vec_perm_const_1): Update after rename of
5126 aarch64_evpc_rev. Handle SVE permutes too, trying
5127 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
5128 than aarch64_evpc_tbl.
5129 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
5130 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
5131 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
5132 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
5133 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
5134 (aarch64_expand_sve_vcond): New functions.
5135 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
5136 of aarch64_vector_mode_p.
5137 (aarch64_dwarf_poly_indeterminate_value): New function.
5138 (aarch64_compute_pressure_classes): Likewise.
5139 (aarch64_can_change_mode_class): Likewise.
5140 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
5141 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
5142 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
5143 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
5144 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
5145 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
5146 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
5147 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
5149 (Dn, Dl, Dr): Accept const as well as const_vector.
5150 (Dz): Likewise. Compare against CONST0_RTX.
5151 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
5152 of "vector" where appropriate.
5153 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
5154 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
5155 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
5156 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
5157 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
5158 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
5159 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
5160 (v_int_equiv): Extend to SVE modes.
5161 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
5163 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
5164 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
5165 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
5166 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
5167 (SVE_COND_FP_CMP): New int iterators.
5168 (perm_hilo): Handle the new unpack unspecs.
5169 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
5171 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
5172 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
5173 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
5174 (aarch64_equality_operator, aarch64_constant_vector_operand)
5175 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
5176 (aarch64_sve_nonimmediate_operand): Likewise.
5177 (aarch64_sve_general_operand): Likewise.
5178 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
5179 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
5180 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
5181 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
5182 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
5183 (aarch64_sve_float_arith_immediate): Likewise.
5184 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
5185 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
5186 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
5187 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
5188 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
5189 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
5190 (aarch64_sve_float_arith_operand): Likewise.
5191 (aarch64_sve_float_arith_with_sub_operand): Likewise.
5192 (aarch64_sve_float_mul_operand): Likewise.
5193 (aarch64_sve_vec_perm_operand): Likewise.
5194 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
5195 (aarch64_mov_operand): Accept const_poly_int and const_vector.
5196 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
5197 as well as const_vector.
5198 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
5199 in file. Use CONST0_RTX and CONSTM1_RTX.
5200 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
5201 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
5202 Use aarch64_simd_imm_zero.
5203 * config/aarch64/aarch64-sve.md: New file.
5204 * config/aarch64/aarch64.md: Include it.
5205 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
5206 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
5207 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
5208 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
5209 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
5210 (sve): New attribute.
5211 (enabled): Disable instructions with the sve attribute unless
5213 (movqi, movhi): Pass CONST_POLY_INT operaneds through
5214 aarch64_expand_mov_immediate.
5215 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
5216 CNT[BHSD] immediates.
5217 (movti): Split CONST_POLY_INT moves into two halves.
5218 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
5219 Split additions that need a temporary here if the destination
5220 is the stack pointer.
5221 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
5222 (*add<mode>3_poly_1): New instruction.
5223 (set_clobber_cc): New expander.
5225 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5227 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
5228 parameter and use it instead of GET_MODE_SIZE (innermode). Use
5229 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
5230 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
5231 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
5232 Change innermode from fixed_mode_size to machine_mode.
5233 (simplify_subreg): Update call accordingly. Handle a constant-sized
5234 subreg of a variable-length CONST_VECTOR.
5236 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5237 Alan Hayward <alan.hayward@arm.com>
5238 David Sherwood <david.sherwood@arm.com>
5240 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
5241 (add_offset_to_base): New function, split out from...
5242 (create_mem_ref): ...here. When handling a scale other than 1,
5243 check first whether the address is valid without the offset.
5244 Add it into the base if so, leaving the index and scale as-is.
5246 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5249 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
5250 fold_for_warn before checking if arg2 is INTEGER_CST.
5252 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
5254 * config/rs6000/predicates.md (load_multiple_operation): Delete.
5255 (store_multiple_operation): Delete.
5256 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
5257 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
5258 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
5259 guarded by TARGET_STRING.
5260 (rs6000_output_load_multiple): Delete.
5261 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
5262 OPTION_MASK_STRING / TARGET_STRING handling.
5263 (print_operand) <'N', 'O'>: Add comment that these are unused now.
5264 (const rs6000_opt_masks) <"string">: Change mask to 0.
5265 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
5266 (MASK_STRING): Delete.
5267 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
5269 (load_multiple): Delete.
5276 (store_multiple): Delete.
5283 (movmemsi_8reg): Delete.
5284 (corresponding unnamed define_insn): Delete.
5285 (movmemsi_6reg): Delete.
5286 (corresponding unnamed define_insn): Delete.
5287 (movmemsi_4reg): Delete.
5288 (corresponding unnamed define_insn): Delete.
5289 (movmemsi_2reg): Delete.
5290 (corresponding unnamed define_insn): Delete.
5291 (movmemsi_1reg): Delete.
5292 (corresponding unnamed define_insn): Delete.
5293 * config/rs6000/rs6000.opt (mno-string): New.
5294 (mstring): Replace by deprecation warning stub.
5295 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
5297 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5299 * regrename.c (regrename_do_replace): If replacing the same
5300 reg multiple times, try to reuse last created gen_raw_REG.
5303 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
5304 main to workaround a bug in GDB.
5306 2018-01-12 Tom de Vries <tom@codesourcery.com>
5309 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
5311 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
5313 PR rtl-optimization/80481
5314 * ira-color.c (get_cap_member): New function.
5315 (allocnos_conflict_by_live_ranges_p): Use it.
5316 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
5317 (setup_slot_coalesced_allocno_live_ranges): Ditto.
5319 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
5322 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
5323 (*saddl_se_1): Ditto.
5325 (*ssubl_se_1): Ditto.
5327 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5329 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
5330 rather than wi::to_widest for DR_INITs.
5331 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
5332 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
5333 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
5335 (vect_analyze_group_access_1): Note that here.
5337 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5339 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
5340 polynomial type sizes.
5342 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5344 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
5345 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
5346 (gimple_add_tmp_var): Likewise.
5348 2018-01-12 Martin Liska <mliska@suse.cz>
5350 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
5351 (gimple_alloc_sizes): Likewise.
5352 (dump_gimple_statistics): Use PRIu64 in printf format.
5353 * gimple.h: Change uint64_t to int.
5355 2018-01-12 Martin Liska <mliska@suse.cz>
5357 * tree-core.h: Use uint64_t instead of int.
5358 * tree.c (tree_node_counts): Likewise.
5359 (tree_node_sizes): Likewise.
5360 (dump_tree_statistics): Use PRIu64 in printf format.
5362 2018-01-12 Martin Liska <mliska@suse.cz>
5364 * Makefile.in: As qsort_chk is implemented in vec.c, add
5365 vec.o to linkage of gencfn-macros.
5366 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
5367 passing the info to record_node_allocation_statistics.
5368 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
5370 * ggc-common.c (struct ggc_usage): Add operator== and use
5371 it in operator< and compare function.
5372 * mem-stats.h (struct mem_usage): Likewise.
5373 * vec.c (struct vec_usage): Remove operator< and compare
5374 function. Can be simply inherited.
5376 2018-01-12 Martin Jambor <mjambor@suse.cz>
5379 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
5380 * tree-ssa-math-opts.c: Include domwalk.h.
5381 (convert_mult_to_fma_1): New function.
5382 (fma_transformation_info): New type.
5383 (fma_deferring_state): Likewise.
5384 (cancel_fma_deferring): New function.
5385 (result_of_phi): Likewise.
5386 (last_fma_candidate_feeds_initial_phi): Likewise.
5387 (convert_mult_to_fma): Added deferring logic, split actual
5388 transformation to convert_mult_to_fma_1.
5389 (math_opts_dom_walker): New type.
5390 (math_opts_dom_walker::after_dom_children): New method, body moved
5391 here from pass_optimize_widening_mul::execute, added deferring logic
5393 (pass_optimize_widening_mul::execute): Moved most of code to
5394 math_opts_dom_walker::after_dom_children.
5395 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
5396 * config/i386/i386.c (ix86_option_override_internal): Added
5397 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
5399 2018-01-12 Richard Biener <rguenther@suse.de>
5402 * dwarf2out.c (gen_variable_die): Do not reset old_die for
5403 inline instance vars.
5405 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
5408 * config/rx/rx.c (rx_is_restricted_memory_address):
5411 2018-01-12 Richard Biener <rguenther@suse.de>
5413 PR tree-optimization/80846
5414 * target.def (split_reduction): New target hook.
5415 * targhooks.c (default_split_reduction): New function.
5416 * targhooks.h (default_split_reduction): Declare.
5417 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
5418 target requests first reduce vectors by combining low and high
5420 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
5421 (get_vectype_for_scalar_type_and_size): Export.
5422 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
5423 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
5424 * doc/tm.texi: Regenerate.
5425 * config/i386/i386.c (ix86_split_reduction): Implement
5426 TARGET_VECTORIZE_SPLIT_REDUCTION.
5428 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5431 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
5432 in PIC mode except for TARGET_VXWORKS_RTP.
5433 * config/sparc/sparc.c: Include cfgrtl.h.
5434 (TARGET_INIT_PIC_REG): Define.
5435 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
5436 (sparc_pic_register_p): New predicate.
5437 (sparc_legitimate_address_p): Use it.
5438 (sparc_legitimize_pic_address): Likewise.
5439 (sparc_delegitimize_address): Likewise.
5440 (sparc_mode_dependent_address_p): Likewise.
5441 (gen_load_pcrel_sym): Remove 4th parameter.
5442 (load_got_register): Adjust call to above. Remove obsolete stuff.
5443 (sparc_expand_prologue): Do not call load_got_register here.
5444 (sparc_flat_expand_prologue): Likewise.
5445 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
5446 (sparc_use_pseudo_pic_reg): New function.
5447 (sparc_init_pic_reg): Likewise.
5448 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5449 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
5451 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
5453 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
5454 Add item for branch_cost.
5456 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5458 PR rtl-optimization/83565
5459 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
5460 not extend the result to a larger mode for rotate operations.
5461 (num_sign_bit_copies1): Likewise.
5463 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5466 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
5468 Use values-Xc.o for -pedantic.
5469 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
5471 2018-01-12 Martin Liska <mliska@suse.cz>
5474 * ipa-devirt.c (final_warning_record::grow_type_warnings):
5476 (possible_polymorphic_call_targets): Use it.
5477 (ipa_devirt): Likewise.
5479 2018-01-12 Martin Liska <mliska@suse.cz>
5481 * profile-count.h (enum profile_quality): Use 0 as invalid
5482 enum value of profile_quality.
5484 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
5486 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
5487 -mext-string options.
5489 2018-01-12 Richard Biener <rguenther@suse.de>
5491 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
5492 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
5493 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
5495 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
5497 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5499 * configure.ac (--with-long-double-format): Add support for the
5500 configuration option to change the default long double format on
5502 * config.gcc (powerpc*-linux*-*): Likewise.
5503 * configure: Regenerate.
5504 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
5505 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
5506 used without modification.
5508 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5510 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
5511 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
5512 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
5513 MISC_BUILTIN_SPEC_BARRIER.
5514 (rs6000_init_builtins): Likewise.
5515 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
5517 (speculation_barrier): New define_insn.
5518 * doc/extend.texi: Document __builtin_speculation_barrier.
5520 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5523 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
5524 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
5525 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
5527 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
5528 integral modes instead of "ss" and "sd".
5529 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
5530 vectors with 32-bit and 64-bit elements.
5531 (vecdupssescalarmodesuffix): New mode attribute.
5532 (vec_dup<mode>): Use it.
5534 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
5537 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
5538 frame if argument is passed on stack.
5540 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5543 * ree.c (combine_reaching_defs): Optimize also
5544 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
5545 reg2=any_extend(exp); reg1=reg2;, formatting fix.
5547 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5550 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
5552 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5555 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
5556 after they are computed.
5558 2018-01-11 Bin Cheng <bin.cheng@arm.com>
5560 PR tree-optimization/83695
5561 * gimple-loop-linterchange.cc
5562 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
5563 reset cached scev information after interchange.
5564 (pass_linterchange::execute): Remove call to scev_reset_htab.
5566 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5568 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
5569 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
5570 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
5571 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
5572 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
5573 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
5574 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
5575 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
5576 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
5577 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
5578 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
5579 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
5580 (V_lane_reg): Likewise.
5581 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
5583 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
5584 (vfmal_lane_low<mode>_intrinsic,
5585 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
5586 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
5587 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
5588 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
5589 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
5590 vfmsl_lane_high<mode>_intrinsic): New define_insns.
5592 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5594 * config/arm/arm-cpus.in (fp16fml): New feature.
5595 (ALL_SIMD): Add fp16fml.
5596 (armv8.2-a): Add fp16fml as an option.
5597 (armv8.3-a): Likewise.
5598 (armv8.4-a): Add fp16fml as part of fp16.
5599 * config/arm/arm.h (TARGET_FP16FML): Define.
5600 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
5602 * config/arm/arm-modes.def (V2HF): Define.
5603 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
5604 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
5605 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
5606 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
5607 vfmsl_low, vfmsl_high): New set of builtins.
5608 * config/arm/iterators.md (PLUSMINUS): New code iterator.
5609 (vfml_op): New code attribute.
5610 (VFMLHALVES): New int iterator.
5611 (VFML, VFMLSEL): New mode attributes.
5612 (V_reg): Define mapping for V2HF.
5613 (V_hi, V_lo): New mode attributes.
5614 (VF_constraint): Likewise.
5615 (vfml_half, vfml_half_selector): New int attributes.
5616 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
5618 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
5619 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
5621 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
5622 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
5623 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
5624 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
5626 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
5627 Document new effective target and option set.
5629 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5631 * config/arm/arm-cpus.in (armv8_4): New feature.
5632 (ARMv8_4a): New fgroup.
5633 (armv8.4-a): New arch.
5634 * config/arm/arm-tables.opt: Regenerate.
5635 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
5636 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
5637 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
5638 Add matching rules for -march=armv8.4-a and extensions.
5639 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
5641 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
5644 * config/rx/rx.md (BW): New mode attribute.
5645 (sync_lock_test_and_setsi): Add mode suffix to insn output.
5647 2018-01-11 Richard Biener <rguenther@suse.de>
5649 PR tree-optimization/83435
5650 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
5651 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
5652 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
5654 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5655 Alan Hayward <alan.hayward@arm.com>
5656 David Sherwood <david.sherwood@arm.com>
5658 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
5660 (aarch64_classify_address): Initialize it. Track polynomial offsets.
5661 (aarch64_print_address_internal): Use it to check for a zero offset.
5663 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5664 Alan Hayward <alan.hayward@arm.com>
5665 David Sherwood <david.sherwood@arm.com>
5667 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
5668 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
5669 Return a poly_int64 rather than a HOST_WIDE_INT.
5670 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
5671 rather than a HOST_WIDE_INT.
5672 * config/aarch64/aarch64.h (aarch64_frame): Protect with
5673 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
5674 hard_fp_offset, frame_size, initial_adjust, callee_offset and
5675 final_offset from HOST_WIDE_INT to poly_int64.
5676 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
5677 to_constant when getting the number of units in an Advanced SIMD
5679 (aarch64_builtin_vectorized_function): Check for a constant number
5681 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
5683 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
5684 attribute instead of GET_MODE_NUNITS.
5685 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
5686 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
5687 GET_MODE_SIZE for fixed-size registers.
5688 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
5689 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
5690 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
5691 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
5692 (aarch64_print_operand, aarch64_print_address_internal)
5693 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
5694 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
5695 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
5696 Handle polynomial GET_MODE_SIZE.
5697 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
5698 wider than SImode without modification.
5699 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
5700 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
5701 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
5702 passing and returning SVE modes.
5703 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
5704 rather than GEN_INT.
5705 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
5706 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
5707 (aarch64_allocate_and_probe_stack_space): Likewise.
5708 (aarch64_layout_frame): Cope with polynomial offsets.
5709 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
5710 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
5712 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
5713 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
5714 poly_int64 rather than a HOST_WIDE_INT.
5715 (aarch64_get_separate_components, aarch64_process_components)
5716 (aarch64_expand_prologue, aarch64_expand_epilogue)
5717 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
5718 (aarch64_anchor_offset): New function, split out from...
5719 (aarch64_legitimize_address): ...here.
5720 (aarch64_builtin_vectorization_cost): Handle polynomial
5721 TYPE_VECTOR_SUBPARTS.
5722 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
5724 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
5725 number of elements from the PARALLEL rather than the mode.
5726 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
5727 rather than GET_MODE_BITSIZE.
5728 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
5729 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
5730 (aarch64_expand_vec_perm_const_1): Handle polynomial
5731 d->perm.length () and d->perm elements.
5732 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
5733 Apply to_constant to d->perm elements.
5734 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
5735 polynomial CONST_VECTOR_NUNITS.
5736 (aarch64_move_pointer): Take amount as a poly_int64 rather
5738 (aarch64_progress_pointer): Avoid temporary variable.
5739 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
5740 the mode attribute instead of GET_MODE.
5742 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5743 Alan Hayward <alan.hayward@arm.com>
5744 David Sherwood <david.sherwood@arm.com>
5746 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
5747 x exists before using it.
5748 (aarch64_add_constant_internal): Rename to...
5749 (aarch64_add_offset_1): ...this. Replace regnum with separate
5750 src and dest rtxes. Handle the case in which they're different,
5751 including when the offset is zero. Replace scratchreg with an rtx.
5752 Use 2 additions if there is no spare register into which we can
5753 move a 16-bit constant.
5754 (aarch64_add_constant): Delete.
5755 (aarch64_add_offset): Replace reg with separate src and dest
5756 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
5757 Use aarch64_add_offset_1.
5758 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
5759 an rtx rather than an int. Take the delta as a poly_int64
5760 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
5761 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
5762 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
5763 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
5764 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
5766 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
5767 aarch64_add_constant.
5769 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5771 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
5772 Use scalar_float_mode.
5774 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5776 * config/aarch64/aarch64-simd.md
5777 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
5778 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
5779 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
5780 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
5781 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
5782 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
5783 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
5784 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
5785 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
5786 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
5788 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5791 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
5792 targ_options->x_arm_arch_string is non NULL.
5794 2018-01-11 Tamar Christina <tamar.christina@arm.com>
5796 * config/aarch64/aarch64.h
5797 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
5799 2018-01-11 Sudakshina Das <sudi.das@arm.com>
5802 * expmed.c (emit_store_flag_force): Swap if const op0
5803 and change VOIDmode to mode of op0.
5805 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5807 PR rtl-optimization/83761
5808 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
5809 than bytes to mode_for_size.
5811 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5814 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
5815 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
5818 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5821 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
5822 when in layout mode.
5823 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
5824 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
5827 2018-01-10 Michael Collison <michael.collison@arm.com>
5829 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
5830 * config/aarch64/aarch64-option-extension.def: Add
5831 AARCH64_OPT_EXTENSION of 'fp16fml'.
5832 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5833 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
5834 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
5835 * config/aarch64/constraints.md (Ui7): New constraint.
5836 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
5837 (VFMLA_SEL_W): Ditto.
5840 (VFMLA16_LOW): New int iterator.
5841 (VFMLA16_HIGH): Ditto.
5842 (UNSPEC_FMLAL): New unspec.
5843 (UNSPEC_FMLSL): Ditto.
5844 (UNSPEC_FMLAL2): Ditto.
5845 (UNSPEC_FMLSL2): Ditto.
5846 (f16mac): New code attribute.
5847 * config/aarch64/aarch64-simd-builtins.def
5848 (aarch64_fmlal_lowv2sf): Ditto.
5849 (aarch64_fmlsl_lowv2sf): Ditto.
5850 (aarch64_fmlalq_lowv4sf): Ditto.
5851 (aarch64_fmlslq_lowv4sf): Ditto.
5852 (aarch64_fmlal_highv2sf): Ditto.
5853 (aarch64_fmlsl_highv2sf): Ditto.
5854 (aarch64_fmlalq_highv4sf): Ditto.
5855 (aarch64_fmlslq_highv4sf): Ditto.
5856 (aarch64_fmlal_lane_lowv2sf): Ditto.
5857 (aarch64_fmlsl_lane_lowv2sf): Ditto.
5858 (aarch64_fmlal_laneq_lowv2sf): Ditto.
5859 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
5860 (aarch64_fmlalq_lane_lowv4sf): Ditto.
5861 (aarch64_fmlsl_lane_lowv4sf): Ditto.
5862 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
5863 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
5864 (aarch64_fmlal_lane_highv2sf): Ditto.
5865 (aarch64_fmlsl_lane_highv2sf): Ditto.
5866 (aarch64_fmlal_laneq_highv2sf): Ditto.
5867 (aarch64_fmlsl_laneq_highv2sf): Ditto.
5868 (aarch64_fmlalq_lane_highv4sf): Ditto.
5869 (aarch64_fmlsl_lane_highv4sf): Ditto.
5870 (aarch64_fmlalq_laneq_highv4sf): Ditto.
5871 (aarch64_fmlsl_laneq_highv4sf): Ditto.
5872 * config/aarch64/aarch64-simd.md:
5873 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
5874 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5875 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
5876 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5877 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
5878 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
5879 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
5880 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
5881 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
5882 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
5883 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
5884 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
5885 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
5886 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
5887 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
5888 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
5889 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
5890 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
5891 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
5892 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
5893 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
5894 (vfmlsl_low_u32): Ditto.
5895 (vfmlalq_low_u32): Ditto.
5896 (vfmlslq_low_u32): Ditto.
5897 (vfmlal_high_u32): Ditto.
5898 (vfmlsl_high_u32): Ditto.
5899 (vfmlalq_high_u32): Ditto.
5900 (vfmlslq_high_u32): Ditto.
5901 (vfmlal_lane_low_u32): Ditto.
5902 (vfmlsl_lane_low_u32): Ditto.
5903 (vfmlal_laneq_low_u32): Ditto.
5904 (vfmlsl_laneq_low_u32): Ditto.
5905 (vfmlalq_lane_low_u32): Ditto.
5906 (vfmlslq_lane_low_u32): Ditto.
5907 (vfmlalq_laneq_low_u32): Ditto.
5908 (vfmlslq_laneq_low_u32): Ditto.
5909 (vfmlal_lane_high_u32): Ditto.
5910 (vfmlsl_lane_high_u32): Ditto.
5911 (vfmlal_laneq_high_u32): Ditto.
5912 (vfmlsl_laneq_high_u32): Ditto.
5913 (vfmlalq_lane_high_u32): Ditto.
5914 (vfmlslq_lane_high_u32): Ditto.
5915 (vfmlalq_laneq_high_u32): Ditto.
5916 (vfmlslq_laneq_high_u32): Ditto.
5917 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
5918 (AARCH64_FL_FOR_ARCH8_4): New.
5919 (AARCH64_ISA_F16FML): New ISA flag.
5920 (TARGET_F16FML): New feature flag for fp16fml.
5921 (doc/invoke.texi): Document new fp16fml option.
5923 2018-01-10 Michael Collison <michael.collison@arm.com>
5925 * config/aarch64/aarch64-builtins.c:
5926 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
5927 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5928 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
5929 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
5930 (AARCH64_ISA_SHA3): New ISA flag.
5931 (TARGET_SHA3): New feature flag for sha3.
5932 * config/aarch64/iterators.md (sha512_op): New int attribute.
5933 (CRYPTO_SHA512): New int iterator.
5934 (UNSPEC_SHA512H): New unspec.
5935 (UNSPEC_SHA512H2): Ditto.
5936 (UNSPEC_SHA512SU0): Ditto.
5937 (UNSPEC_SHA512SU1): Ditto.
5938 * config/aarch64/aarch64-simd-builtins.def
5939 (aarch64_crypto_sha512hqv2di): New builtin.
5940 (aarch64_crypto_sha512h2qv2di): Ditto.
5941 (aarch64_crypto_sha512su0qv2di): Ditto.
5942 (aarch64_crypto_sha512su1qv2di): Ditto.
5943 (aarch64_eor3qv8hi): Ditto.
5944 (aarch64_rax1qv2di): Ditto.
5945 (aarch64_xarqv2di): Ditto.
5946 (aarch64_bcaxqv8hi): Ditto.
5947 * config/aarch64/aarch64-simd.md:
5948 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
5949 (aarch64_crypto_sha512su0qv2di): Ditto.
5950 (aarch64_crypto_sha512su1qv2di): Ditto.
5951 (aarch64_eor3qv8hi): Ditto.
5952 (aarch64_rax1qv2di): Ditto.
5953 (aarch64_xarqv2di): Ditto.
5954 (aarch64_bcaxqv8hi): Ditto.
5955 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
5956 (vsha512h2q_u64): Ditto.
5957 (vsha512su0q_u64): Ditto.
5958 (vsha512su1q_u64): Ditto.
5959 (veor3q_u16): Ditto.
5960 (vrax1q_u64): Ditto.
5962 (vbcaxq_u16): Ditto.
5963 * config/arm/types.md (crypto_sha512): New type attribute.
5964 (crypto_sha3): Ditto.
5965 (doc/invoke.texi): Document new sha3 option.
5967 2018-01-10 Michael Collison <michael.collison@arm.com>
5969 * config/aarch64/aarch64-builtins.c:
5970 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
5971 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5972 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
5973 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
5974 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
5975 (AARCH64_ISA_SM4): New ISA flag.
5976 (TARGET_SM4): New feature flag for sm4.
5977 * config/aarch64/aarch64-simd-builtins.def
5978 (aarch64_sm3ss1qv4si): Ditto.
5979 (aarch64_sm3tt1aq4si): Ditto.
5980 (aarch64_sm3tt1bq4si): Ditto.
5981 (aarch64_sm3tt2aq4si): Ditto.
5982 (aarch64_sm3tt2bq4si): Ditto.
5983 (aarch64_sm3partw1qv4si): Ditto.
5984 (aarch64_sm3partw2qv4si): Ditto.
5985 (aarch64_sm4eqv4si): Ditto.
5986 (aarch64_sm4ekeyqv4si): Ditto.
5987 * config/aarch64/aarch64-simd.md:
5988 (aarch64_sm3ss1qv4si): Ditto.
5989 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
5990 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
5991 (aarch64_sm4eqv4si): Ditto.
5992 (aarch64_sm4ekeyqv4si): Ditto.
5993 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
5994 (sm3part_op): Ditto.
5995 (CRYPTO_SM3TT): Ditto.
5996 (CRYPTO_SM3PART): Ditto.
5997 (UNSPEC_SM3SS1): New unspec.
5998 (UNSPEC_SM3TT1A): Ditto.
5999 (UNSPEC_SM3TT1B): Ditto.
6000 (UNSPEC_SM3TT2A): Ditto.
6001 (UNSPEC_SM3TT2B): Ditto.
6002 (UNSPEC_SM3PARTW1): Ditto.
6003 (UNSPEC_SM3PARTW2): Ditto.
6004 (UNSPEC_SM4E): Ditto.
6005 (UNSPEC_SM4EKEY): Ditto.
6006 * config/aarch64/constraints.md (Ui2): New constraint.
6007 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
6008 * config/arm/types.md (crypto_sm3): New type attribute.
6009 (crypto_sm4): Ditto.
6010 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
6011 (vsm3tt1aq_u32): Ditto.
6012 (vsm3tt1bq_u32): Ditto.
6013 (vsm3tt2aq_u32): Ditto.
6014 (vsm3tt2bq_u32): Ditto.
6015 (vsm3partw1q_u32): Ditto.
6016 (vsm3partw2q_u32): Ditto.
6017 (vsm4eq_u32): Ditto.
6018 (vsm4ekeyq_u32): Ditto.
6019 (doc/invoke.texi): Document new sm4 option.
6021 2018-01-10 Michael Collison <michael.collison@arm.com>
6023 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
6024 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
6025 (AARCH64_FL_FOR_ARCH8_4): New.
6026 (AARCH64_FL_V8_4): New flag.
6027 (doc/invoke.texi): Document new armv8.4-a option.
6029 2018-01-10 Michael Collison <michael.collison@arm.com>
6031 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6032 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
6033 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
6034 * config/aarch64/aarch64-option-extension.def: Add
6035 AARCH64_OPT_EXTENSION of 'sha2'.
6036 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
6037 (crypto): Disable sha2 and aes if crypto disabled.
6038 (crypto): Enable aes and sha2 if enabled.
6039 (simd): Disable sha2 and aes if simd disabled.
6040 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
6042 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
6043 (TARGET_SHA2): New feature flag for sha2.
6044 (TARGET_AES): New feature flag for aes.
6045 * config/aarch64/aarch64-simd.md:
6046 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
6047 conditional on TARGET_AES.
6048 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
6049 (aarch64_crypto_sha1hsi): Make pattern conditional
6051 (aarch64_crypto_sha1hv4si): Ditto.
6052 (aarch64_be_crypto_sha1hv4si): Ditto.
6053 (aarch64_crypto_sha1su1v4si): Ditto.
6054 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
6055 (aarch64_crypto_sha1su0v4si): Ditto.
6056 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
6057 (aarch64_crypto_sha256su0v4si): Ditto.
6058 (aarch64_crypto_sha256su1v4si): Ditto.
6059 (doc/invoke.texi): Document new aes and sha2 options.
6061 2018-01-10 Martin Sebor <msebor@redhat.com>
6063 PR tree-optimization/83781
6064 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
6067 2018-01-11 Martin Sebor <msebor@gmail.com>
6068 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6070 PR tree-optimization/83501
6071 PR tree-optimization/81703
6073 * tree-ssa-strlen.c (get_string_cst): Rename...
6074 (get_string_len): ...to this. Handle global constants.
6075 (handle_char_store): Adjust.
6077 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
6078 Jim Wilson <jimw@sifive.com>
6080 * config/riscv/riscv-protos.h (riscv_output_return): New.
6081 * config/riscv/riscv.c (struct machine_function): New naked_p field.
6082 (riscv_attribute_table, riscv_output_return),
6083 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
6084 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
6085 (riscv_compute_frame_info): Only compute frame->mask if not a naked
6087 (riscv_expand_prologue): Add early return for naked function.
6088 (riscv_expand_epilogue): Likewise.
6089 (riscv_function_ok_for_sibcall): Return false for naked function.
6090 (riscv_set_current_function): New.
6091 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
6092 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
6093 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
6094 * doc/extend.texi (RISC-V Function Attributes): New.
6096 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6098 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
6099 check for 128-bit long double before checking TCmode.
6100 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
6101 128-bit long doubles before checking TFmode or TCmode.
6102 (FLOAT128_IBM_P): Likewise.
6104 2018-01-10 Martin Sebor <msebor@redhat.com>
6106 PR tree-optimization/83671
6107 * builtins.c (c_strlen): Unconditionally return zero for the empty
6109 Use -Warray-bounds for warnings.
6110 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
6111 for non-constant array indices with COMPONENT_REF, arrays of
6112 arrays, and pointers to arrays.
6113 (gimple_fold_builtin_strlen): Determine and set length range for
6114 non-constant character arrays.
6116 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
6119 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
6122 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
6124 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
6126 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6129 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
6130 VECTOR_MEM_ALTIVEC_OR_VSX_P.
6131 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
6132 indexed_or_indirect_operand predicate.
6133 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
6134 (*vsx_le_perm_load_v8hi): Likewise.
6135 (*vsx_le_perm_load_v16qi): Likewise.
6136 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
6137 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
6138 (*vsx_le_perm_store_v8hi): Likewise.
6139 (*vsx_le_perm_store_v16qi): Likewise.
6140 (eight unnamed splitters): Likewise.
6142 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6144 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
6145 * config/rs6000/emmintrin.h: Likewise.
6146 * config/rs6000/mmintrin.h: Likewise.
6147 * config/rs6000/xmmintrin.h: Likewise.
6149 2018-01-10 David Malcolm <dmalcolm@redhat.com>
6152 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
6154 * tree.c (tree_nop_conversion): Return true for location wrapper
6156 (maybe_wrap_with_location): New function.
6157 (selftest::check_strip_nops): New function.
6158 (selftest::test_location_wrappers): New function.
6159 (selftest::tree_c_tests): Call it.
6160 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
6161 (maybe_wrap_with_location): New decl.
6162 (EXPR_LOCATION_WRAPPER_P): New macro.
6163 (location_wrapper_p): New inline function.
6164 (tree_strip_any_location_wrapper): New inline function.
6166 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
6169 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
6170 stack_realign_offset for the largest alignment of stack slot
6172 (ix86_find_max_used_stack_alignment): New function.
6173 (ix86_finalize_stack_frame_flags): Use it. Set
6174 max_used_stack_alignment if we don't realign stack.
6175 * config/i386/i386.h (machine_function): Add
6176 max_used_stack_alignment.
6178 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
6180 * config/arm/arm.opt (-mbranch-cost): New option.
6181 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
6184 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
6187 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
6188 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
6190 2018-01-10 Richard Biener <rguenther@suse.de>
6193 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
6194 early out so it also covers the case where we have a non-NULL
6197 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6199 PR tree-optimization/83753
6200 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
6201 for non-strided grouped accesses if the number of elements is 1.
6203 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6206 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
6207 * i386.h (TARGET_USE_GATHER): Define.
6208 * x86-tune.def (X86_TUNE_USE_GATHER): New.
6210 2018-01-10 Martin Liska <mliska@suse.cz>
6213 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
6214 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
6216 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
6217 CLEANUP_NO_PARTITIONING is not set.
6219 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6221 * doc/rtl.texi: Remove documentation of (const ...) wrappers
6222 for vectors, as a partial revert of r254296.
6223 * rtl.h (const_vec_p): Delete.
6224 (const_vec_duplicate_p): Don't test for vector CONSTs.
6225 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
6226 * expmed.c (make_tree): Likewise.
6229 * common.md (E, F): Use CONSTANT_P instead of checking for
6231 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
6232 checking for CONST_VECTOR.
6234 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6237 * predict.c (force_edge_cold): Handle in more sane way edges
6240 2018-01-09 Carl Love <cel@us.ibm.com>
6242 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
6244 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
6245 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
6246 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
6247 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
6248 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
6249 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
6250 * config/rs6000/rs6000-protos.h: Add extern defition for
6251 rs6000_generate_float2_double_code.
6252 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
6254 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
6255 (float2_v2df): Add define_expand.
6257 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
6260 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
6261 op_mode in the force_to_mode call.
6263 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6265 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
6266 instead of checking each element individually.
6267 (aarch64_evpc_uzp): Likewise.
6268 (aarch64_evpc_zip): Likewise.
6269 (aarch64_evpc_ext): Likewise.
6270 (aarch64_evpc_rev): Likewise.
6271 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
6272 instead of checking each element individually. Return true without
6274 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
6275 whether all selected elements come from the same input, instead of
6276 checking each element individually. Remove calls to gen_rtx_REG,
6277 start_sequence and end_sequence and instead assert that no rtl is
6280 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6282 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
6283 order of HIGH and CONST checks.
6285 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6287 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
6288 if the destination isn't an SSA_NAME.
6290 2018-01-09 Richard Biener <rguenther@suse.de>
6292 PR tree-optimization/83668
6293 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
6295 (canonicalize_loop_form): ... here, renamed from ...
6296 (canonicalize_loop_closed_ssa_form): ... this and amended to
6297 swap successor edges for loop exit blocks to make us use
6298 the RPO order we need for initial schedule generation.
6300 2018-01-09 Joseph Myers <joseph@codesourcery.com>
6302 PR tree-optimization/64811
6303 * match.pd: When optimizing comparisons with Inf, avoid
6304 introducing or losing exceptions from comparisons with NaN.
6306 2018-01-09 Martin Liska <mliska@suse.cz>
6309 * asan.c (shadow_mem_size): Add gcc_assert.
6311 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
6313 Don't save registers in main().
6316 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
6317 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
6318 * config/avr/avr.c (avr_set_current_function): Don't error if
6319 naked, OS_task or OS_main are specified at the same time.
6320 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
6322 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
6324 * common/config/avr/avr-common.c (avr_option_optimization_table):
6325 Switch on -mmain-is-OS_task for optimizing compilations.
6327 2018-01-09 Richard Biener <rguenther@suse.de>
6329 PR tree-optimization/83572
6330 * graphite.c: Include cfganal.h.
6331 (graphite_transform_loops): Connect infinite loops to exit
6332 and remove fake edges at the end.
6334 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6336 * ipa-inline.c (edge_badness): Revert accidental checkin.
6338 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6341 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
6342 symbols; not inline clones.
6344 2018-01-09 Jakub Jelinek <jakub@redhat.com>
6347 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
6348 hard registers. Formatting fixes.
6350 PR preprocessor/83722
6351 * gcc.c (try_generate_repro): Pass
6352 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
6353 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
6356 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
6357 Kito Cheng <kito.cheng@gmail.com>
6359 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
6360 (riscv_leaf_function_p): Delete.
6361 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
6363 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6365 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
6367 (do_ifelse): New function.
6368 (do_isel): New function.
6369 (do_sub3): New function.
6370 (do_add3): New function.
6371 (do_load_mask_compare): New function.
6372 (do_overlap_load_compare): New function.
6373 (expand_compare_loop): New function.
6374 (expand_block_compare): Call expand_compare_loop() when appropriate.
6375 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
6377 (-mblock-compare-inline-loop-limit): New option.
6379 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6382 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
6383 Reverse order of second and third operands in first alternative.
6384 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
6385 of first and second elements in UNSPEC_VPERMR vector.
6386 (altivec_expand_vec_perm_le): Likewise.
6388 2018-01-08 Jeff Law <law@redhat.com>
6390 PR rtl-optimizatin/81308
6391 * tree-switch-conversion.c (cfg_altered): New file scoped static.
6392 (process_switch): If group_case_labels makes a change, then set
6394 (pass_convert_switch::execute): If a switch is converted, then
6395 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
6397 PR rtl-optimization/81308
6398 * recog.c (split_all_insns): Conditionally cleanup the CFG after
6401 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
6403 PR target/83663 - Revert r255946
6404 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
6405 generation for cases where splatting a value is not useful.
6406 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
6407 across a vec_duplicate and a paradoxical subreg forming a vector
6408 mode to a vec_concat.
6410 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6412 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
6413 -march=armv8.3-a variants.
6414 * config/arm/t-multilib: Likewise.
6415 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
6417 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6419 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
6421 (cceq_ior_compare_complement): Give it a name so I can use it, and
6422 change boolean_or_operator predicate to boolean_operator so it can
6423 be used to generate a crand.
6424 (eqne): New code iterator.
6425 (bd/bd_neg): New code_attrs.
6426 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
6427 a single define_insn.
6428 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
6429 decrement (bdnzt/bdnzf/bdzt/bdzf).
6430 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
6431 with the new names of the branch decrement patterns, and added the
6432 names of the branch decrement conditional patterns.
6434 2018-01-08 Richard Biener <rguenther@suse.de>
6436 PR tree-optimization/83563
6437 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
6440 2018-01-08 Richard Biener <rguenther@suse.de>
6443 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
6445 2018-01-08 Richard Biener <rguenther@suse.de>
6447 PR tree-optimization/83685
6448 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
6449 references to abnormals.
6451 2018-01-08 Richard Biener <rguenther@suse.de>
6454 * dwarf2out.c (output_indirect_strings): Handle empty
6455 skeleton_debug_str_hash.
6456 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
6458 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6460 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
6461 (emit_store_direct): Likewise.
6462 (arc_trampoline_adjust_address): Likewise.
6463 (arc_asm_trampoline_template): New function.
6464 (arc_initialize_trampoline): Use asm_trampoline_template.
6465 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
6466 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
6467 * config/arc/arc.md (flush_icache): Delete pattern.
6469 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6471 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
6472 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
6475 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6478 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
6479 by not USED_FOR_TARGET.
6480 (make_pass_resolve_sw_modes): Likewise.
6482 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6484 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
6487 2018-01-08 Richard Biener <rguenther@suse.de>
6490 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
6492 2018-01-08 Richard Biener <rguenther@suse.de>
6495 * match.pd ((t * 2) / 2) -> t): Add missing :c.
6497 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
6500 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
6501 basic blocks with a small number of successors.
6502 (convert_control_dep_chain_into_preds): Improve handling of
6504 (dump_predicates): Split apart into...
6505 (dump_pred_chain): ...here...
6506 (dump_pred_info): ...and here.
6507 (can_one_predicate_be_invalidated_p): Add debugging printfs.
6508 (can_chain_union_be_invalidated_p): Improve check for invalidation
6510 (uninit_uses_cannot_happen): Avoid unnecessary if
6511 convert_control_dep_chain_into_preds yielded nothing.
6513 2018-01-06 Martin Sebor <msebor@redhat.com>
6515 PR tree-optimization/83640
6516 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
6517 subtracting negative offset from size.
6518 (builtin_access::overlap): Adjust offset bounds of the access to fall
6519 within the size of the object if possible.
6521 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
6523 PR rtl-optimization/83699
6524 * expmed.c (extract_bit_field_1): Restrict the vector usage of
6525 extract_bit_field_as_subreg to cases in which the extracted
6526 value is also a vector.
6528 * lra-constraints.c (process_alt_operands): Test for the equivalence
6529 substitutions when detecting a possible reload cycle.
6531 2018-01-06 Jakub Jelinek <jakub@redhat.com>
6534 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
6535 by default if flag_selective_schedling{,2}. Formatting fixes.
6537 PR rtl-optimization/83682
6538 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
6539 if it has non-VECTOR_MODE element mode.
6540 (vec_duplicate_p): Likewise.
6543 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
6544 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
6546 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6549 * config/i386/i386-builtin.def
6550 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
6551 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
6552 Require also OPTION_MASK_ISA_AVX512F in addition to
6553 OPTION_MASK_ISA_GFNI.
6554 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
6555 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
6556 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
6557 to OPTION_MASK_ISA_GFNI.
6558 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
6559 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
6560 OPTION_MASK_ISA_AVX512BW.
6561 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
6562 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
6563 addition to OPTION_MASK_ISA_GFNI.
6564 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
6565 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
6566 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
6567 to OPTION_MASK_ISA_GFNI.
6568 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
6569 a requirement for all ISAs rather than any of them with a few
6571 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
6573 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
6574 bitmasks to be enabled with 3 exceptions, instead of requiring any
6575 enabled ISA with lots of exceptions.
6576 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
6577 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
6578 Change avx512bw in isa attribute to avx512f.
6579 * config/i386/sgxintrin.h: Add license boilerplate.
6580 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
6581 to __AVX512F__ and __AVX512VL to __AVX512VL__.
6582 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
6583 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
6585 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
6586 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
6587 temporarily sse2 rather than sse if not enabled already.
6590 * config/i386/sse.md (VI248_VLBW): Rename to ...
6591 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
6592 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
6593 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
6594 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
6595 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
6596 mode iterator instead of VI248_VLBW.
6598 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
6600 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
6601 (record_modified): Skip clobbers; add debug output.
6602 (param_change_prob): Use sreal frequencies.
6604 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6606 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
6607 punt for user-aligned variables.
6609 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6611 * tree-chrec.c (chrec_contains_symbols): Return true for
6614 2018-01-05 Sudakshina Das <sudi.das@arm.com>
6617 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
6618 of (x|y) == x for BICS pattern.
6620 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6622 PR tree-optimization/83605
6623 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
6624 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
6627 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
6629 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
6630 * config/epiphany/rtems.h: New file.
6632 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6633 Uros Bizjak <ubizjak@gmail.com>
6636 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
6637 QIreg_operand instead of register_operand predicate.
6638 * config/i386/i386.c (ix86_rop_should_change_byte_p,
6639 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
6640 comments instead of -fmitigate[-_]rop.
6642 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6645 * cgraphunit.c (symbol_table::compile): Switch to text_section
6646 before calling assembly_start debug hook.
6647 * run-rtl-passes.c (run_rtl_passes): Likewise.
6650 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6652 * tree-vrp.c (extract_range_from_binary_expr_1): Check
6653 range_int_cst_p rather than !symbolic_range_p before calling
6654 extract_range_from_multiplicative_op_1.
6656 2018-01-04 Jeff Law <law@redhat.com>
6658 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
6659 redundant test in assertion.
6661 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6663 * doc/rtl.texi: Document machine_mode wrapper classes.
6665 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6667 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
6670 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6672 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
6673 the VEC_PERM_EXPR fold to fail.
6675 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6678 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
6679 to switched_sections.
6681 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6684 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
6687 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
6690 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
6691 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
6693 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6696 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
6697 is BLKmode and bitpos not zero or mode change is needed.
6699 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6702 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
6705 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
6708 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
6709 instead of MULT rtx. Update all corresponding splitters.
6711 (*ssub<modesuffix>): Ditto.
6713 (*cmp_sadd_di): Update split patterns.
6714 (*cmp_sadd_si): Ditto.
6715 (*cmp_sadd_sidi): Ditto.
6716 (*cmp_ssub_di): Ditto.
6717 (*cmp_ssub_si): Ditto.
6718 (*cmp_ssub_sidi): Ditto.
6719 * config/alpha/predicates.md (const23_operand): New predicate.
6720 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
6721 Look for ASHIFT, not MULT inner operand.
6722 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
6724 2018-01-04 Martin Liska <mliska@suse.cz>
6726 PR gcov-profile/83669
6727 * gcov.c (output_intermediate_file): Add version to intermediate
6729 * doc/gcov.texi: Document new field 'version' in intermediate
6730 file format. Fix location of '-k' option of gcov command.
6732 2018-01-04 Martin Liska <mliska@suse.cz>
6735 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
6737 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6739 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
6741 2018-01-03 Martin Sebor <msebor@redhat.com>
6743 PR tree-optimization/83655
6744 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
6745 checking calls with invalid arguments.
6747 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6749 * tree-vect-stmts.c (vect_get_store_rhs): New function.
6750 (vectorizable_mask_load_store): Delete.
6751 (vectorizable_call): Return false for masked loads and stores.
6752 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
6753 instead of gimple_assign_rhs1.
6754 (vectorizable_load): Handle IFN_MASK_LOAD.
6755 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
6757 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6759 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
6761 (vectorizable_mask_load_store): ...here.
6762 (vectorizable_load): ...and here.
6764 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6766 * tree-vect-stmts.c (vect_build_all_ones_mask)
6767 (vect_build_zero_merge_argument): New functions, split out from...
6768 (vectorizable_load): ...here.
6770 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6772 * tree-vect-stmts.c (vect_check_store_rhs): New function,
6774 (vectorizable_mask_load_store): ...here.
6775 (vectorizable_store): ...and here.
6777 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6779 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
6781 (vectorizable_mask_load_store): ...here.
6783 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6785 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
6786 (vect_model_store_cost): Take a vec_load_store_type instead of a
6788 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
6789 (vect_model_store_cost): Take a vec_load_store_type instead of a
6791 (vectorizable_mask_load_store): Update accordingly.
6792 (vectorizable_store): Likewise.
6793 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
6795 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6797 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
6798 IFN_MASK_LOAD calls here rather than...
6799 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
6801 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6802 Alan Hayward <alan.hayward@arm.com>
6803 David Sherwood <david.sherwood@arm.com>
6805 * expmed.c (extract_bit_field_1): For vector extracts,
6806 fall back to extract_bit_field_as_subreg if vec_extract
6809 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6810 Alan Hayward <alan.hayward@arm.com>
6811 David Sherwood <david.sherwood@arm.com>
6813 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
6814 they are variable or constant sized.
6815 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
6816 slots for constant-sized data.
6818 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6819 Alan Hayward <alan.hayward@arm.com>
6820 David Sherwood <david.sherwood@arm.com>
6822 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
6823 handling COND_EXPRs with boolean comparisons, try to find a better
6824 basis for the mask type than the boolean itself.
6826 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6828 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
6829 is calculated and how it can be overridden.
6830 * genmodes.c (max_bitsize_mode_any_mode): New variable.
6831 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
6833 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
6836 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6837 Alan Hayward <alan.hayward@arm.com>
6838 David Sherwood <david.sherwood@arm.com>
6840 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
6841 Remove the mode argument.
6842 (aarch64_simd_valid_immediate): Remove the mode and inverse
6844 * config/aarch64/iterators.md (bitsize): New iterator.
6845 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
6846 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
6847 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
6848 aarch64_simd_valid_immediate.
6849 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
6850 (aarch64_reg_or_bic_imm): Likewise.
6851 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
6852 with an insn_type enum and msl with a modifier_type enum.
6853 Replace element_width with a scalar_mode. Change the shift
6854 to unsigned int. Add constructors for scalar_float_mode and
6855 scalar_int_mode elements.
6856 (aarch64_vect_float_const_representable_p): Delete.
6857 (aarch64_can_const_movi_rtx_p)
6858 (aarch64_simd_scalar_immediate_valid_for_move)
6859 (aarch64_simd_make_constant): Update call to
6860 aarch64_simd_valid_immediate.
6861 (aarch64_advsimd_valid_immediate_hs): New function.
6862 (aarch64_advsimd_valid_immediate): Likewise.
6863 (aarch64_simd_valid_immediate): Remove mode and inverse
6864 arguments. Rewrite to use the above. Use const_vec_duplicate_p
6865 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
6866 and aarch64_float_const_representable_p on the result.
6867 (aarch64_output_simd_mov_immediate): Remove mode argument.
6868 Update call to aarch64_simd_valid_immediate and use of
6869 simd_immediate_info.
6870 (aarch64_output_scalar_simd_mov_immediate): Update call
6873 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6874 Alan Hayward <alan.hayward@arm.com>
6875 David Sherwood <david.sherwood@arm.com>
6877 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
6878 (mode_nunits): Likewise CONST_MODE_NUNITS.
6879 * machmode.def (ADJUST_NUNITS): Document.
6880 * genmodes.c (mode_data::need_nunits_adj): New field.
6881 (blank_mode): Update accordingly.
6882 (adj_nunits): New variable.
6883 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
6885 (emit_mode_size_inline): Set need_bytesize_adj for all modes
6886 listed in adj_nunits.
6887 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
6888 listed in adj_nunits. Don't emit case statements for such modes.
6889 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
6890 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
6891 nothing if adj_nunits is nonnull.
6892 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
6893 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
6894 (emit_mode_fbit): Update use of print_maybe_const_decl.
6895 (emit_move_size): Likewise. Treat the array as non-const
6897 (emit_mode_adjustments): Handle adj_nunits.
6899 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6901 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
6902 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
6903 (VECTOR_MODES): Use it.
6904 (make_vector_modes): Take the prefix as an argument.
6906 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6907 Alan Hayward <alan.hayward@arm.com>
6908 David Sherwood <david.sherwood@arm.com>
6910 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
6911 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
6912 for MODE_VECTOR_BOOL.
6913 * machmode.def (VECTOR_BOOL_MODE): Document.
6914 * genmodes.c (VECTOR_BOOL_MODE): New macro.
6915 (make_vector_bool_mode): New function.
6916 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
6918 * lto-streamer-in.c (lto_input_mode_table): Likewise.
6919 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
6921 * stor-layout.c (int_mode_for_mode): Likewise.
6922 * tree.c (build_vector_type_for_mode): Likewise.
6923 * varasm.c (output_constant_pool_2): Likewise.
6924 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
6925 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
6926 for MODE_VECTOR_BOOL.
6927 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
6928 of mode class checks.
6929 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
6930 instead of a list of mode class checks.
6931 (expand_vector_scalar_condition): Likewise.
6932 (type_for_widest_vector_mode): Handle BImode as an inner mode.
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_size): Change from unsigned short to
6940 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
6941 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6942 or if measurement_type is not polynomial.
6943 (fixed_size_mode::includes_p): Check for constant-sized modes.
6944 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
6945 return a poly_uint16 rather than an unsigned short.
6946 (emit_mode_size): Change the type of mode_size from unsigned short
6947 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
6948 (emit_mode_adjustments): Cope with polynomial vector sizes.
6949 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6951 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6953 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
6954 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
6955 * caller-save.c (setup_save_areas): Likewise.
6956 (replace_reg_with_saved_mem): Likewise.
6957 * calls.c (emit_library_call_value_1): Likewise.
6958 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
6959 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
6960 (gen_lowpart_for_combine): Likewise.
6961 * convert.c (convert_to_integer_1): Likewise.
6962 * cse.c (equiv_constant, cse_insn): Likewise.
6963 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
6964 (cselib_subst_to_values): Likewise.
6965 * dce.c (word_dce_process_block): Likewise.
6966 * df-problems.c (df_word_lr_mark_ref): Likewise.
6967 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
6968 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
6969 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
6970 (rtl_for_decl_location): Likewise.
6971 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
6972 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
6973 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
6974 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
6975 (expand_expr_real_1): Likewise.
6976 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
6977 (pad_below): Likewise.
6978 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6979 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
6980 * ira.c (get_subreg_tracking_sizes): Likewise.
6981 * ira-build.c (ira_create_allocno_objects): Likewise.
6982 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
6983 (ira_sort_regnos_for_alter_reg): Likewise.
6984 * ira-costs.c (record_operand_costs): Likewise.
6985 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
6986 (resolve_simple_move): Likewise.
6987 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
6988 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
6989 (lra_constraints): Likewise.
6990 (CONST_POOL_OK_P): Reject variable-sized modes.
6991 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
6992 (add_pseudo_to_slot, lra_spill): Likewise.
6993 * omp-low.c (omp_clause_aligned_alignment): Likewise.
6994 * optabs-query.c (get_best_extraction_insn): Likewise.
6995 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
6996 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
6997 (expand_mult_highpart, valid_multiword_target_p): Likewise.
6998 * recog.c (offsettable_address_addr_space_p): Likewise.
6999 * regcprop.c (maybe_mode_change): Likewise.
7000 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
7001 * regrename.c (build_def_use): Likewise.
7002 * regstat.c (dump_reg_info): Likewise.
7003 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
7004 (find_reloads, find_reloads_subreg_address): Likewise.
7005 * reload1.c (eliminate_regs_1): Likewise.
7006 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
7007 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
7008 (simplify_binary_operation_1, simplify_subreg): Likewise.
7009 * targhooks.c (default_function_arg_padding): Likewise.
7010 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
7011 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
7012 (verify_gimple_assign_ternary): Likewise.
7013 * tree-inline.c (estimate_move_cost): Likewise.
7014 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7015 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
7016 (get_address_cost_ainc): Likewise.
7017 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
7018 (vect_supportable_dr_alignment): Likewise.
7019 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7020 (vectorizable_reduction): Likewise.
7021 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
7022 (vectorizable_operation, vectorizable_load): Likewise.
7023 * tree.c (build_same_sized_truth_vector_type): Likewise.
7024 * valtrack.c (cleanup_auto_inc_dec): Likewise.
7025 * var-tracking.c (emit_note_insn_var_location): Likewise.
7026 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
7027 (ADDR_VEC_ALIGN): Likewise.
7029 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7030 Alan Hayward <alan.hayward@arm.com>
7031 David Sherwood <david.sherwood@arm.com>
7033 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
7035 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7036 or if measurement_type is polynomial.
7037 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
7038 * combine.c (make_extraction): Likewise.
7039 * dse.c (find_shift_sequence): Likewise.
7040 * dwarf2out.c (mem_loc_descriptor): Likewise.
7041 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
7042 (extract_bit_field, extract_low_bits): Likewise.
7043 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
7044 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
7045 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
7046 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
7047 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7048 * reload.c (find_reloads): Likewise.
7049 * reload1.c (alter_reg): Likewise.
7050 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
7051 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
7052 * tree-if-conv.c (predicate_mem_writes): Likewise.
7053 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
7054 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
7055 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
7056 * valtrack.c (dead_debug_insert_temp): Likewise.
7057 * varasm.c (mergeable_constant_section): Likewise.
7058 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
7060 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7061 Alan Hayward <alan.hayward@arm.com>
7062 David Sherwood <david.sherwood@arm.com>
7064 * expr.c (expand_assignment): Cope with polynomial mode sizes
7065 when assigning to a CONCAT.
7067 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7068 Alan Hayward <alan.hayward@arm.com>
7069 David Sherwood <david.sherwood@arm.com>
7071 * machmode.h (mode_precision): Change from unsigned short to
7073 (mode_to_precision): Return a poly_uint16 rather than an unsigned
7075 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
7076 or if measurement_type is not polynomial.
7077 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
7078 in which the mode is already known to be a scalar_int_mode.
7079 * genmodes.c (emit_mode_precision): Change the type of mode_precision
7080 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
7082 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7083 for GET_MODE_PRECISION.
7084 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7085 for GET_MODE_PRECISION.
7086 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
7088 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
7089 (expand_field_assignment, make_extraction): Likewise.
7090 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
7091 (get_last_value): Likewise.
7092 * convert.c (convert_to_integer_1): Likewise.
7093 * cse.c (cse_insn): Likewise.
7094 * expr.c (expand_expr_real_1): Likewise.
7095 * lra-constraints.c (simplify_operand_subreg): Likewise.
7096 * optabs-query.c (can_atomic_load_p): Likewise.
7097 * optabs.c (expand_atomic_load): Likewise.
7098 (expand_atomic_store): Likewise.
7099 * ree.c (combine_reaching_defs): Likewise.
7100 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
7101 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
7102 * tree.h (type_has_mode_precision_p): Likewise.
7103 * ubsan.c (instrument_si_overflow): Likewise.
7105 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7106 Alan Hayward <alan.hayward@arm.com>
7107 David Sherwood <david.sherwood@arm.com>
7109 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
7110 polynomial numbers of units.
7111 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
7112 (valid_vector_subparts_p): New function.
7113 (build_vector_type): Remove temporary shim and take the number
7114 of units as a poly_uint64 rather than an int.
7115 (build_opaque_vector_type): Take the number of units as a
7116 poly_uint64 rather than an int.
7117 * tree.c (build_vector_from_ctor): Handle polynomial
7118 TYPE_VECTOR_SUBPARTS.
7119 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
7120 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
7121 (build_vector_from_val): If the number of units is variable,
7122 use build_vec_duplicate_cst for constant operands and
7123 VEC_DUPLICATE_EXPR otherwise.
7124 (make_vector_type): Remove temporary is_constant ().
7125 (build_vector_type, build_opaque_vector_type): Take the number of
7126 units as a poly_uint64 rather than an int.
7127 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
7129 * cfgexpand.c (expand_debug_expr): Likewise.
7130 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
7131 (store_constructor, expand_expr_real_1): Likewise.
7132 (const_scalar_mask_from_tree): Likewise.
7133 * fold-const-call.c (fold_const_reduction): Likewise.
7134 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
7135 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
7136 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
7137 (fold_relational_const): Likewise.
7138 (native_interpret_vector): Likewise. Change the size from an
7139 int to an unsigned int.
7140 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
7141 TYPE_VECTOR_SUBPARTS.
7142 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
7143 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
7144 duplicating a non-constant operand into a variable-length vector.
7145 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
7146 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
7147 * ipa-icf.c (sem_variable::equals): Likewise.
7148 * match.pd: Likewise.
7149 * omp-simd-clone.c (simd_clone_subparts): Likewise.
7150 * print-tree.c (print_node): Likewise.
7151 * stor-layout.c (layout_type): Likewise.
7152 * targhooks.c (default_builtin_vectorization_cost): Likewise.
7153 * tree-cfg.c (verify_gimple_comparison): Likewise.
7154 (verify_gimple_assign_binary): Likewise.
7155 (verify_gimple_assign_ternary): Likewise.
7156 (verify_gimple_assign_single): Likewise.
7157 * tree-pretty-print.c (dump_generic_node): Likewise.
7158 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7159 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
7160 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
7161 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
7162 (vect_shift_permute_load_chain): Likewise.
7163 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
7164 (expand_vector_condition, optimize_vector_constructor): Likewise.
7165 (lower_vec_perm, get_compute_type): Likewise.
7166 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7167 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
7168 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
7169 (vect_recog_mask_conversion_pattern): Likewise.
7170 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
7171 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
7172 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7173 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
7174 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
7175 (vectorizable_shift, vectorizable_operation, vectorizable_store)
7176 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
7177 (supportable_widening_operation): Likewise.
7178 (supportable_narrowing_operation): Likewise.
7179 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
7181 * varasm.c (output_constant): Likewise.
7183 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7184 Alan Hayward <alan.hayward@arm.com>
7185 David Sherwood <david.sherwood@arm.com>
7187 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
7188 so that both the length == 3 and length != 3 cases set up their
7189 own permute vectors. Add comments explaining why we know the
7190 number of elements is constant.
7191 (vect_permute_load_chain): Likewise.
7193 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7194 Alan Hayward <alan.hayward@arm.com>
7195 David Sherwood <david.sherwood@arm.com>
7197 * machmode.h (mode_nunits): Change from unsigned char to
7199 (ONLY_FIXED_SIZE_MODES): New macro.
7200 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
7201 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
7202 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
7204 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
7205 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
7206 or if measurement_type is not polynomial.
7207 * genmodes.c (ZERO_COEFFS): New macro.
7208 (emit_mode_nunits_inline): Make mode_nunits_inline return a
7210 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
7211 Use ZERO_COEFFS when emitting initializers.
7212 * data-streamer.h (bp_pack_poly_value): New function.
7213 (bp_unpack_poly_value): Likewise.
7214 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7215 for GET_MODE_NUNITS.
7216 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7217 for GET_MODE_NUNITS.
7218 * tree.c (make_vector_type): Remove temporary shim and make
7219 the real function take the number of units as a poly_uint64
7221 (build_vector_type_for_mode): Handle polynomial nunits.
7222 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
7223 * emit-rtl.c (const_vec_series_p_1): Likewise.
7224 (gen_rtx_CONST_VECTOR): Likewise.
7225 * fold-const.c (test_vec_duplicate_folding): Likewise.
7226 * genrecog.c (validate_pattern): Likewise.
7227 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
7228 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7229 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
7230 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
7231 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
7232 * rtlanal.c (subreg_get_info): Likewise.
7233 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7234 (vect_grouped_load_supported): Likewise.
7235 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
7236 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
7237 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
7238 (simplify_const_unary_operation, simplify_binary_operation_1)
7239 (simplify_const_binary_operation, simplify_ternary_operation)
7240 (test_vector_ops_duplicate, test_vector_ops): Likewise.
7241 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
7242 instead of CONST_VECTOR_NUNITS.
7243 * varasm.c (output_constant_pool_2): Likewise.
7244 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
7245 explicit-encoded elements in the XVEC for variable-length vectors.
7247 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7249 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
7251 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7252 Alan Hayward <alan.hayward@arm.com>
7253 David Sherwood <david.sherwood@arm.com>
7255 * coretypes.h (fixed_size_mode): Declare.
7256 (fixed_size_mode_pod): New typedef.
7257 * builtins.h (target_builtins::x_apply_args_mode)
7258 (target_builtins::x_apply_result_mode): Change type to
7259 fixed_size_mode_pod.
7260 * builtins.c (apply_args_size, apply_result_size, result_vector)
7261 (expand_builtin_apply_args_1, expand_builtin_apply)
7262 (expand_builtin_return): Update accordingly.
7264 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7266 * cse.c (hash_rtx_cb): Hash only the encoded elements.
7267 * cselib.c (cselib_hash_rtx): Likewise.
7268 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
7269 CONST_VECTOR encoding.
7271 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7272 Jeff Law <law@redhat.com>
7275 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
7276 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
7277 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
7278 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
7281 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
7282 explicitly probe *sp in a noreturn function if there were any callee
7283 register saves or frame pointer is needed.
7285 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7288 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
7289 BLKmode for ternary, binary or unary expressions.
7292 * var-tracking.c (delete_vta_debug_insn): New inline function.
7293 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
7294 insns from get_insns () to NULL instead of each bb separately.
7295 Use delete_vta_debug_insn. No longer static.
7296 (vt_debug_insns_local, variable_tracking_main_1): Adjust
7297 delete_vta_debug_insns callers.
7298 * rtl.h (delete_vta_debug_insns): Declare.
7299 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
7300 instead of variable_tracking_main.
7302 2018-01-03 Martin Sebor <msebor@redhat.com>
7304 PR tree-optimization/83603
7305 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
7306 arguments past the endof the argument list in functions declared
7307 without a prototype.
7308 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
7309 Avoid checking when arguments are null.
7311 2018-01-03 Martin Sebor <msebor@redhat.com>
7314 * doc/extend.texi (attribute const): Fix a typo.
7315 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
7316 issuing -Wsuggest-attribute for void functions.
7318 2018-01-03 Martin Sebor <msebor@redhat.com>
7320 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
7321 offset_int::from instead of wide_int::to_shwi.
7322 (maybe_diag_overlap): Remove assertion.
7323 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
7324 * gimple-ssa-sprintf.c (format_directive): Same.
7325 (parse_directive): Same.
7326 (sprintf_dom_walker::compute_format_length): Same.
7327 (try_substitute_return_value): Same.
7329 2018-01-03 Jeff Law <law@redhat.com>
7332 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
7333 non-constant residual for zero at runtime and avoid probing in
7334 that case. Reorganize code for trailing problem to mirror handling
7337 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7339 PR tree-optimization/83501
7340 * tree-ssa-strlen.c (get_string_cst): New.
7341 (handle_char_store): Call get_string_cst.
7343 2018-01-03 Martin Liska <mliska@suse.cz>
7345 PR tree-optimization/83593
7346 * tree-ssa-strlen.c: Include tree-cfg.h.
7347 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
7348 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
7349 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
7351 (strlen_dom_walker::before_dom_children): Call
7352 gimple_purge_dead_eh_edges. Dump tranformation with details
7354 (strlen_dom_walker::before_dom_children): Update call by adding
7355 new argument cleanup_eh.
7356 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
7358 2018-01-03 Martin Liska <mliska@suse.cz>
7361 * cif-code.def (VARIADIC_THUNK): New enum value.
7362 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
7365 2018-01-03 Jan Beulich <jbeulich@suse.com>
7367 * sse.md (mov<mode>_internal): Tighten condition for when to use
7368 vmovdqu<ssescalarsize> for TI and OI modes.
7370 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7372 Update copyright years.
7374 2018-01-03 Martin Liska <mliska@suse.cz>
7377 * ipa-visibility.c (function_and_variable_visibility): Skip
7378 functions with noipa attribure.
7380 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7382 * gcc.c (process_command): Update copyright notice dates.
7383 * gcov-dump.c (print_version): Ditto.
7384 * gcov.c (print_version): Ditto.
7385 * gcov-tool.c (print_version): Ditto.
7386 * gengtype.c (create_file): Ditto.
7387 * doc/cpp.texi: Bump @copying's copyright year.
7388 * doc/cppinternals.texi: Ditto.
7389 * doc/gcc.texi: Ditto.
7390 * doc/gccint.texi: Ditto.
7391 * doc/gcov.texi: Ditto.
7392 * doc/install.texi: Ditto.
7393 * doc/invoke.texi: Ditto.
7395 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7397 * vector-builder.h (vector_builder::m_full_nelts): Change from
7398 unsigned int to poly_uint64.
7399 (vector_builder::full_nelts): Update prototype accordingly.
7400 (vector_builder::new_vector): Likewise.
7401 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
7402 (vector_builder::operator ==): Likewise.
7403 (vector_builder::finalize): Likewise.
7404 * int-vector-builder.h (int_vector_builder::int_vector_builder):
7405 Take the number of elements as a poly_uint64 rather than an
7407 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
7408 from unsigned int to poly_uint64.
7409 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
7410 (vec_perm_indices::new_vector): Likewise.
7411 (vec_perm_indices::length): Likewise.
7412 (vec_perm_indices::nelts_per_input): Likewise.
7413 (vec_perm_indices::input_nelts): Likewise.
7414 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
7415 number of elements per input as a poly_uint64 rather than an
7416 unsigned int. Use the original encoding for variable-length
7417 vectors, rather than clamping each individual element.
7418 For the second and subsequent elements in each pattern,
7419 clamp the step and base before clamping their sum.
7420 (vec_perm_indices::series_p): Handle polynomial element counts.
7421 (vec_perm_indices::all_in_range_p): Likewise.
7422 (vec_perm_indices_to_tree): Likewise.
7423 (vec_perm_indices_to_rtx): Likewise.
7424 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
7425 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
7426 (tree_vector_builder::new_binary_operation): Handle polynomial
7427 element counts. Return false if we need to know the number
7428 of elements at compile time.
7429 * fold-const.c (fold_vec_perm): Punt if the number of elements
7430 isn't known at compile time.
7432 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7434 * vec-perm-indices.h (vec_perm_builder): Change element type
7435 from HOST_WIDE_INT to poly_int64.
7436 (vec_perm_indices::element_type): Update accordingly.
7437 (vec_perm_indices::clamp): Handle polynomial element_types.
7438 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7439 (vec_perm_indices::all_in_range_p): Likewise.
7440 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
7442 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
7443 polynomial vec_perm_indices element types.
7444 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
7445 * fold-const.c (fold_vec_perm): Likewise.
7446 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
7447 * tree-vect-generic.c (lower_vec_perm): Likewise.
7448 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7449 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
7450 element type to HOST_WIDE_INT.
7452 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7453 Alan Hayward <alan.hayward@arm.com>
7454 David Sherwood <david.sherwood@arm.com>
7456 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
7457 rather than an int. Use plus_constant.
7458 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
7459 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
7461 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7462 Alan Hayward <alan.hayward@arm.com>
7463 David Sherwood <david.sherwood@arm.com>
7465 * calls.c (emit_call_1, expand_call): Change struct_value_size from
7466 a HOST_WIDE_INT to a poly_int64.
7468 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7469 Alan Hayward <alan.hayward@arm.com>
7470 David Sherwood <david.sherwood@arm.com>
7472 * calls.c (load_register_parameters): Cope with polynomial
7473 mode sizes. Require a constant size for BLKmode parameters
7474 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
7475 forces a parameter to be padded at the lsb end in order to
7476 fill a complete number of words, require the parameter size
7477 to be ordered wrt UNITS_PER_WORD.
7479 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7480 Alan Hayward <alan.hayward@arm.com>
7481 David Sherwood <david.sherwood@arm.com>
7483 * reload1.c (spill_stack_slot_width): Change element type
7484 from unsigned int to poly_uint64_pod.
7485 (alter_reg): Treat mode sizes as polynomial.
7487 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7488 Alan Hayward <alan.hayward@arm.com>
7489 David Sherwood <david.sherwood@arm.com>
7491 * reload.c (complex_word_subreg_p): New function.
7492 (reload_inner_reg_of_subreg, push_reload): Use it.
7494 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7495 Alan Hayward <alan.hayward@arm.com>
7496 David Sherwood <david.sherwood@arm.com>
7498 * lra-constraints.c (process_alt_operands): Reject matched
7499 operands whose sizes aren't ordered.
7500 (match_reload): Refer to this check here.
7502 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7503 Alan Hayward <alan.hayward@arm.com>
7504 David Sherwood <david.sherwood@arm.com>
7506 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
7507 that the mode size is in the set {1, 2, 4, 8, 16}.
7509 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7510 Alan Hayward <alan.hayward@arm.com>
7511 David Sherwood <david.sherwood@arm.com>
7513 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
7514 Use plus_constant instead of gen_rtx_PLUS.
7516 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7517 Alan Hayward <alan.hayward@arm.com>
7518 David Sherwood <david.sherwood@arm.com>
7520 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
7521 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
7522 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
7523 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
7524 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
7525 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
7526 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
7527 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
7528 * config/i386/i386.c (ix86_push_rounding): ...this new function.
7529 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
7531 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
7532 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
7533 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
7534 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
7535 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
7536 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
7537 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
7538 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
7539 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
7540 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
7542 * expr.c (emit_move_resolve_push): Treat the input and result
7543 of PUSH_ROUNDING as a poly_int64.
7544 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
7545 (emit_push_insn): Likewise.
7546 * lra-eliminations.c (mark_not_eliminable): Likewise.
7547 * recog.c (push_operand): Likewise.
7548 * reload1.c (elimination_effects): Likewise.
7549 * rtlanal.c (nonzero_bits1): Likewise.
7550 * calls.c (store_one_arg): Likewise. Require the padding to be
7551 known at compile time.
7553 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7554 Alan Hayward <alan.hayward@arm.com>
7555 David Sherwood <david.sherwood@arm.com>
7557 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
7558 Use plus_constant instead of gen_rtx_PLUS.
7560 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7561 Alan Hayward <alan.hayward@arm.com>
7562 David Sherwood <david.sherwood@arm.com>
7564 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
7567 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7568 Alan Hayward <alan.hayward@arm.com>
7569 David Sherwood <david.sherwood@arm.com>
7571 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
7572 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
7573 via stack temporaries. Treat the mode size as polynomial too.
7575 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7576 Alan Hayward <alan.hayward@arm.com>
7577 David Sherwood <david.sherwood@arm.com>
7579 * expr.c (expand_expr_real_2): When handling conversions involving
7580 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
7581 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
7582 as a poly_uint64 too.
7584 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7585 Alan Hayward <alan.hayward@arm.com>
7586 David Sherwood <david.sherwood@arm.com>
7588 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
7590 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7591 Alan Hayward <alan.hayward@arm.com>
7592 David Sherwood <david.sherwood@arm.com>
7594 * combine.c (can_change_dest_mode): Handle polynomial
7595 REGMODE_NATURAL_SIZE.
7596 * expmed.c (store_bit_field_1): Likewise.
7597 * expr.c (store_constructor): Likewise.
7598 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
7599 and polynomial REGMODE_NATURAL_SIZE.
7600 (gen_lowpart_common): Likewise.
7601 * reginfo.c (record_subregs_of_mode): Likewise.
7602 * rtlanal.c (read_modify_subreg_p): Likewise.
7604 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7605 Alan Hayward <alan.hayward@arm.com>
7606 David Sherwood <david.sherwood@arm.com>
7608 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
7609 numbers of elements.
7611 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7612 Alan Hayward <alan.hayward@arm.com>
7613 David Sherwood <david.sherwood@arm.com>
7615 * match.pd: Cope with polynomial numbers of vector elements.
7617 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7618 Alan Hayward <alan.hayward@arm.com>
7619 David Sherwood <david.sherwood@arm.com>
7621 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
7622 in a POINTER_PLUS_EXPR.
7624 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7625 Alan Hayward <alan.hayward@arm.com>
7626 David Sherwood <david.sherwood@arm.com>
7628 * omp-simd-clone.c (simd_clone_subparts): New function.
7629 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
7630 (ipa_simd_modify_function_body): Likewise.
7632 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7633 Alan Hayward <alan.hayward@arm.com>
7634 David Sherwood <david.sherwood@arm.com>
7636 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
7637 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
7638 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
7639 (expand_vector_condition, vector_element): Likewise.
7640 (subparts_gt): New function.
7641 (get_compute_type): Use subparts_gt.
7642 (count_type_subparts): Delete.
7643 (expand_vector_operations_1): Use subparts_gt instead of
7644 count_type_subparts.
7646 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7647 Alan Hayward <alan.hayward@arm.com>
7648 David Sherwood <david.sherwood@arm.com>
7650 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
7651 (vect_compile_time_alias): ...this new function. Do the calculation
7652 on poly_ints rather than trees.
7653 (vect_prune_runtime_alias_test_list): Update call accordingly.
7655 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7656 Alan Hayward <alan.hayward@arm.com>
7657 David Sherwood <david.sherwood@arm.com>
7659 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
7661 (vect_schedule_slp_instance): Likewise.
7663 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7664 Alan Hayward <alan.hayward@arm.com>
7665 David Sherwood <david.sherwood@arm.com>
7667 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
7668 constant and extern definitions for variable-length vectors.
7669 (vect_get_constant_vectors): Note that the number of units
7670 is known to be constant.
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 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
7677 of units as polynomial. Choose between WIDE and NARROW based
7680 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7681 Alan Hayward <alan.hayward@arm.com>
7682 David Sherwood <david.sherwood@arm.com>
7684 * tree-vect-stmts.c (simd_clone_subparts): New function.
7685 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
7687 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7688 Alan Hayward <alan.hayward@arm.com>
7689 David Sherwood <david.sherwood@arm.com>
7691 * tree-vect-stmts.c (vectorizable_call): Treat the number of
7692 vectors as polynomial. Use build_index_vector for
7695 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7696 Alan Hayward <alan.hayward@arm.com>
7697 David Sherwood <david.sherwood@arm.com>
7699 * tree-vect-stmts.c (get_load_store_type): Treat the number of
7700 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
7701 for variable-length vectors.
7702 (vectorizable_mask_load_store): Treat the number of units as
7703 polynomial, asserting that it is constant if the condition has
7704 already been enforced.
7705 (vectorizable_store, vectorizable_load): Likewise.
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-loop.c (vectorizable_live_operation): Treat the number
7712 of units as polynomial. Punt if we can't tell at compile time
7713 which vector contains the final result.
7715 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7716 Alan Hayward <alan.hayward@arm.com>
7717 David Sherwood <david.sherwood@arm.com>
7719 * tree-vect-loop.c (vectorizable_induction): Treat the number
7720 of units as polynomial. Punt on SLP inductions. Use an integer
7721 VEC_SERIES_EXPR for variable-length integer reductions. Use a
7722 cast of such a series for variable-length floating-point
7725 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7726 Alan Hayward <alan.hayward@arm.com>
7727 David Sherwood <david.sherwood@arm.com>
7729 * tree.h (build_index_vector): Declare.
7730 * tree.c (build_index_vector): New function.
7731 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
7732 of units as polynomial, forcibly converting it to a constant if
7733 vectorizable_reduction has already enforced the condition.
7734 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
7735 to create a {1,2,3,...} vector.
7736 (vectorizable_reduction): Treat the number of units as polynomial.
7737 Choose vectype_in based on the largest scalar element size rather
7738 than the smallest number of units. Enforce the restrictions
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-data-refs.c (vector_alignment_reachable_p): Treat the
7746 number of units as polynomial.
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 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
7753 * target.def (autovectorize_vector_sizes): Return the vector sizes
7754 by pointer, using vector_sizes rather than a bitmask.
7755 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
7756 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
7757 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
7759 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
7760 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
7761 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
7762 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
7763 * omp-general.c (omp_max_vf): Likewise.
7764 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7765 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
7766 * tree-vect-loop.c (vect_analyze_loop): Likewise.
7767 * tree-vect-slp.c (vect_slp_bb): Likewise.
7768 * doc/tm.texi: Regenerate.
7769 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
7771 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
7772 the vector size as a poly_uint64 rather than an unsigned int.
7773 (current_vector_size): Change from an unsigned int to a poly_uint64.
7774 (get_vectype_for_scalar_type): Update accordingly.
7775 * tree.h (build_truth_vector_type): Take the size and number of
7776 units as a poly_uint64 rather than an unsigned int.
7777 (build_vector_type): Add a temporary overload that takes
7778 the number of units as a poly_uint64 rather than an unsigned int.
7779 * tree.c (make_vector_type): Likewise.
7780 (build_truth_vector_type): Take the number of units as a poly_uint64
7781 rather than an unsigned int.
7783 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7784 Alan Hayward <alan.hayward@arm.com>
7785 David Sherwood <david.sherwood@arm.com>
7787 * target.def (get_mask_mode): Take the number of units and length
7788 as poly_uint64s rather than unsigned ints.
7789 * targhooks.h (default_get_mask_mode): Update accordingly.
7790 * targhooks.c (default_get_mask_mode): Likewise.
7791 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
7792 * doc/tm.texi: Regenerate.
7794 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7795 Alan Hayward <alan.hayward@arm.com>
7796 David Sherwood <david.sherwood@arm.com>
7798 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
7799 * omp-general.c (omp_max_vf): Likewise.
7800 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
7801 (expand_omp_simd): Handle polynomial safelen.
7802 * omp-low.c (omplow_simd_context): Add a default constructor.
7803 (omplow_simd_context::max_vf): Change from int to poly_uint64.
7804 (lower_rec_simd_input_clauses): Update accordingly.
7805 (lower_rec_input_clauses): Likewise.
7807 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7808 Alan Hayward <alan.hayward@arm.com>
7809 David Sherwood <david.sherwood@arm.com>
7811 * tree-vectorizer.h (vect_nunits_for_cost): New function.
7812 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
7813 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
7814 (vect_analyze_slp_cost): Likewise.
7815 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
7816 (vect_model_load_cost): Likewise.
7818 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7819 Alan Hayward <alan.hayward@arm.com>
7820 David Sherwood <david.sherwood@arm.com>
7822 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
7823 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
7824 from an unsigned int * to a poly_uint64_pod *.
7825 (calculate_unrolling_factor): New function.
7826 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
7828 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7829 Alan Hayward <alan.hayward@arm.com>
7830 David Sherwood <david.sherwood@arm.com>
7832 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
7833 from an unsigned int to a poly_uint64.
7834 (_loop_vec_info::slp_unrolling_factor): Likewise.
7835 (_loop_vec_info::vectorization_factor): Change from an int
7837 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
7838 (vect_get_num_vectors): New function.
7839 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
7840 (vect_get_num_copies): Use vect_get_num_vectors.
7841 (vect_analyze_data_ref_dependences): Change max_vf from an int *
7842 to an unsigned int *.
7843 (vect_analyze_data_refs): Change min_vf from an int * to a
7845 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7846 than an unsigned HOST_WIDE_INT.
7847 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
7848 (vect_analyze_data_ref_dependence): Change max_vf from an int *
7849 to an unsigned int *.
7850 (vect_analyze_data_ref_dependences): Likewise.
7851 (vect_compute_data_ref_alignment): Handle polynomial vf.
7852 (vect_enhance_data_refs_alignment): Likewise.
7853 (vect_prune_runtime_alias_test_list): Likewise.
7854 (vect_shift_permute_load_chain): Likewise.
7855 (vect_supportable_dr_alignment): Likewise.
7856 (dependence_distance_ge_vf): Take the vectorization factor as a
7857 poly_uint64 rather than an unsigned HOST_WIDE_INT.
7858 (vect_analyze_data_refs): Change min_vf from an int * to a
7860 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
7861 vfm1 as a poly_uint64 rather than an int. Make the same change
7862 for the returned bound_scalar.
7863 (vect_gen_vector_loop_niters): Handle polynomial vf.
7864 (vect_do_peeling): Likewise. Update call to
7865 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
7866 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
7868 * tree-vect-loop.c (vect_determine_vectorization_factor)
7869 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
7870 (vect_get_known_peeling_cost): Likewise.
7871 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
7872 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
7873 (vect_transform_loop): Likewise. Use the lowest possible VF when
7874 updating the upper bounds of the loop.
7875 (vect_min_worthwhile_factor): Make static. Return an unsigned int
7877 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
7878 polynomial unroll factors.
7879 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
7880 (vect_make_slp_decision): Likewise.
7881 (vect_supported_load_permutation_p): Likewise, and polynomial
7883 (vect_analyze_slp_cost): Handle polynomial vf.
7884 (vect_slp_analyze_node_operations): Likewise.
7885 (vect_slp_analyze_bb_1): Likewise.
7886 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7887 than an unsigned HOST_WIDE_INT.
7888 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
7889 (vectorizable_load): Handle polynomial vf.
7890 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
7892 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
7894 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7895 Alan Hayward <alan.hayward@arm.com>
7896 David Sherwood <david.sherwood@arm.com>
7898 * match.pd: Handle bit operations involving three constants
7899 and try to fold one pair.
7901 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7903 * tree-vect-loop-manip.c: Include gimple-fold.h.
7904 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
7905 niters_maybe_zero parameters. Handle other cases besides a step of 1.
7906 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
7907 Add a path that uses a step of VF instead of 1, but disable it
7909 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
7910 and niters_no_overflow parameters. Update calls to
7911 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
7912 Create a new SSA name if the latter choses to use a ste other
7913 than zero, and return it via niters_vector_mult_vf_var.
7914 * tree-vect-loop.c (vect_transform_loop): Update calls to
7915 vect_do_peeling, vect_gen_vector_loop_niters and
7916 slpeel_make_loop_iterate_ntimes.
7917 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
7918 (vect_gen_vector_loop_niters): Update declarations after above changes.
7920 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
7922 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
7923 128-bit round to integer instructions.
7924 (ceil<mode>2): Likewise.
7925 (btrunc<mode>2): Likewise.
7926 (round<mode>2): Likewise.
7928 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7930 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
7931 unaligned VSX load/store on P8/P9.
7932 (expand_block_clear): Allow the use of unaligned VSX
7933 load/store on P8/P9.
7935 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7937 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
7939 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
7940 swap associated with both a load and a store.
7942 2018-01-02 Andrew Waterman <andrew@sifive.com>
7944 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
7945 * config/riscv/riscv.md (clear_cache): Use it.
7947 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
7949 * web.c: Remove out-of-date comment.
7951 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7953 * expr.c (fixup_args_size_notes): Check that any existing
7954 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
7955 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
7956 (emit_single_push_insn): ...here.
7958 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7960 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
7961 (const_vector_encoded_nelts): New function.
7962 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
7963 (const_vector_int_elt, const_vector_elt): Declare.
7964 * emit-rtl.c (const_vector_int_elt_1): New function.
7965 (const_vector_elt): Likewise.
7966 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
7967 of CONST_VECTOR_ELT.
7969 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7971 * expr.c: Include rtx-vector-builder.h.
7972 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
7973 directly on the tree encoding.
7974 (const_vector_from_tree): Likewise.
7975 * optabs.c: Include rtx-vector-builder.h.
7976 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
7977 sequence of "u" values.
7978 * vec-perm-indices.c: Include rtx-vector-builder.h.
7979 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
7980 directly on the vec_perm_indices encoding.
7982 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7984 * doc/rtl.texi (const_vector): Describe new encoding scheme.
7985 * Makefile.in (OBJS): Add rtx-vector-builder.o.
7986 * rtx-vector-builder.h: New file.
7987 * rtx-vector-builder.c: Likewise.
7988 * rtl.h (rtx_def::u2): Add a const_vector field.
7989 (CONST_VECTOR_NPATTERNS): New macro.
7990 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
7991 (CONST_VECTOR_DUPLICATE_P): Likewise.
7992 (CONST_VECTOR_STEPPED_P): Likewise.
7993 (CONST_VECTOR_ENCODED_ELT): Likewise.
7994 (const_vec_duplicate_p): Check for a duplicated vector encoding.
7995 (unwrap_const_vec_duplicate): Likewise.
7996 (const_vec_series_p): Check for a non-duplicated vector encoding.
7997 Say that the function only returns true for integer vectors.
7998 * emit-rtl.c: Include rtx-vector-builder.h.
7999 (gen_const_vec_duplicate_1): Delete.
8000 (gen_const_vector): Call gen_const_vec_duplicate instead of
8001 gen_const_vec_duplicate_1.
8002 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
8003 (gen_const_vec_duplicate): Use rtx_vector_builder.
8004 (gen_const_vec_series): Likewise.
8005 (gen_rtx_CONST_VECTOR): Likewise.
8006 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
8007 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8008 Build a new vector rather than modifying a CONST_VECTOR in-place.
8009 (handle_special_swappables): Update call accordingly.
8010 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
8011 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8012 Build a new vector rather than modifying a CONST_VECTOR in-place.
8013 (handle_special_swappables): Update call accordingly.
8015 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8017 * simplify-rtx.c (simplify_const_binary_operation): Use
8018 CONST_VECTOR_ELT instead of XVECEXP.
8020 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8022 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
8023 the selector elements to be different from the data elements
8024 if the selector is a VECTOR_CST.
8025 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
8026 ssizetype for the selector.
8028 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8030 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
8031 before testing each element individually.
8032 * tree-vect-generic.c (lower_vec_perm): Likewise.
8034 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8036 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
8037 * selftest-run-tests.c (selftest::run_tests): Call it.
8038 * vector-builder.h (vector_builder::operator ==): New function.
8039 (vector_builder::operator !=): Likewise.
8040 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
8041 (vec_perm_indices::all_from_input_p): New function.
8042 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8043 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
8044 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
8045 instead of reading the VECTOR_CST directly. Detect whether both
8046 vector inputs are the same before constructing the vec_perm_indices,
8047 and update the number of inputs argument accordingly. Use the
8048 utility functions added above. Only construct sel2 if we need to.
8050 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8052 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
8053 the broadcast of the low byte.
8054 (expand_mult_highpart): Use an explicit encoding for the permutes.
8055 * optabs-query.c (can_mult_highpart_p): Likewise.
8056 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
8057 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8058 (vectorizable_bswap): Likewise.
8059 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
8060 explicit encoding for the power-of-2 permutes.
8061 (vect_permute_store_chain): Likewise.
8062 (vect_grouped_load_supported): Likewise.
8063 (vect_permute_load_chain): Likewise.
8065 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8067 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
8068 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
8069 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
8070 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8071 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
8072 (vect_gen_perm_mask_any): Likewise.
8074 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8076 * int-vector-builder.h: New file.
8077 * vec-perm-indices.h: Include int-vector-builder.h.
8078 (vec_perm_indices): Redefine as an int_vector_builder.
8079 (auto_vec_perm_indices): Delete.
8080 (vec_perm_builder): Redefine as a stand-alone class.
8081 (vec_perm_indices::vec_perm_indices): New function.
8082 (vec_perm_indices::clamp): Likewise.
8083 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
8084 (vec_perm_indices::new_vector): New function.
8085 (vec_perm_indices::new_expanded_vector): Update for new
8086 vec_perm_indices class.
8087 (vec_perm_indices::rotate_inputs): New function.
8088 (vec_perm_indices::all_in_range_p): Operate directly on the
8089 encoded form, without computing elided elements.
8090 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
8091 encoding. Update for new vec_perm_indices class.
8092 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
8093 the given vec_perm_builder.
8094 (expand_vec_perm_var): Update vec_perm_builder constructor.
8095 (expand_mult_highpart): Use vec_perm_builder instead of
8096 auto_vec_perm_indices.
8097 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
8098 vec_perm_indices instead of auto_vec_perm_indices. Use a single
8099 or double series encoding as appropriate.
8100 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
8101 vec_perm_indices instead of auto_vec_perm_indices.
8102 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8103 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8104 (vect_permute_store_chain): Likewise.
8105 (vect_grouped_load_supported): Likewise.
8106 (vect_permute_load_chain): Likewise.
8107 (vect_shift_permute_load_chain): Likewise.
8108 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8109 (vect_transform_slp_perm_load): Likewise.
8110 (vect_schedule_slp_instance): Likewise.
8111 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8112 (vectorizable_mask_load_store): Likewise.
8113 (vectorizable_bswap): Likewise.
8114 (vectorizable_store): Likewise.
8115 (vectorizable_load): Likewise.
8116 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
8117 vec_perm_indices instead of auto_vec_perm_indices. Use
8118 tree_to_vec_perm_builder to read the vector from a tree.
8119 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
8120 vec_perm_builder instead of a vec_perm_indices.
8121 (have_whole_vector_shift): Use vec_perm_builder and
8122 vec_perm_indices instead of auto_vec_perm_indices. Leave the
8123 truncation to calc_vec_perm_mask_for_shift.
8124 (vect_create_epilog_for_reduction): Likewise.
8125 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
8126 from auto_vec_perm_indices to vec_perm_indices.
8127 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8128 instead of changing individual elements.
8129 (aarch64_vectorize_vec_perm_const): Use new_vector to install
8130 the vector in d.perm.
8131 * config/arm/arm.c (expand_vec_perm_d::perm): Change
8132 from auto_vec_perm_indices to vec_perm_indices.
8133 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8134 instead of changing individual elements.
8135 (arm_vectorize_vec_perm_const): Use new_vector to install
8136 the vector in d.perm.
8137 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
8138 Update vec_perm_builder constructor.
8139 (rs6000_expand_interleave): Likewise.
8140 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
8141 (rs6000_expand_interleave): Likewise.
8143 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8145 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
8146 to qimode could truncate the indices.
8147 * optabs.c (expand_vec_perm_var): Likewise.
8149 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8151 * Makefile.in (OBJS): Add vec-perm-indices.o.
8152 * vec-perm-indices.h: New file.
8153 * vec-perm-indices.c: Likewise.
8154 * target.h (vec_perm_indices): Replace with a forward class
8156 (auto_vec_perm_indices): Move to vec-perm-indices.h.
8157 * optabs.h: Include vec-perm-indices.h.
8158 (expand_vec_perm): Delete.
8159 (selector_fits_mode_p, expand_vec_perm_var): Declare.
8160 (expand_vec_perm_const): Declare.
8161 * target.def (vec_perm_const_ok): Replace with...
8162 (vec_perm_const): ...this new hook.
8163 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
8164 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
8165 * doc/tm.texi: Regenerate.
8166 * optabs.def (vec_perm_const): Delete.
8167 * doc/md.texi (vec_perm_const): Likewise.
8168 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
8169 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
8170 expand_vec_perm for constant permutation vectors. Assert that
8171 the mode of variable permutation vectors is the integer equivalent
8172 of the mode that is being permuted.
8173 * optabs-query.h (selector_fits_mode_p): Declare.
8174 * optabs-query.c: Include vec-perm-indices.h.
8175 (selector_fits_mode_p): New function.
8176 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
8177 is defined, instead of checking whether the vec_perm_const_optab
8178 exists. Use targetm.vectorize.vec_perm_const instead of
8179 targetm.vectorize.vec_perm_const_ok. Check whether the indices
8180 fit in the vector mode before using a variable permute.
8181 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
8182 vec_perm_indices instead of an rtx.
8183 (expand_vec_perm): Replace with...
8184 (expand_vec_perm_const): ...this new function. Take the selector
8185 as a vec_perm_indices rather than an rtx. Also take the mode of
8186 the selector. Update call to shift_amt_for_vec_perm_mask.
8187 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
8188 Use vec_perm_indices::new_expanded_vector to expand the original
8189 selector into bytes. Check whether the indices fit in the vector
8190 mode before using a variable permute.
8191 (expand_vec_perm_var): Make global.
8192 (expand_mult_highpart): Use expand_vec_perm_const.
8193 * fold-const.c: Includes vec-perm-indices.h.
8194 * tree-ssa-forwprop.c: Likewise.
8195 * tree-vect-data-refs.c: Likewise.
8196 * tree-vect-generic.c: Likewise.
8197 * tree-vect-loop.c: Likewise.
8198 * tree-vect-slp.c: Likewise.
8199 * tree-vect-stmts.c: Likewise.
8200 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
8202 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
8203 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
8204 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
8205 (aarch64_vectorize_vec_perm_const): ...this new function.
8206 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8207 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8208 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
8209 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
8210 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8211 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8212 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
8214 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
8215 check for NEON modes.
8216 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
8217 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
8218 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
8219 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
8221 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
8222 the old VEC_PERM_CONST conditions.
8223 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
8224 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
8225 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
8226 (ia64_vectorize_vec_perm_const_ok): Merge into...
8227 (ia64_vectorize_vec_perm_const): ...this new function.
8228 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
8229 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
8230 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
8231 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
8232 * config/mips/mips.c (mips_expand_vec_perm_const)
8233 (mips_vectorize_vec_perm_const_ok): Merge into...
8234 (mips_vectorize_vec_perm_const): ...this new function.
8235 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
8236 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
8237 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
8238 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
8239 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
8240 (rs6000_expand_vec_perm_const): Delete.
8241 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
8243 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8244 (altivec_expand_vec_perm_const_le): Take each operand individually.
8245 Operate on constant selectors rather than rtxes.
8246 (altivec_expand_vec_perm_const): Likewise. Update call to
8247 altivec_expand_vec_perm_const_le.
8248 (rs6000_expand_vec_perm_const): Delete.
8249 (rs6000_vectorize_vec_perm_const_ok): Delete.
8250 (rs6000_vectorize_vec_perm_const): New function.
8251 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8252 an element count and rtx array.
8253 (rs6000_expand_extract_even): Update call accordingly.
8254 (rs6000_expand_interleave): Likewise.
8255 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
8256 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
8257 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
8258 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
8259 (rs6000_expand_vec_perm_const): Delete.
8260 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8261 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8262 (altivec_expand_vec_perm_const_le): Take each operand individually.
8263 Operate on constant selectors rather than rtxes.
8264 (altivec_expand_vec_perm_const): Likewise. Update call to
8265 altivec_expand_vec_perm_const_le.
8266 (rs6000_expand_vec_perm_const): Delete.
8267 (rs6000_vectorize_vec_perm_const_ok): Delete.
8268 (rs6000_vectorize_vec_perm_const): New function. Remove stray
8269 reference to the SPE evmerge intructions.
8270 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8271 an element count and rtx array.
8272 (rs6000_expand_extract_even): Update call accordingly.
8273 (rs6000_expand_interleave): Likewise.
8274 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
8275 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
8277 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8279 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8281 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
8282 vector mode and that that mode matches the mode of the data
8284 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
8285 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
8286 directly using expand_vec_perm_1 when forcing selectors into
8288 (expand_vec_perm_var): New function, split out from expand_vec_perm.
8290 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8292 * optabs-query.h (can_vec_perm_p): Delete.
8293 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
8294 * optabs-query.c (can_vec_perm_p): Split into...
8295 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
8296 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
8297 particular selector is valid.
8298 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8299 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8300 (vect_grouped_load_supported): Likewise.
8301 (vect_shift_permute_load_chain): Likewise.
8302 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8303 (vect_transform_slp_perm_load): Likewise.
8304 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8305 (vectorizable_bswap): Likewise.
8306 (vect_gen_perm_mask_checked): Likewise.
8307 * fold-const.c (fold_ternary_loc): Likewise. Don't take
8308 implementations of variable permutation vectors into account
8309 when deciding which selector to use.
8310 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
8311 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
8312 with a false third argument.
8313 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
8314 to test whether the constant selector is valid and can_vec_perm_var_p
8315 to test whether a variable selector is valid.
8317 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8319 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
8320 * optabs-query.c (can_vec_perm_p): Likewise.
8321 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
8322 instead of vec_perm_indices.
8323 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
8324 (vect_gen_perm_mask_checked): Likewise,
8325 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
8326 (vect_gen_perm_mask_checked): Likewise,
8328 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8330 * optabs-query.h (qimode_for_vec_perm): Declare.
8331 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
8332 (qimode_for_vec_perm): ...this new function.
8333 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
8335 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8337 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
8338 does not have a conditional at the top.
8340 2018-01-02 Richard Biener <rguenther@suse.de>
8342 * ipa-inline.c (big_speedup_p): Fix expression.
8344 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8347 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
8350 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8354 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
8355 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
8356 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
8357 cond_taken_branch_cost 3->4.
8359 2018-01-01 Jakub Jelinek <jakub@redhat.com>
8361 PR tree-optimization/83581
8362 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
8363 TODO_cleanup_cfg if any changes have been made.
8366 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
8367 convert_modes if target mode has the right side, but different mode
8371 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
8372 last argument when extracting from CONCAT. If either from_real or
8373 from_imag is NULL, use expansion through memory. If result is not
8374 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
8375 the parts directly to inner mode, if even that fails, use expansion
8379 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
8380 check for bswap in mode rather than HImode and use that in expand_unop
8383 Copyright (C) 2018 Free Software Foundation, Inc.
8385 Copying and distribution of this file, with or without modification,
8386 are permitted in any medium without royalty provided the copyright
8387 notice and this notice are preserved.