1 2018-03-08 Jakub Jelinek <jakub@redhat.com>
4 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
5 has ',' character inside of it.
7 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
10 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
11 as clobbering CC_REGNUM.
13 2018-03-08 Richard Biener <rguenther@suse.de>
16 * tree-scalar-evolution.c: Include tree-into-ssa.h.
17 (follow_copies_to_constant): Do not follow SSA names registered
20 2018-03-08 Richard Biener <rguenther@suse.de>
22 PR tree-optimization/84178
23 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
25 (version_loop_for_if_conversion): Delay update_ssa call.
26 (tree_if_conversion): Delay update_ssa until after predicate
29 2018-03-08 David Malcolm <dmalcolm@redhat.com>
31 PR tree-optimization/84178
32 * tree-if-conv.c (release_bb_predicate): Remove the
33 the assertion that the stmts have NULL use_ops.
34 Discard the statements, asserting that they haven't
35 yet been added to a BB.
37 2018-03-08 Richard Biener <rguenther@suse.de>
39 PR tree-optimization/84746
40 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
41 (phi_translate): Pass in destination ANTIC_OUT set.
42 (phi_translate_1): Likewise. For a simplified result lookup
43 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
44 (phi_translate_set): Adjust.
45 (do_pre_regular_insertion): Likewise.
46 (do_pre_partial_partial_insertion): Likewise.
48 2018-03-08 Martin Liska <mliska@suse.cz>
51 * doc/gcov.texi: Document usage of profile files.
52 * gcov-io.h: Document changes in the format.
54 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
58 * dwarf2out.c (struct dw_line_info_table): Update comments for
60 (FORCE_RESET_NEXT_VIEW): New.
61 (FORCE_RESETTING_VIEW_P): New.
62 (RESETTING_VIEW_P): Check for -1 too.
63 (ZERO_VIEW_P): Likewise.
64 (new_line_info_table): Force-reset next view.
65 (dwarf2out_begin_function): Likewise.
66 (dwarf2out_source_line): Simplify zero_view_p initialization.
67 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
68 view directly. Omit view when omitting .loc at line 0.
70 2018-03-08 Jakub Jelinek <jakub@redhat.com>
72 PR tree-optimization/84740
73 * tree-switch-conversion.c (process_switch): Call build_constructors
74 only if info.phi_count is non-zero.
76 PR tree-optimization/84739
77 * tree-tailcall.c (find_tail_calls): Check call arguments against
78 DECL_ARGUMENTS (current_function_decl) rather than
79 DECL_ARGUMENTS (func) when checking for tail recursion.
81 2018-03-07 Jakub Jelinek <jakub@redhat.com>
83 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
84 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
85 Volker Reichelt's entry and add entries for people that perform
86 GCC fuzzy testing and report numerous bugs.
88 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
91 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
92 readonly data in sdata, if that is disabled.
93 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
94 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
95 -mreadonly-in-sdata option.
97 2018-03-07 Martin Sebor <msebor@redhat.com>
99 PR tree-optimization/84468
100 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
101 basic block when looking for nul assignment.
103 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
106 * except.h (output_function_exception_table): Adjust prototype.
107 * except.c (output_function_exception_table): Remove FNNAME parameter
108 and add SECTION parameter. Ouput one part of the table at a time.
109 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
110 the first part of the exception table and emit unwind directives.
111 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
112 (i386_pe_seh_cold_init): Likewise.
113 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
114 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
115 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
116 (ix86_output_call_insn): Emit a nop in one more case for SEH.
117 * config/i386/winnt.c: Include except.h.
118 (struct seh_frame_state): Add reg_offset, after_prologue and
119 in_cold_section fields.
120 (i386_pe_seh_end_prologue): Set seh->after_prologue.
121 (i386_pe_seh_cold_init): New function.
122 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
123 to seh->in_cold_section.
124 (seh_emit_push): Record the offset of the push.
125 (seh_emit_save): Record the offet of the save.
126 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
127 Test seh->after_prologue to disregard the epilogue.
128 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
129 (i386_pe_end_cold_function): New function.
131 2018-03-07 Jakub Jelinek <jakub@redhat.com>
134 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
135 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
138 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
140 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
141 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
144 * multiple_target.c: Include tree-inline.h and intl.h.
145 (expand_target_clones): Diagnose and fail if node->definition and
146 !tree_versionable_function_p (node->decl).
148 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
150 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
152 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
153 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
154 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
156 2018-03-06 Jakub Jelinek <jakub@redhat.com>
159 * combine.c (try_combine): Use reg_or_subregno instead of handling
160 just paradoxical SUBREGs and REGs.
162 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
164 * config/arc/arc.c (arc_finalize_pic): Remove function.
165 (arc_must_save_register): We use single base PIC register, remove
166 checks to save/restore the PIC register.
167 (arc_expand_prologue): Likewise.
168 * config/arc/arc-protos.h (arc_set_default_type_attributes):
170 (arc_verify_short): Likewise.
171 (arc_attr_type): Likewise.
172 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
173 (walk_stores): Likewise.
174 (arc_address_cost): Make it static.
175 (arc_verify_short): Likewise.
176 (branch_dest): Likewise.
177 (arc_attr_type): Likewise.
178 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
179 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
180 (arc_final_prescan_insn): Remove inserting the nops due to
181 hardware hazards. It is done in reorg step.
182 (insn_length_variant_t): Remove.
183 (insn_length_parameters_t): Likewise.
184 (arc_insn_length_parameters): Likewise.
185 (arc_get_insn_variants): Likewise.
186 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
188 2018-03-06 Jakub Jelinek <jakub@redhat.com>
191 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
194 PR tree-optimization/84687
195 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
197 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
199 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
201 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
202 Rename to ppc_speculation_barrier.
203 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
204 __builtin_ppc_speculation_barrier.
206 2018-03-05 Jakub Jelinek <jakub@redhat.com>
209 * combine.c (combine_simplify_rtx): Don't try to simplify if
210 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
213 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
215 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
216 to 32 bytes when compiling for POWER9.
218 2018-03-05 Jakub Jelinek <jakub@redhat.com>
221 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
222 regparm >= 3 with no arg reg available also for calls with
223 flag_force_indirect_call. Pass decl to ix86_function_regparm.
226 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
228 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
230 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
233 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
235 2018-03-05 Richard Biener <rguenther@suse.de>
237 PR tree-optimization/84486
238 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
239 When inserting a __builtin_assume_aligned call set the LHS
240 SSA name alignment info accordingly.
242 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
244 PR tree-optimization/84114
245 * config/aarch64/aarch64.c (aarch64_reassociation_width)
246 Avoid reassociation of FLOAT_MODE addition.
248 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
250 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
251 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
252 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
253 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
254 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
255 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
256 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
258 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
259 __builtin_ia32_wbinvd): New builtins.
260 (SPECIAL_ARGS2): New.
261 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
262 (SPECIAL_ARGS2): New.
263 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
264 (ix86_valid_target_attribute_inner_p): Ditto.
265 (ix86_init_mmx_sse_builtins): Add special_args2.
266 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
267 TARGET_WBNOINVD_P): New.
268 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
269 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
270 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
271 * config/i386/immintrin.h (_wbinvd): New intrinsic.
272 * config/i386/pconfigintrin.h: New file.
273 * config/i386/wbnoinvdintrin.h: Ditto.
274 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
275 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
277 2018-03-05 Richard Biener <rguenther@suse.de>
279 PR tree-optimization/84670
280 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
282 (BB_VISITED_WITH_VISITED_SUCCS): New define.
283 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
284 (compute_antic_aux): Only assert the number of values in ANTIC_IN
285 doesn't grow if all successors (recursively) were visited at least
288 2018-03-05 Richard Biener <rguenther@suse.de>
290 PR tree-optimization/84650
291 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
292 if executed in the loop pipeline.
294 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
296 * doc/configfiles.texi (Configuration Files): Move info about
297 conditionalizing $target-protos.h to...
298 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
299 differs from $target-protos.h.
301 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
302 Chung-Ju Wu <jasonwucj@gmail.com>
304 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
305 * config/nds32/nds32-multiple.md (setmemsi): Define.
306 * config/nds32/nds32-memory-manipulation.c
307 (nds32_gen_dup_4_byte_to_word_value): New.
308 (emit_setmem_word_loop): New.
309 (emit_setmem_byte_loop): New.
310 (nds32_expand_setmem_loop): New.
311 (nds32_expand_setmem_loop_v3m): New.
312 (nds32_expand_setmem_unroll): New.
313 (nds32_expand_setmem): New.
315 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
316 Chung-Ju Wu <jasonwucj@gmail.com>
318 * config/nds32/nds32-memory-manipulation.c
319 (nds32_emit_load_store): New.
320 (nds32_emit_post_inc_load_store): New.
321 (nds32_emit_mem_move): New.
322 (nds32_emit_mem_move_block): New.
323 (nds32_expand_movmemsi_loop_unknown_size): New.
324 (nds32_expand_movmemsi_loop_known_size): New.
325 (nds32_expand_movmemsi_loop): New.
326 (nds32_expand_movmemsi_unroll): New.
327 (nds32_expand_movmemqi): Rename ...
328 (nds32_expand_movmemsi): ... to this.
329 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
330 (movmemsi): ... to this.
331 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
332 (nds32_expand_movmemsi): ... to this.
334 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
335 Monk Chiang <sh.chiang04@gmail.com>
336 Chung-Ju Wu <jasonwucj@gmail.com>
338 * config/nds32/nds32-protos.h
339 (nds32_expand_load_multiple): New arguments.
340 (nds32_expand_store_multiple): Ditto.
341 (nds32_valid_multiple_load_store): Rename ...
342 (nds32_valid_multiple_load_store_p): ... to this.
343 * config/nds32/nds32-memory-manipulation.c
344 (nds32_expand_load_multiple): Refine implementation.
345 (nds32_expand_store_multiple): Ditto.
346 * config/nds32/nds32-multiple.md
347 (load_multiple): Update nds32_expand_load_multiple interface.
348 (store_multiple): Update nds32_expand_store_multiple interface.
349 * config/nds32/nds32-predicates.c
350 (nds32_valid_multiple_load_store): Rename ...
351 (nds32_valid_multiple_load_store_p): ... to this and refine
353 * config/nds32/predicates.md
354 (nds32_load_multiple_and_update_address_operation): New predicate.
355 (nds32_store_multiple_and_update_address_operation): New predicate.
357 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
358 Chung-Ju Wu <jasonwucj@gmail.com>
360 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
361 (combo): New attribute.
362 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
364 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
366 * config/nds32/nds32.opt: Change -mcmodel= default value.
368 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
369 Monk Chiang <sh.chiang04@gmail.com>
370 Chung-Ju Wu <jasonwucj@gmail.com>
372 * config/nds32/constants.md (unspec_element): New enum.
373 * config/nds32/constraints.md (Umw): New constraint.
374 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
375 * config/nds32/nds32-intrinsic.md: Likewise.
376 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
377 (nds32_valid_smw_lwm_base_p): New.
378 (nds32_output_smw_single_word): New.
379 (nds32_output_lmw_single_word): New.
380 (nds32_expand_unaligned_load): New.
381 (nds32_expand_unaligned_store): New.
382 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
383 (nds32_output_smw_single_word): Declare.
384 (nds32_output_lmw_single_word): Declare.
385 (nds32_expand_unaligned_load): Declare.
386 (nds32_expand_unaligned_store): Declare.
387 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
388 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
389 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
390 NDS32_BUILTIN_UASTORE_DW.
391 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
394 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
395 Kito Cheng <kito.cheng@gmail.com>
396 Chung-Ju Wu <jasonwucj@gmail.com>
398 * config/nds32/nds32-intrinsic.c
399 (nds32_expand_builtin_null_ftype_reg): Delete.
400 (nds32_expand_builtin_reg_ftype_imm): Ditto.
401 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
402 (nds32_read_argument): New.
403 (nds32_legitimize_target): Ditto.
404 (nds32_legitimize_argument): Ditto.
405 (nds32_check_constant_argument): Ditto.
406 (nds32_expand_unop_builtin): Ditto.
407 (nds32_expand_unopimm_builtin): Ditto.
408 (nds32_expand_binop_builtin): Ditto.
409 (nds32_builtin_decl_impl): Ditto.
410 (builtin_description): Ditto.
411 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
412 (nds32_init_builtins_impl): Ditto.
413 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
414 (nds32_builtin_decl): New.
415 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
416 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
418 2018-03-02 Jeff Law <law@redhat.com>
420 * reorg.c (stop_search_p): Handle DEBUG_INSN.
421 (redundant_insn, fill_simple_delay_slots): Likewise.
422 (fill_slots_from_thread): Likewise.
423 * resource.c (mark_referenced_resources): Likewise.
424 (mark_set_resources, find_dead_or_set_registers): Likewise.
426 2018-03-02 Jakub Jelinek <jakub@redhat.com>
428 * substring-locations.h (format_warning_va): Formatting fix for
430 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
432 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
433 * substring-locations.c: Include intl.h.
434 (format_warning_va): Turned into small wrapper around
435 format_warning_n_va, renamed to ...
436 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
437 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
439 (format_warning_at_substring_n): New function.
440 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
441 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
442 format_warning_at_substring with just a shorter name instead of
443 const function pointer.
444 (fmtwarn_n): New function.
445 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
446 appropriate, get rid of all the fmtstr temporaries, move conditionals
447 with G_() wrapped string literals directly into fmtwarn arguments,
448 cast dir.len to (int), formatting fixes.
450 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
452 * doc/invoke.texi: Remove "Cilk Plus" references.
454 2018-03-02 Jakub Jelinek <jakub@redhat.com>
455 Richard Biener <rguenther@suse.de>
458 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
459 for error or warning attributes if CALL_FROM_THUNK_P is set.
462 2018-03-02 Jakub Jelinek <jakub@redhat.com>
465 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
466 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
469 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
470 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
472 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
473 instead of -1U in last predictors element's probability member.
475 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
478 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
479 arguments if they are comparable.
481 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
483 PR tree-optimization/84634
484 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
485 masks and masked_loop_p with a single loop_masks, making sure it's
486 null for bb vectorization.
488 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
490 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
491 (vect_analyze_data_ref_access): Use loop->safe_len rather than
492 loop->force_vectorize to check whether there is no alias.
494 2018-03-02 Jakub Jelinek <jakub@redhat.com>
497 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
499 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
501 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
502 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
503 instead of a loop around prev_real_insn.
504 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
508 * config/i386/i386.c (ix86_print_operand): Use conditional
509 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
512 2018-03-02 Richard Biener <rguenther@suse.de>
514 PR tree-optimization/84427
515 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
516 (bitmap_set_subtract_values): Rewrite to handle multiple
519 (prune_clobbered_mems): Likewise.
520 (phi_translate): Take edge instead of pred/phiblock.
521 (phi_translate_1): Likewise.
522 (phi_translate_set): Likewise. Insert all translated
523 exprs for a value into the set, keeping possibly multiple
524 expressions per value.
525 (compute_antic_aux): Adjust for phi_translate changes.
526 When intersecting union the expressions and prune those
527 not in the final value set, keeping possibly multiple
528 expressions per value. Do not use value-insertion
529 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
530 all expressions. Add verification that the value-sets
531 only shrink during iteration.
532 (compute_partial_antic_aux): Adjust for the phi_translate changes.
533 (do_pre_regular_insertion): Likewise.
534 (do_pre_partial_partial_insertion): Likewise.
536 2018-03-02 Richard Biener <rguenther@suse.de>
539 * config/darwin.c (saved_debug_info_level): New static global.
540 (darwin_asm_lto_start): Disable debug info generation for LTO out.
541 (darwin_asm_lto_end): Restore debug info generation settings.
543 2018-03-01 Martin Liska <mliska@suse.cz>
546 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
549 2018-03-01 Richard Biener <rguenther@suse.de>
552 * dwarf2out.c (gen_variable_die): Properly handle late VLA
553 type annotation with LTO when debug was disabled at compile-time.
555 2018-03-01 Matthew Fortune <mfortune@gmail.com>
557 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
559 (mips_final_postscan_insn): Likewise.
561 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
563 PR rtl-optimization/84528
564 * alias.c (init_alias_target): Add commentary.
565 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
566 a unique base value if the frame pointer is not eliminated
567 to the stack pointer.
569 2018-03-01 Tom de Vries <tom@codesourcery.com>
571 PR rtl-optimization/83327
572 * lra-int.h (hard_regs_spilled_into): Declare.
573 * lra.c (hard_regs_spilled_into): Define.
574 (init_reg_info): Init hard_regs_spilled_into.
575 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
576 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
577 (process_bb_lives): Handle hard_regs_spilled_into.
578 (lra_create_live_ranges_1): Before doing liveness propagation, clear
579 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
581 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
583 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
584 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
585 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
586 * config/rs6000/aix72.h: New file.
588 2018-02-28 Jakub Jelinek <jakub@redhat.com>
590 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
591 instead of warning_at with conditional singular and plural messages
595 * stor-layout.c (update_alignment_for_field): For
596 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
597 && !DECL_PACKED (field), do the alignment update, just use
598 only desired_align instead of MAX (type_align, desired_align)
600 (place_field): Don't do known_align < desired_align handling
601 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
602 is non-NULL, instead do it after rli->prev_field handling and
603 only if not within a bitfield word. For DECL_PACKED (field)
604 use type_align of BITS_PER_UNIT.
606 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
608 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
609 superfluous parentheses and trailing spaces.
611 2018-02-28 Richard Biener <rguenther@suse.de>
613 PR tree-optimization/84584
614 * graphite-scop-detection.c (scop_detection::add_scop): Discard
615 SCoPs with fake exit edge.
617 2018-02-28 Martin Liska <mliska@suse.cz>
620 * timevar.c (timer::print): Fix format to properly print 100%
623 2018-02-28 Richard Biener <rguenther@suse.de>
626 * genmatch.c (capture_info::walk_match): Do not mark
627 captured expressions without operands as expr_p given
628 they act more like predicates and should be subject to
629 "lost tail" side-effect preserving.
631 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
633 PR rtl-optimization/81611
634 * auto-inc-dec.c (attempt_change): Move dead note from
635 mem_insn if it's the next use of regno
636 (find_address): Take address use of reg holding
637 non-incremented value. Add parm to limit search to the named
639 (merge_in_block): Attempt to use a mem insn that is the next
640 use of the original regno.
642 2018-02-27 Martin Sebor <msebor@redhat.com>
645 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
646 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
648 2018-02-27 Martin Sebor <msebor@redhat.com>
651 * diagnostic-core.h (warning_n, error_n, inform_n): Change
652 n argument to unsigned HOST_WIDE_INT.
653 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
654 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
655 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
656 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
658 2018-02-27 Richard Biener <rguenther@suse.de>
660 PR tree-optimization/84512
661 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
662 Do not use the estimate returned from record_stmt_cost for
663 the scalar iteration cost but sum properly using add_stmt_cost.
665 2018-02-27 Richard Biener <rguenther@suse.de>
667 PR tree-optimization/84466
668 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
669 Adjust last change to less strictly validate use operands.
671 2018-02-27 Martin Liska <mliska@suse.cz>
673 PR gcov-profile/84548
674 * gcov.c (process_file): Allow partial overlap and consider it
675 also as group functions.
676 (output_lines): Properly calculate range of lines for a group.
678 2018-02-27 Martin Liska <mliska@suse.cz>
680 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
681 'ggc' suffixes. Change first column width.
682 (timer::print): Fix formatting of the column.
684 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
686 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
687 preserve inline entry blocks for the sake of debug inline
688 entry point markers alone.
689 (remove_unused_locals): Suggest in comments a better place to
690 force the preservation of inline entry blocks that are
691 otherwise unused, but do not preserve them.
693 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
695 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
697 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
700 * config/i386/constraints.md (Bs): Replace
701 ix86_indirect_branch_register with
702 TARGET_INDIRECT_BRANCH_REGISTER.
704 * config/i386/i386.md (indirect_jump): Likewise.
705 (tablejump): Likewise.
706 (*sibcall_memory): Likewise.
707 (*sibcall_value_memory): Likewise.
708 Peepholes of indirect call and jump via memory: Likewise.
709 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
710 (*sibcall_value_GOT_32): Likewise.
711 * config/i386/predicates.md (indirect_branch_operand): Likewise.
712 (GOT_memory_operand): Likewise.
713 (call_insn_operand): Likewise.
714 (sibcall_insn_operand): Likewise.
715 (GOT32_symbol_operand): Likewise.
716 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
718 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
720 PR rtl-optimization/83496
721 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
722 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
723 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
724 redundant insn, if any.
725 (relax_delay_slots): Likewise.
726 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
728 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
730 PR tree-optimization/83965
731 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
732 that grouped statements are part of a reduction chain. Return
733 true if the statement is not marked as a reduction itself but
735 (vect_recog_dot_prod_pattern): Don't check whether the statement
736 is part of a group here.
737 (vect_recog_sad_pattern): Likewise.
738 (vect_recog_widen_sum_pattern): Likewise.
740 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
743 * final.c (rest_of_clean_state): Also look for calls inside sequences.
745 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
748 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
750 (ix86_output_indirect_function_return): New prototype.
751 (ix86_split_simple_return_pop_internal): Likewise.
752 * config/i386/i386.c (indirect_return_via_cx): New.
753 (indirect_return_via_cx_bnd): Likewise.
754 (indirect_thunk_name): Handle return va CX_REG.
755 (output_indirect_thunk_function): Create alias for
756 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
757 (ix86_output_indirect_jmp): Remove the bool argument.
758 (ix86_output_indirect_function_return): New function.
759 (ix86_split_simple_return_pop_internal): Likewise.
760 * config/i386/i386.md (*indirect_jump): Don't pass false
761 to ix86_output_indirect_jmp.
762 (*tablejump_1): Likewise.
763 (simple_return_pop_internal): Change it to define_insn_and_split.
764 Call ix86_split_simple_return_pop_internal to split it for
766 (simple_return_indirect_internal): Call
767 ix86_output_indirect_function_return instead of
768 ix86_output_indirect_jmp.
770 2018-02-26 Jakub Jelinek <jakub@redhat.com>
773 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
774 memset and value initialization afterwards.
776 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
778 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
780 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
783 * common/config/aarch64/aarch64-common.c
784 (aarch_option_optimization_table[]): Switch
785 off fomit-frame-pointer
787 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
788 Chung-Ju Wu <jasonwucj@gmail.com>
790 * config/nds32/nds32-multiple.md (load_multiple): Disallow
792 (store_multiple): Ditto.
794 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
796 * config.gcc: Add --with-cpu support for nds32 target.
797 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
798 * config/nds32/nds32.opt: Add -mcpu= option.
800 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
802 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
803 isel=yes): Warn for these deprecated options.
805 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
807 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
808 ISA_2_5_MASKS_EMBEDDED.
810 2018-02-23 Jakub Jelinek <jakub@redhat.com>
812 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
813 p->max as pointers rather than using iterative_hash_expr.
815 2018-02-23 Carl Love <cel@us.ibm.com>
817 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
818 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
820 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
821 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
822 P8V_BUILTIN_VEC_VUNSIGNED2.
824 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
827 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
828 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
830 * lra-constraints.c (curr_insn_transform): Set up
831 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
833 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
834 * lra-eliminations.c (spill_pseudos): Ditto.
835 (process_insn_for_elimination): Ditto.
836 * lra-lives.c (reg_early_clobber_p): Use the new macros.
837 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
838 LRA_NON_CLOBBERED_ALT.
840 2018-02-22 Martin Sebor <msebor@redhat.com>
842 PR tree-optimization/84480
843 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
844 to maybe_diag_stxncpy_trunc. Call it.
845 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
846 from gimple_fold_builtin_strcpy. Print inlining stack.
847 (handle_builtin_stxncpy): Print inlining stack.
848 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
850 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
853 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
854 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
855 and -fcheck-pointer-bounds are used together.
856 (indirect_thunk_prefix): New enum.
857 (indirect_thunk_need_prefix): New function.
858 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
859 "_nt" instead of "_bnd" for NOTRACK prefix.
860 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
861 (output_indirect_thunk_function): Likewise.
863 (ix86_code_end): Update output_indirect_thunk_function calls.
864 (ix86_output_indirect_branch_via_reg): Replace
865 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
866 (ix86_output_indirect_branch_via_push): Likewise.
867 (ix86_output_function_return): Likewise.
868 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
869 incompatible with -fcf-protection=branch and
870 -fcheck-pointer-bounds.
872 2018-02-22 Steve Ellcey <sellcey@cavium.com>
875 * config/aarch64/aarch64.c (aarch64_print_address_internal):
876 Change gcc_assert call to output_operand_lossage.
878 2018-02-22 Steve Ellcey <sellcey@cavium.com>
880 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
882 2018-02-22 DJ Delorie <dj@redhat.com>
883 Sebastian Perta <sebastian.perta@renesas.com>
884 Oleg Endo <olegendo@gcc.gnu.org>
886 * config/rx/rx.c (rx_rtx_costs): New function.
887 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
889 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
891 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
893 2018-02-22 Martin Liska <mliska@suse.cz>
896 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
897 Add "native" as a possible value.
899 2018-02-22 Martin Liska <mliska@suse.cz>
902 * config/i386/i386.c (ix86_option_override_internal):
903 Add "native" as a possible value for -march and -mtune.
905 2018-02-22 Jakub Jelinek <jakub@redhat.com>
908 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
909 to all type variants.
911 PR tree-optimization/84503
912 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
913 width as info->bitpos + info->bitsize - start.
914 (merged_store_group::merge_overlapping): Simplify width computation.
915 (check_no_overlap): New function.
916 (imm_store_chain_info::try_coalesce_bswap): Compute expected
917 start + width and last_order of the group, fail if check_no_overlap
919 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
920 to group if check_no_overlap fails.
922 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
924 * config/rs6000/altivec.md: Delete contraint arguments to
925 define_expand, define_split, and define_peephole2, and in
926 define_insn_and_split if always unused.
927 * config/rs6000/darwin.md: Ditto.
928 * config/rs6000/dfp.md: Ditto.
929 * config/rs6000/rs6000.md: Ditto.
930 * config/rs6000/sync.md: Ditto.
931 * config/rs6000/vector.md: Ditto.
932 * config/rs6000/vsx.md: Ditto.
934 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
936 * config/rs6000/altivec.md: Write output control strings as braced
937 blocks instead of double-quoted strings.
938 * config/rs6000/darwin.md: Ditto.
939 * config/rs6000/rs6000.md: Ditto.
940 * config/rs6000/vector.md: Ditto.
941 * config/rs6000/vsx.md: Ditto.
943 2018-02-21 Jason Merrill <jason@redhat.com>
945 PR c++/84314 - ICE with templates and fastcall attribute.
946 * attribs.c (build_type_attribute_qual_variant): Remove assert.
948 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
950 * ipa-cp.c (determine_versionability): Fix comment typos.
952 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
955 * ipa-cp.c (determine_versionability): Do not version functions caling
958 2018-02-21 Martin Liska <mliska@suse.cz>
961 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
962 Add "native" as a possible value.
963 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
964 the macro when native cpu detection is available.
966 2018-02-21 Martin Liska <mliska@suse.cz>
969 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
970 Add "native" as a possible value.
971 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
972 when native cpu detection is available.
974 2018-02-21 Jakub Jelinek <jakub@redhat.com>
975 Martin Sebor <msebor@redhat.com>
977 PR tree-optimization/84478
978 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
980 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
981 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
982 support which is conservatively correct, for 2 only stay conservative
983 for maxlen. Formatting and comment capitalization fixes. Add STRICT
984 argument to the 2 argument get_range_strlen, adjust 6 arg
985 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
987 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
988 (gimple_fold_builtin_strlen): Pass true as last argument to
991 2018-02-20 Martin Sebor <msebor@redhat.com>
994 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
995 (builtin_memref::set_base_and_offset): Same. Handle inner references.
996 (builtin_memref::builtin_memref): Factor out parts into
997 set_base_and_offset and call it.
999 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1002 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
1003 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
1004 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
1005 search at the associated MODE_INT.
1007 2018-02-20 Jeff Law <law@redhat.com>
1010 PR tree-optimization/81592
1012 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
1013 for range data rather than using global data.
1014 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
1015 range data rather than using global data.
1016 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
1017 pass it to children as needed.
1018 (struct directive::fmtresult): Similarly.
1019 (struct directive::set_width): Similarly.
1020 (struct directive::set_precision): Similarly.
1021 (format_integer, format_directive, parse_directive): Similarly.
1022 (format_none): Accept unnamed vr_values parameter.
1023 (format_percent, format_floating, format_character): Similarly.
1024 (format_string, format_plain): Similarly.
1025 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
1026 the EVRP range analyzer for range data rather than using global data.
1027 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
1028 gimple-ssa-evrp-analyze.h
1029 (class sprintf_dom_walker): Add after_dom_children member function.
1030 Add evrp_range_analyzer member.
1031 (sprintf_dom_walker::before_dom_children): Call into the EVRP
1032 range analyzer as needed.
1033 (sprintf_dom_walker::after_dom_children): New member function.
1034 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
1036 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
1037 (evrp_range_analyzer::pop_to_marker): Likewise.
1039 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1041 PR tree-optimization/84419
1042 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
1043 with the required type if its current type is compatible but
1046 2018-02-20 Jakub Jelinek <jakub@redhat.com>
1049 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
1050 after vectorization.
1052 2018-02-20 Martin Liska <mliska@suse.cz>
1055 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
1056 possible values if we don't have a hint.
1058 2018-02-20 Martin Liska <mliska@suse.cz>
1062 * final.c (shorten_branches): Build align_tab array with one
1064 * opts.c (finish_options): Add alignment option limit check.
1065 (MAX_CODE_ALIGN): Likewise.
1066 (MAX_CODE_ALIGN_VALUE): Likewise.
1067 * doc/invoke.texi: Document maximum allowed option value for
1068 all -falign-* options.
1070 2018-02-19 Jakub Jelinek <jakub@redhat.com>
1073 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
1074 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
1075 * var-tracking.c (emit_note_insn_var_location): Remove all references
1076 to NOTE_INSN_CALL_ARG_LOCATION.
1077 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
1078 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
1079 Use copy_rtx_if_shared.
1080 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
1081 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
1082 (dwarf2out_var_location): Remove handling of
1083 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
1085 * final.c (final_scan_insn): Remove all references to
1086 NOTE_INSN_CALL_ARG_LOCATION.
1087 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
1088 before dumping final insns.
1089 * except.c (emit_note_eh_region_end): Remove all references to
1090 NOTE_INSN_CALL_ARG_LOCATION.
1091 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
1092 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
1093 * config/arc/arc.c (hwloop_optimize): Likewise.
1094 * config/arm/arm.c (create_fix_barrier): Likewise.
1095 * config/s390/s390.c (s390_chunkify_start): Likewise.
1096 * config/sh/sh.c (find_barrier): Likewise.
1097 * config/i386/i386.c (rest_of_insert_endbranch,
1098 ix86_seh_fixup_eh_fallthru): Likewise.
1099 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
1100 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
1101 * config/frv/frv.c (frv_function_prologue): Likewise.
1102 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
1104 (note_outside_basic_block_p): Remove all references to
1105 NOTE_INSN_CALL_ARG_LOCATION.
1106 * gengtype.c (adjust_field_rtx_def): Likewise.
1107 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
1109 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
1110 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
1113 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
1116 PR tree-optimization/84452
1117 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
1118 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
1121 2018-02-19 Martin Liska <mliska@suse.cz>
1124 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
1126 2018-02-19 Martin Liska <mliska@suse.cz>
1127 Richard Sandiford <richard.sandiford@linaro.org>
1129 PR tree-optimization/82491
1130 * gimple-fold.c (get_base_constructor): Make earlier bail out
1133 2018-02-19 Carl Love <cel@us.ibm.com>
1135 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
1136 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
1138 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
1139 P8V_BUILTIN_VEC_NEG.
1141 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
1143 * config/rl78/rl78.md (movdf): New define expand.
1145 2018-02-19 Martin Liska <mliska@suse.cz>
1148 * doc/invoke.texi: Fix typo.
1149 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
1151 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
1153 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
1154 handle rs6000_single_float and rs6000_double_float specially for
1157 2018-02-16 Jeff Law <law@redhat.com>
1159 * config/rx/rx.c (add_pop_cfi_notes): New function.;
1162 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1165 * ipa-inline.c (inline_small_functions): Fix a typo.
1167 2018-02-16 Nathan Sidwell <nathan@acm.org>
1169 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
1171 2018-02-16 Carl Love <cel@us.ibm.com>
1173 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
1174 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
1175 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
1176 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
1177 expansion to P8V_BUILTIN_VEC_FLOAT2.
1179 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
1181 PR rtl-optimization/70023
1182 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
1183 src_regno into account.
1185 2018-02-16 Carl Love <cel@us.ibm.com>
1187 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
1188 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
1189 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
1190 * config/rs6000/rs6000.c: Remove case statements for
1191 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
1192 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
1193 and P9V_BUILTIN_VEC_VINSERT4B.
1194 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
1195 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
1196 * config/rs6000/vsx.md:
1197 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
1200 2018-02-16 Carl Love <cel@us.ibm.com>
1202 * config/rs6000/altivec.h: Add builtin names vec_extract4b
1204 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
1206 * config/rs6000/rs6000-c.c: Add the definitions for
1207 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
1208 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
1209 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
1210 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
1211 definition for insert4b and define insn *insert3b_internal.
1212 * doc/extend.texi: Add documentation for vec_extract4b.
1214 2018-02-16 Nathan Sidwell <nathan@acm.org>
1216 * doc/extend.texi (Backwards Compatibility): Mention friend
1217 injection. Note for-scope is deprecated.
1218 * doc/invoke.texi (-ffriend-injection): Deprecate.
1220 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
1222 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
1223 that moved to I2, also allow destinations that are a paradoxical
1224 subreg (instead of a normal reg).
1226 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
1229 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
1232 2018-02-16 Richard Biener <rguenther@suse.de>
1234 PR tree-optimization/84037
1235 PR tree-optimization/84016
1237 * config/i386/i386.c (ix86_builtin_vectorization_cost):
1238 Adjust vec_construct for the fact we need additional higher latency
1239 128bit inserts for AVX256 and AVX512 vector builds.
1240 (ix86_add_stmt_cost): Scale vector construction cost for
1243 2018-02-16 Richard Biener <rguenther@suse.de>
1245 PR tree-optimization/84417
1246 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
1247 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
1248 (non_rewritable_lvalue_p): Likewise, use poly-ints.
1250 2018-02-16 Martin Liska <mliska@suse.cz>
1253 * internal-fn.def (ASAN_CHECK): Set proper flags.
1254 (ASAN_MARK): Likewise.
1256 2018-02-16 Julia Koval <julia.koval@intel.com>
1258 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
1259 from PTA_CANNONLAKE.
1261 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1264 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
1265 Use ++iter rather than iter++ for std::list iterators.
1266 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
1267 defer deleting them until all nodes in the forest are processed. Do
1268 free even leaf nodes. Change to_process into auto_vec.
1271 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
1272 * vec.h (vec_default_construct): Use memset instead of placement new
1273 if BROKEN_VALUE_INITIALIZATION is defined.
1274 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
1275 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
1278 PR rtl-optimization/83723
1279 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
1280 * lra.c (lra_substitute_pseudo): Likewise. If true, use
1281 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
1283 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
1285 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
1287 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
1289 PR rtl-optimization/81443
1290 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
1291 from inner REGs to paradoxical SUBREGs.
1293 2018-02-16 Richard Biener <rguenther@suse.de>
1295 PR tree-optimization/84399
1296 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1297 For operands we can analyze at their definition make sure we can
1298 analyze them at each use as well.
1300 2018-02-16 Richard Biener <rguenther@suse.de>
1302 PR tree-optimization/84190
1303 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
1304 volatile accesses if the decl isn't volatile.
1306 2018-02-15 Jason Merrill <jason@redhat.com>
1308 PR c++/84314 - ICE with templates and fastcall attribute.
1309 * attribs.c (build_type_attribute_qual_variant): Don't clobber
1310 TYPE_CANONICAL on an existing type.
1312 2018-02-15 Jakub Jelinek <jakub@redhat.com>
1314 PR tree-optimization/84383
1315 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
1316 dstoff nor call operand_equal_p if dstbase is NULL.
1318 PR tree-optimization/84334
1319 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
1320 also a CONSTANT_CLASS_P, punt.
1322 2018-02-14 Jim Wilson <jimw@sifive.com>
1324 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
1325 first SMALL_OPERAND check. New local min_second_step. Move assert
1326 to where locals are set. Add TARGET_RVC support.
1327 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
1329 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
1331 * doc/invoke.texi: Correct -Wformat-overflow code sample.
1333 2018-02-14 Martin Sebor <msebor@redhat.com>
1335 PR tree-optimization/83698
1336 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
1337 arrays constrain the offset range to their bounds.
1338 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
1339 (builtin_access::overlap): Avoid setting the size of overlap if it's
1341 (maybe_diag_overlap): Also consider arrays when deciding what values
1342 of offsets to include in diagnostics.
1344 2018-02-14 Martin Sebor <msebor@redhat.com>
1347 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
1348 that correspond to the kind of a declaration.
1350 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
1353 * config/pa/pa.md: Load address of PIC label using the linkage table
1354 if the label is nonlocal.
1356 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
1358 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
1359 warning message if user requests -maltivec=be.
1360 * doc/invoke.texi: Document deprecation of -maltivec=be.
1362 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
1365 * config/rs6000/rs6000-c.c: Update definitions for
1366 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
1367 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
1369 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1372 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
1373 add _get_ssp intrinsics. Remove argument from
1374 __builtin_ia32_rdssp[d|q].
1375 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
1376 * config/i386/i386-builtin.def: Remove argument from
1377 __builtin_ia32_rdssp[d|q].
1378 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
1379 ix86_expand_special_args_builtin for _rdssp[d|q].
1380 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
1381 Clear register before usage.
1382 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
1383 Add documentation for new _get_ssp and _inc_ssp intrinsics.
1385 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
1387 PR tree-optimization/84357
1388 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
1389 operand 1 of an ARRAY_REF too.
1391 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
1394 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
1395 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1397 (set_of_reg): New struct.
1398 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
1399 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
1400 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1402 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
1403 Split into bitclr, bitset, bitinvert patterns if appropriate.
1404 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
1405 use rx_fuse_in_memory_bitop.
1406 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
1407 to named insn, correct maximum insn length.
1409 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
1412 * machmode.def: Define a complex mode for PARTIAL_INT.
1413 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
1415 * doc/rtl.texi: Document CSPImode.
1416 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
1418 (msp430_hard_regno_nregs_with_padding): Likewise.
1420 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
1423 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
1425 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
1427 PR rtl-optimization/84169
1428 * combine.c (try_combine): New variable split_i2i3. Set it to true if
1429 we generated a parallel as new i3 and we split that to new i2 and i3
1430 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
1431 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
1432 those to i2, not i1. Partially rewrite this scan code.
1434 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1437 * stor-layout.c (place_field): For variable length fields, adjust
1438 offset_align afterwards not just based on the field's alignment,
1439 but also on the size.
1442 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
1443 of exps and logs in the use_exp2 case.
1445 2018-02-13 Jeff Law <law@redhat.com>
1447 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
1450 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
1453 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
1457 * common.opt (gas-loc-support, gas-locview-support): New.
1458 (ginline-points, ginternal-reset-location-views): New.
1459 * doc/invoke.texi: Document them. Use @itemx where intended.
1460 (gvariable-location-views): Adjust.
1461 * target.def (reset_location_view): New.
1462 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
1463 (TARGET_RESET_LOCATION_VIEW): New.
1464 * doc/tm.texi: Rebuilt.
1465 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
1466 (dwarf2out_default_as_locview_support): New.
1467 (output_asm_line_debug_info): Use option variables.
1468 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
1469 (output_loc_list): Likewise.
1470 (add_high_low_attributes): Check option variables.
1471 Don't output entry view attribute in strict mode.
1472 (gen_inlined_subroutine_die): Check option variables.
1473 (dwarf2out_inline_entry): Likewise.
1474 (init_sections_and_labels): Likewise.
1475 (dwarf2out_early_finish): Likewise.
1476 (maybe_reset_location_view): New, from...
1477 (dwarf2out_var_location): ... here. Call it.
1478 * debug.h (dwarf2out_default_as_loc_support): Declare.
1479 (dwarf2out_default_as_locview_support): Declare.
1480 * hooks.c (hook_int_rtx_insn_0): New.
1481 * hooks.h (hook_int_rtx_insn_0): Declare.
1482 * toplev.c (process_options): Take -gas-loc-support and
1483 -gas-locview-support from dwarf2out. Enable
1484 -gvariable-location-views by default only with locview
1485 assembler support. Enable -ginternal-reset-location-views by
1486 default only if the target defines the corresponding hook.
1487 Enable -ginline-points by default if location views are
1488 enabled; force it disabled if statement frontiers are
1490 * tree-inline.c (expand_call_inline): Check option variables.
1491 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
1493 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
1495 PR tree-optimization/84321
1496 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
1497 handling. Also check whether the anti-range contains any values
1498 that satisfy the mask; switch to a VR_RANGE if not.
1500 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
1503 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
1505 2018-02-13 Martin Jambor <mjambor@suse.cz>
1508 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
1509 of call statements, also set location of a load to a temporary.
1511 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
1513 * config/rl78/rl78.c (add_vector_labels): New function.
1514 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
1515 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
1516 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
1517 which checks that no arguments are passed.
1518 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
1519 * doc/extend.texi: Documentation for the new attribute.
1521 2018-02-13 Andreas Schwab <schwab@suse.de>
1523 * config/riscv/linux.h (CPP_SPEC): Define.
1525 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1528 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
1529 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
1530 OPTION_MASK_ISA_AES as first argument to def_builtin_const
1531 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
1532 instead of OPTION_MASK_ISA_PCLMUL as first argument to
1533 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
1534 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
1535 temporarily for AES and PCLMUL builtins.
1537 PR tree-optimization/84339
1538 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
1539 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
1543 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
1544 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
1545 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
1547 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
1549 * omp-simd-clone.h: New file.
1550 * omp-simd-clone.c: Include omp-simd-clone.h.
1551 (expand_simd_clones): No longer static.
1552 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
1553 cgraph.h and omp-simd-clone.h.
1554 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
1555 (vect_recog_widen_shift_pattern): Formatting fix.
1556 (vect_pattern_recog_1): Don't check optab for calls.
1559 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
1560 operands[2] into a REG before using gen_lowpart on it.
1562 2018-02-12 Jeff Law <law@redhat.com>
1565 * config/sh/sh.c (find_barrier): Consider a sibling call
1568 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
1569 successfully back substituting a reg.
1571 2018-02-12 Richard Biener <rguenther@suse.de>
1573 PR tree-optimization/84037
1574 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
1575 parameter, move visited init to caller.
1576 (vect_slp_analyze_operations): Separate cost from validity
1577 check, initialize visited once for all instances.
1578 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
1580 * tree-vect-stmts.c (vect_model_simple_cost): Make early
1582 (vect_model_promotion_demotion_cost): Likewise.
1583 (vectorizable_bswap): Guard cost modeling with !slp_node
1584 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
1586 (vectorizable_call): Likewise.
1587 (vectorizable_conversion): Likewise.
1588 (vectorizable_assignment): Likewise.
1589 (vectorizable_shift): Likewise.
1590 (vectorizable_operation): Likewise.
1591 (vectorizable_store): Likewise.
1592 (vectorizable_load): Likewise.
1593 (vectorizable_condition): Likewise.
1594 (vectorizable_comparison): Likewise.
1596 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
1599 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
1600 (ASAN_MARK): Fix fnspec to account for return value, change pointer
1601 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
1603 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1606 * params.def (inline-min-speedup): Increase from 8 to 15.
1607 (max-inline-insns-auto): Decrease from 40 to 30.
1608 * ipa-split.c (consider_split): Add some buffer for function to
1609 be considered inlining candidate.
1610 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
1613 2018-02-12 Richard Biener <rguenther@suse.de>
1615 PR tree-optimization/84037
1616 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
1617 matched stmts if we cannot swap the non-matched ones.
1619 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1621 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
1622 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
1623 _mm_maskz_scalef_round_ss): New intrinsics.
1624 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
1625 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
1626 __builtin_ia32_scalefss_round): Remove.
1627 (__builtin_ia32_scalefsd_mask_round,
1628 __builtin_ia32_scalefss_mask_round): New intrinsics.
1629 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
1630 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
1631 ((match_operand:VF_128 2 "<round_nimm_predicate>"
1632 "<round_constraint>")): Changed to ...
1633 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
1634 "<round_scalar_constraint>")): ... this.
1635 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
1636 %0, %1, %2<round_op3>}"): Changed to ...
1637 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
1638 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
1639 %2<round_scalar_mask_op3>}"): ... this.
1640 * config/i386/subst.md (round_scalar_nimm_predicate): New.
1642 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1644 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
1645 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
1646 (_mm_maskz_sqrt_round_ss): New intrinsics.
1647 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
1648 (__builtin_ia32_sqrtsd_mask_round)
1649 (__builtin_ia32_sqrtss_mask_round): New builtins.
1650 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
1651 (__builtin_ia32_sqrtss_round): Remove.
1652 (__builtin_ia32_sqrtsd_mask_round)
1653 (__builtin_ia32_sqrtss_mask_round): New builtins.
1654 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
1655 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
1656 ((match_operand:VF_128 1 "vector_operand"
1657 "xBm,<round_constraint>")): Changed to ...
1658 ((match_operand:VF_128 1 "vector_operand"
1659 "xBm,<round_scalar_constraint>")): ... this.
1660 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
1661 %0, %2, %<iptr>1<round_op3>}): Changed to ...
1662 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
1663 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
1664 %<iptr>1<round_scalar_mask_op3>}): ... this.
1665 ((set_attr "prefix" "<round_prefix>")): Changed to ...
1666 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
1668 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
1671 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
1672 Cast vec_cmpeq result to correct type.
1673 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
1674 Cast vec_cmpgt result to correct type.
1676 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
1678 * final.c (final_scan_insn_1): Renamed from...
1679 (final_scan_insn): ... this. New wrapper, to recover
1680 seen from the outermost call in recursive ones.
1681 * config/sparc/sparc.c (output_return): Drop seen from call.
1682 (output_sibcall): Likewise.
1683 * config/visium/visium.c (output_branch): Likewise.
1685 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
1687 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
1690 2018-02-10 Alan Modra <amodra@gmail.com>
1693 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
1694 Specify LR as an input.
1696 2018-02-10 Jakub Jelinek <jakub@redhat.com>
1699 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
1700 remove_member_access_dummy_vars): New functions.
1701 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
1702 lower_omp_1, execute_lower_omp): Use them.
1704 PR rtl-optimization/84308
1705 * shrink-wrap.c (spread_components): Release todo vector.
1707 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
1709 PR rtl-optimization/57193
1710 * ira-color.c (struct allocno_color_data): Add member
1711 conflict_allocno_hard_prefs.
1712 (update_conflict_allocno_hard_prefs): New.
1713 (bucket_allocno_compare_func): Add a preference based on
1714 conflict_allocno_hard_prefs.
1715 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
1716 (color_allocnos): Remove a dead code. Initiate
1717 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
1719 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1722 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
1723 constraint from =wa to wa. Avoid a subreg on the output operand,
1724 instead use a pseudo and subreg it in a move.
1725 (p9_xxbrd_<mode>): Changed to ...
1726 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
1727 (p9_xxbrd_v2df): New expander.
1728 (p9_xxbrw_<mode>): Changed to ...
1729 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
1730 (p9_xxbrw_v4sf): New expander.
1732 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1734 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
1736 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
1739 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
1740 multiply in 32-bit mode.
1741 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
1742 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
1745 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1747 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
1748 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
1749 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
1750 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
1752 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
1755 * dwarf2out.c (is_trivial_indirect_ref): New function.
1756 (dwarf2out_late_global_decl): Do not generate a location
1757 attribute for variables that have a non-trivial DECL_VALUE_EXPR
1758 and that are not defined in the current unit.
1760 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1762 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
1763 instead of a libcall for UNORDERED.
1765 2018-02-09 Tamar Christina <tamar.christina@arm.com>
1768 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
1769 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
1771 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1774 * config/s390/s390.c (s390_set_current_function): Invoke
1775 s390_indirect_branch_settings also if fndecl didn't change.
1777 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1779 * config/rs6000/rs6000.md (blockage): Set length to zero.
1781 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1783 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
1785 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1788 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
1789 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
1793 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
1794 PARALLEL incoming that failed vt_get_decl_and_offset check.
1797 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
1798 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
1800 (get_variable_section): For decls in named .bss* sections pass true as
1801 second argument to bss_initializer_p.
1803 2018-02-09 Marek Polacek <polacek@redhat.com>
1804 Jakub Jelinek <jakub@redhat.com>
1807 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
1808 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
1809 Sync some changes from cxx_fold_indirect_ref.
1811 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1813 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
1815 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
1816 (BLOCK_INLINE_ENTRY_LABEL): New.
1817 (dwarf2out_var_location): Disregard inline entry markers.
1818 (inline_entry_data): New struct.
1819 (inline_entry_data_hasher): New hashtable type.
1820 (inline_entry_data_hasher::hash): New.
1821 (inline_entry_data_hasher::equal): New.
1822 (inline_entry_data_table): New variable.
1823 (add_high_low_attributes): Add DW_AT_entry_pc and
1824 DW_AT_GNU_entry_view attributes if a pending entry is found
1825 in inline_entry_data_table. Add old entry_pc attribute only
1826 if debug nonbinding markers are disabled.
1827 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
1828 markers are enabled.
1829 (block_within_block_p, dwarf2out_inline_entry): New.
1830 (dwarf2out_finish): Check that no entries remained in
1831 inline_entry_data_table.
1832 * final.c (reemit_insn_block_notes): Handle inline entry notes.
1833 (final_scan_insn, notice_source_line): Likewise.
1834 (rest_of_clean_state): Skip inline entry markers.
1835 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
1837 * gimple.c (gimple_build_debug_inline_entry): New.
1838 * gimple.h (enum gimple_debug_subcode): Add
1839 GIMPLE_DEBUG_INLINE_ENTRY.
1840 (gimple_build_debug_inline_entry): Declare.
1841 (gimple_debug_inline_entry_p): New.
1842 (gimple_debug_nonbind_marker_p): Adjust.
1843 * insn-notes.def (INLINE_ENTRY): New.
1844 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
1845 inline entry marker notes.
1846 (print_insn): Likewise.
1847 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
1848 (INSN_DEBUG_MARKER_KIND): Likewise.
1849 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
1850 * tree-inline.c (expand_call_inline): Build and insert
1851 debug_inline_entry stmt.
1852 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
1853 inline entry blocks early, if nonbind markers are enabled.
1854 (dump_scope_block): Dump fragment info.
1855 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
1856 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
1857 (gimple_build_debug_inline_entry): New.
1858 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
1859 Enable/disable inline entry points too.
1860 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
1861 (DEBUG_INSN): Describe inline entry markers.
1863 * common.opt (gvariable-location-views): New.
1864 (gvariable-location-views=incompat5): New.
1865 * config.in: Rebuilt.
1866 * configure: Rebuilt.
1867 * configure.ac: Test assembler for view support.
1868 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
1869 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
1870 * dwarf2out.c (var_loc_view): New typedef.
1871 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
1872 (dwarf2out_locviews_in_attribute): New.
1873 (dwarf2out_locviews_in_loclist): New.
1874 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
1875 (enum dw_line_info_opcode): Add LI_adv_address.
1876 (struct dw_line_info_table): Add view.
1877 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
1878 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
1879 (zero_view_p): New variable.
1880 (ZERO_VIEW_P): New macro.
1881 (output_asm_line_debug_info): New.
1882 (struct var_loc_node): Add view.
1883 (add_AT_view_list, AT_loc_list): New.
1884 (add_var_loc_to_decl): Add view param. Test it against last.
1885 (new_loc_list): Add view params. Record them.
1886 (AT_loc_list_ptr): Handle loc and view lists.
1887 (view_list_to_loc_list_val_node): New.
1888 (print_dw_val): Handle dw_val_class_view_list.
1889 (size_of_die): Likewise.
1890 (value_format): Likewise.
1891 (loc_list_has_views): New.
1892 (gen_llsym): Set vl_symbol too.
1893 (maybe_gen_llsym, skip_loc_list_entry): New.
1894 (dwarf2out_maybe_output_loclist_view_pair): New.
1895 (output_loc_list): Output view list or entries too.
1896 (output_view_list_offset): New.
1897 (output_die): Handle dw_val_class_view_list.
1898 (output_dwarf_version): New.
1899 (output_compilation_unit_header): Use it.
1900 (output_skeleton_debug_sections): Likewise.
1901 (output_rnglists, output_line_info): Likewise.
1902 (output_pubnames, output_aranges): Update version comments.
1903 (output_one_line_info_table): Output view numbers in asm comments.
1904 (dw_loc_list): Determine current endview, pass it to new_loc_list.
1905 Call maybe_gen_llsym.
1906 (loc_list_from_tree_1): Adjust.
1907 (add_AT_location_description): Create view list attribute if
1908 needed, check it's absent otherwise.
1909 (convert_cfa_to_fb_loc_list): Adjust.
1910 (maybe_emit_file): Call output_asm_line_debug_info for test.
1911 (dwarf2out_var_location): Reset views as needed. Precompute
1912 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
1913 attribute. Set view.
1914 (new_line_info_table): Reset next view.
1915 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
1916 (dwarf2out_source_line): Likewise. Output view resets and labels to
1917 the assembler, or select appropriate line info opcodes.
1918 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
1919 (optimize_string_length): Catch it. Adjust.
1920 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
1921 dw_val_class_view_list, and remove it if no longer needed.
1922 (hash_loc_list): Hash view numbers.
1923 (loc_list_hasher::equal): Compare them.
1924 (optimize_location_lists): Check whether a view list symbol is
1925 needed, and whether the locview attribute is present, and
1926 whether they match. Remove the locview attribute if no longer
1928 (index_location_lists): Call skip_loc_list_entry for test.
1929 (dwarf2out_finish): Call output_asm_line_debug_info for test.
1930 Use output_dwarf_version.
1931 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
1932 (struct dw_val_node): Add val_view_list.
1933 * final.c (SEEN_NEXT_VIEW): New.
1934 (set_next_view_needed): New.
1935 (clear_next_view_needed): New.
1936 (maybe_output_next_view): New.
1937 (final_start_function): Rename to...
1938 (final_start_function_1): ... this. Take pointer to FIRST,
1939 add SEEN parameter. Emit param bindings in the initial view.
1940 (final_start_function): Reintroduce SEEN-less interface.
1941 (final): Rename to...
1942 (final_1): ... this. Take SEEN parameter. Output final pending
1943 next view at the end.
1944 (final): Reintroduce seen-less interface.
1945 (final_scan_insn): Output pending next view before switching
1946 sections or ending a block. Mark the next view as needed when
1947 outputting variable locations. Notify debug backend of section
1948 changes, and of location view changes.
1949 (rest_of_handle_final): Adjust.
1950 * toplev.c (process_options): Autodetect value for debug variable
1951 location views option. Warn on incompat5 without -gdwarf-5.
1952 * doc/invoke.texi (gvariable-location-views): New.
1953 (gvariable-location-views=incompat5): New.
1954 (gno-variable-location-views): New.
1956 2018-02-08 David Malcolm <dmalcolm@redhat.com>
1958 PR tree-optimization/84136
1959 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
1960 that the result of find_edge is non-NULL.
1962 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1965 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
1966 storing integer register in SImode. Fix cost of 256 and 512
1967 byte aligned SSE register store.
1969 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1971 * config/i386/i386.c (ix86_multiplication_cost): Fix
1972 multiplication cost for TARGET_AVX512DQ.
1974 2018-02-08 Marek Polacek <polacek@redhat.com>
1976 PR tree-optimization/84238
1977 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
1980 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1982 PR tree-optimization/84265
1983 * tree-vect-stmts.c (vectorizable_store): Don't treat
1984 VMAT_CONTIGUOUS accesses as grouped.
1985 (vectorizable_load): Likewise.
1987 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1989 PR tree-optimization/81635
1990 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
1991 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
1992 (test_round_for_mask): New functions.
1993 (wide_int_cc_tests): Call test_round_for_mask.
1994 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
1995 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
1996 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
1997 range returned by get_range_info.
1999 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
2002 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
2003 * symtab.c: Include builtins.h
2004 (symtab_node::output_to_lto_symbol_table_p): Move here
2005 from lto-streamer-out.c:output_symbol_p.
2006 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
2007 (output_symbol_p): Move all logic to symtab.c
2008 (produce_symtab): Update.
2010 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
2012 * config/s390/s390-opts.h (enum indirect_branch): Define.
2013 * config/s390/s390-protos.h (s390_return_addr_from_memory)
2014 (s390_indirect_branch_via_thunk)
2015 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
2016 (enum s390_indirect_branch_type): Define.
2017 * config/s390/s390.c (struct s390_frame_layout, struct
2018 machine_function): Remove.
2019 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
2020 (indirect_branch_table_label_no, indirect_branch_table_name):
2022 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
2023 (enum s390_indirect_branch_option): Define.
2024 (s390_return_addr_from_memory): New function.
2025 (s390_handle_string_attribute): New function.
2026 (s390_attribute_table): Add new attribute handler.
2027 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
2028 (s390_indirect_branch_via_thunk): New function.
2029 (s390_indirect_branch_via_inline_thunk): New function.
2030 (s390_function_ok_for_sibcall): When jumping via thunk disallow
2031 sibling call optimization for non z10 compiles.
2032 (s390_emit_call): Force indirect branch target to be a single
2033 register. Add r1 clobber for non-z10 compiles.
2034 (s390_emit_epilogue): Emit return jump via return_use expander.
2035 (s390_reorg): Handle JUMP_INSNs as execute targets.
2036 (s390_option_override_internal): Perform validity checks for the
2037 new command line options.
2038 (s390_indirect_branch_attrvalue): New function.
2039 (s390_indirect_branch_settings): New function.
2040 (s390_set_current_function): Invoke s390_indirect_branch_settings.
2041 (s390_output_indirect_thunk_function): New function.
2042 (s390_code_end): Implement target hook.
2043 (s390_case_values_threshold): Implement target hook.
2044 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
2046 * config/s390/s390.h (struct s390_frame_layout)
2047 (struct machine_function): Move here from s390.c.
2048 (TARGET_INDIRECT_BRANCH_NOBP_RET)
2049 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
2050 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
2051 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
2052 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
2053 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
2054 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
2055 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
2056 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
2057 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
2058 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
2059 (mnemonic attribute): Add values which aren't recognized
2061 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
2062 pattern for branch conversion. Fix mnemonic attribute.
2063 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
2064 indirect branch via thunk if requested.
2065 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
2066 ("*indirect_jump"): Disable for branch conversion using out of
2068 ("indirect_jump_via_thunk<mode>_z10")
2069 ("indirect_jump_via_thunk<mode>")
2070 ("indirect_jump_via_inlinethunk<mode>_z10")
2071 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
2072 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
2073 ("casesi_jump_via_inlinethunk<mode>_z10")
2074 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
2075 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
2076 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
2077 ("*indirect2_jump"): Disable for branch conversion.
2078 ("casesi_jump"): Turn into expander and expand patterns for branch
2080 ("return_use"): New expander.
2081 ("*return"): Emit return via thunk and rename it to ...
2082 ("*return<mode>"): ... this one.
2083 * config/s390/s390.opt: Add new options and and enum for the
2086 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2088 * lra-constraints.c (match_reload): Unconditionally use
2089 gen_lowpart_SUBREG, rather than selecting between that
2090 and equivalent gen_rtx_SUBREG code.
2092 2018-02-08 Richard Biener <rguenther@suse.de>
2094 PR tree-optimization/84233
2095 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
2096 changed flag instead of boguously re-using phi_inserted.
2098 2018-02-08 Martin Jambor <mjambor@suse.cz>
2100 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
2101 static local variables.
2103 2018-02-08 Richard Biener <rguenther@suse.de>
2105 PR tree-optimization/84278
2106 * tree-vect-stmts.c (vectorizable_store): When looking for
2107 smaller vector types to perform grouped strided loads/stores
2108 make sure the mode is supported by the target.
2109 (vectorizable_load): Likewise.
2111 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2113 * config/aarch64/aarch64.c (aarch64_components_for_bb):
2114 Increase LDP/STP opportunities by adding adjacent callee-saves.
2116 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2118 PR rtl-optimization/84068
2119 PR rtl-optimization/83459
2120 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
2122 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
2124 PR tree-optimization/84224
2125 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
2126 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
2129 2018-02-07 Iain Sandoe <iain@codesourcery.com>
2132 * config/rs6000/altivec.md (*restore_world): Remove LR use.
2133 * config/rs6000/predicates.md (restore_world_operation): Adjust op
2134 count, remove one USE.
2136 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
2138 * doc/install.texi (Configuration): Document the
2139 --with-long-double-format={ibm,ieee} PowerPC configuration
2143 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
2144 Convert from define_expand to be define_insn_and_split. Rework
2145 float/double/_Float128 conversions to QI/HI/SImode to work with
2146 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
2147 conversions to QI/HImode types did a store and then a load to
2148 truncate the value. For conversions to VSX registers, don't split
2149 the insn, instead emit the code directly. Use the code iterator
2150 any_fix to combine signed and unsigned conversions.
2151 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
2152 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
2153 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2154 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
2155 (fix_<mode>di2_hw): Likewise.
2156 (fixuns_<mode>di2_hw): Likewise.
2157 (fix_<mode>si2_hw): Likewise.
2158 (fixuns_<mode>si2_hw): Likewise.
2159 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
2160 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2161 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
2162 fix<uns>_trunc<SFDF:mode>si2_p8.
2163 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
2165 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
2166 (fix<uns>_<mode>_mem): Likewise.
2167 (fctiw<u>z_<mode>_mem): Likewise.
2168 (fix<uns>_<mode>_mem): Likewise.
2169 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
2170 the register allocator from doing a direct move to the GPRs to do
2171 a store, and instead use the ISA 3.0 store byte/half-word from
2172 vector register instruction. For IEEE 128-bit floating point,
2173 also optimize stores of 32-bit ints.
2174 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
2176 2018-02-07 Alan Hayward <alan.hayward@arm.com>
2178 * genextract.c (push_pathstr_operand): New function to support
2180 (walk_rtx): Call push_pathstr_operand.
2181 (print_path): Support [a-zA-Z].
2183 2018-02-07 Richard Biener <rguenther@suse.de>
2185 PR tree-optimization/84037
2186 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
2187 (cse_and_gimplify_to_preheader): Declare.
2188 (vect_get_place_in_interleaving_chain): Likewise.
2189 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
2191 (_loop_vec_info::~_loop_vec_info): Delete it.
2192 (cse_and_gimplify_to_preheader): New function.
2193 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
2194 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
2195 (vectorizable_load): Likewise. For grouped stores always base
2196 the IV on the first element.
2197 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
2198 condition before gimplifying.
2200 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2202 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
2203 *DIV_EXPR and *MOD_EXPR.
2205 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
2208 * config/i386/i386.c (ix86_option_override_internal): Mask out
2209 the CF_SET bit when checking -fcf-protection.
2211 2018-02-07 Tom de Vries <tom@codesourcery.com>
2214 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
2217 2018-02-07 Richard Biener <rguenther@suse.de>
2219 PR tree-optimization/84204
2220 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
2223 PR tree-optimization/84205
2224 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
2225 special-case isl_ast_op_zdiv_r.
2227 PR tree-optimization/84223
2228 * graphite-scop-detection.c (gather_bbs::before_dom_children):
2229 Only add conditions from within the region.
2230 (gather_bbs::after_dom_children): Adjust.
2232 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
2235 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
2236 * config/avr/avr.md: Only post-reload split REG-REG moves if
2237 either register is GENERAL_REG_P.
2239 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2241 PR tree-optimization/84235
2242 * tree-ssa-scopedtables.c
2243 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
2244 if the subtraction is performed in floating point type where NaNs are
2245 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
2246 build 1. Formatting fix.
2248 2018-02-06 Jakub Jelinek <jakub@redhat.com>
2251 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
2252 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
2253 and skip it regardless of bb boundaries. Use CALL_P macro,
2254 don't test INSN_P (insn) together with CALL_P or JUMP_P check
2255 unnecessarily, formatting fix.
2257 2018-02-06 Michael Collison <michael.collison@arm.com>
2259 * config/arm/thumb2.md:
2260 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
2261 (*thumb_mov_notscc): Ditto.
2263 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
2266 * config/rs6000/rs6000.md (su code attribute): Use "u" for
2267 unsigned_fix, not "s".
2269 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2271 * configure.ac (gcc_fn_eh_frame_ro): New function.
2272 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
2273 correct .eh_frame permissions.
2274 * configure: Regenerate.
2276 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
2278 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
2281 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2283 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2284 Display warning message for -mno-speculate-indirect-jumps.
2286 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
2288 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
2290 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
2292 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
2294 PR tree-optimization/84225
2295 * tree-eh.c (find_trapping_overflow): Only call
2296 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
2298 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2301 * config/i386/i386.c: Reimplement the check of possible options
2302 -mibt/-mshstk conbination. Change error messages.
2303 * doc/invoke.texi: Fix a typo: remove extra '='.
2305 2018-02-06 Marek Polacek <polacek@redhat.com>
2307 PR tree-optimization/84228
2308 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
2310 2018-02-06 Tamar Christina <tamar.christina@arm.com>
2313 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
2314 emitted arch directives.
2315 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
2316 __ARM_FEATURE_COPROC before changing architectures.
2318 2018-02-06 Richard Biener <rguenther@suse.de>
2320 * config/i386/i386.c (print_reg): Fix typo.
2321 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
2323 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
2325 * configure: Regenerate.
2327 2018-02-05 Martin Sebor <msebor@redhat.com>
2329 PR tree-optimization/83369
2330 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
2333 2018-02-05 Martin Liska <mliska@suse.cz>
2335 * doc/invoke.texi: Cherry-pick upstream r323995.
2337 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
2339 * ira.c (ira_init_register_move_cost): Adjust comment.
2341 2018-02-05 Martin Liska <mliska@suse.cz>
2343 PR gcov-profile/84137
2344 * doc/gcov.texi: Fix typo in documentation.
2346 2018-02-05 Martin Liska <mliska@suse.cz>
2348 PR gcov-profile/83879
2349 * doc/gcov.texi: Document necessity of --dynamic-list-data when
2350 using dlopen functionality.
2352 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
2354 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
2355 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
2356 _mm_maskz_range_ss, _mm_mask_range_round_ss,
2357 _mm_maskz_range_round_ss): New intrinsics.
2358 (__builtin_ia32_rangesd128_round)
2359 (__builtin_ia32_rangess128_round): Remove.
2360 (__builtin_ia32_rangesd128_mask_round,
2361 __builtin_ia32_rangess128_mask_round): New builtins.
2362 * config/i386/i386-builtin.def (__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/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
2367 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
2368 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
2369 "<round_saeonly_constraint>")): Changed to ...
2370 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
2371 "<round_saeonly_scalar_constraint>")): ... this.
2372 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
2373 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
2374 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
2375 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
2376 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
2378 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
2380 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
2382 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
2383 Remove all values except native, 8540 and 8548.
2385 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
2387 * config/i386/i386.c (ix86_output_function_return): Pass
2388 INVALID_REGNUM, instead of -1, as invalid register number to
2389 indirect_thunk_name and output_indirect_thunk.
2391 2018-02-02 Julia Koval <julia.koval@intel.com>
2393 * config.gcc: Add -march=icelake.
2394 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
2395 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
2396 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
2397 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
2398 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
2399 (processor_target_table): Add icelake.
2400 (ix86_option_override_internal): Handle new PTAs.
2401 (get_builtin_code_for_version): Handle icelake.
2402 (M_INTEL_COREI7_ICELAKE): New.
2403 (fold_builtin_cpu): Handle icelake.
2404 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
2405 * doc/invoke.texi: Add -march=icelake.
2407 2018-02-02 Julia Koval <julia.koval@intel.com>
2409 * config/i386/i386.c (ix86_option_override_internal): Change flags type
2410 to wide_int_bitmask.
2411 * wide-int-bitmask.h: New.
2413 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2416 * config/i386/i386.md: Replace Pmode with word_mode in
2417 builtin_setjmp_setup and builtin_longjmp to support x32.
2419 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
2423 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
2425 (rs6000_supported_cpu_names): New static variable.
2426 (linux_cpu_translation_table): Likewise.
2427 (elf_platform) <cpu>: Define new static variable and use it.
2428 Translate kernel AT_PLATFORM name to canonical name if needed.
2429 Error if platform name is unknown.
2431 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
2434 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
2436 2018-02-01 Jeff Law <law@redhat.com>
2439 * config/i386/i386.c (release_scratch_register_on_entry): Add new
2440 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
2441 the scratch if RELEASE_VIA_POP is false.
2442 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
2443 If we have to save a temporary register, decrement SIZE appropriately.
2444 Pass new arguments to release_scratch_register_on_entry.
2445 (ix86_adjust_stack_and_probe): Likewise.
2446 (ix86_emit_probe_stack_range): Pass new arguments to
2447 release_scratch_register_on_entry.
2449 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
2451 PR rtl-optimization/84157
2452 * combine.c (change_zero_ext): Use REG_P predicate in
2453 front of HARD_REGISTER_P predicate.
2455 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
2457 * config/avr/avr.c (avr_option_override): Move disabling of
2458 -fdelete-null-pointer-checks to...
2459 * common/config/avr/avr-common.c (avr_option_optimization_table):
2462 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2464 PR tree-optimization/81635
2465 * tree-data-ref.c (split_constant_offset_1): For types that
2466 wrap on overflow, try to use range info to prove that wrapping
2469 2018-02-01 Renlin Li <renlin.li@arm.com>
2472 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
2474 (aarch64_register_move_cost): Likewise.
2475 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
2477 (REG_CLASS_NAMES): Likewise.
2478 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
2479 TAILCALL_ADDR_REGS. Remove IP registers.
2480 * config/aarch64/aarch64.md (Ucs): Update register constraint.
2482 2018-02-01 Richard Biener <rguenther@suse.de>
2484 * domwalk.h (dom_walker::dom_walker): Add additional constructor
2485 for specifying RPO order and allow NULL for that.
2486 * domwalk.c (dom_walker::dom_walker): Likewise.
2487 (dom_walker::walk): Handle NULL RPO order.
2488 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
2490 (rewrite_update_dom_walker): Likewise.
2491 (mark_def_dom_walker): Likewise.
2493 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2495 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
2496 (aarch64_maybe_expand_sve_subreg_move): Declare.
2497 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
2498 * config/aarch64/predicates.md (aarch64_any_register_operand): New
2500 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
2501 that are semantically a reverse operation.
2502 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
2503 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
2504 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
2506 (aarch64_can_change_mode_class): For big-endian, forbid changes
2507 between two SVE modes if they have different element sizes.
2509 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2511 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
2512 the TImode handling for big-endian targets.
2514 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2516 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
2517 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
2519 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
2520 Remove BSWAP handing for big-endian targets and use the form of
2521 LD1RQ appropariate for the mode.
2523 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2525 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
2526 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
2529 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2532 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
2533 check for operands that need to go through aarch64_sve_reload_be.
2535 2018-02-01 Jakub Jelinek <jakub@redhat.com>
2537 PR tree-optimization/81661
2538 PR tree-optimization/84117
2539 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
2540 * tree-eh.c: Include gimplify.h.
2541 (find_trapping_overflow, replace_trapping_overflow,
2542 rewrite_to_non_trapping_overflow): New functions.
2543 * tree-vect-loop.c: Include tree-eh.h.
2544 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
2545 * tree-data-ref.c: Include tree-eh.h.
2546 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
2548 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
2550 PR rtl-optimization/84123
2551 * combine.c (change_zero_ext): Check if hard register satisfies
2552 can_change_dest_mode before calling gen_lowpart_SUBREG.
2554 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
2557 * ira.c (ira_init_register_move_cost): Remove assert.
2559 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2561 PR rtl-optimization/84071
2562 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
2563 * doc/tm.texi: Regenerate.
2565 2018-01-31 Richard Biener <rguenther@suse.de>
2567 PR tree-optimization/84132
2568 * tree-data-ref.c (analyze_miv_subscript): Properly
2569 check whether evolution_function_is_affine_multivariate_p
2570 before calling gcd_of_steps_may_divide_p.
2572 2018-01-31 Julia Koval <julia.koval@intel.com>
2575 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
2576 * config/i386/i386.md (rdpid_rex64) New.
2577 (rdpid): Make 32bit only.
2579 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
2582 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
2583 an IDENTIFIER_NODE for FUNCTION_TYPE's.
2585 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2588 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
2590 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
2592 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2594 PR rtl-optimization/84071
2595 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
2596 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
2598 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2600 * config/arc/arc.c (arc_handle_aux_attribute): New function.
2601 (arc_attribute_table): Add 'aux' attribute.
2602 (arc_in_small_data_p): Consider aux like variables.
2603 (arc_is_aux_reg_p): New function.
2604 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
2605 (arc_get_aux_arg): New function.
2606 (prepare_move_operands): Handle aux-register access.
2607 (arc_handle_aux_attribute): New function.
2608 * doc/extend.texi (ARC Variable attributes): Add subsection.
2610 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2612 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
2613 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
2614 (arc_attribute_table): Add 'uncached' attribute.
2615 (arc_print_operand): Print '.di' flag for uncached memory
2617 (arc_in_small_data_p): Do not consider for small data the uncached
2619 (arc_is_uncached_mem_p): New function.
2620 * config/arc/predicates.md (compact_store_memory_operand): Check
2621 for uncached memory accesses.
2622 (nonvol_nonimm_operand): Likewise.
2623 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
2625 2018-01-31 Jakub Jelinek <jakub@redhat.com>
2628 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
2629 falign-loops=): Add Optimization flag.
2631 2018-01-30 Jeff Law <law@redhat.com>
2634 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
2635 INT_REGISTERS_SAVED. Check it prior to calling
2636 get_scratch_register_on_entry.
2637 (ix86_adjust_stack_and_probe): Similarly.
2638 (ix86_emit_probe_stack_range): Similarly.
2639 (ix86_expand_prologue): Corresponding changes.
2641 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2644 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
2645 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
2647 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
2650 * lra-constraints.c (curr_insn_transform): Process AND in the
2653 2018-01-30 Jakub Jelinek <jakub@redhat.com>
2655 PR rtl-optimization/83986
2656 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
2657 dependence against last_pending_memory_flush in addition to
2660 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
2662 PR tree-optimization/81611
2663 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
2666 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2669 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
2672 2018-01-30 Richard Biener <rguenther@suse.de>
2673 Jakub Jelinek <jakub@redhat.com>
2675 PR tree-optimization/84111
2676 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
2677 inner loops added during recursion, as they don't have up-to-date
2680 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2683 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
2684 (can_inline_edge_by_limits_p): ... here.
2685 (can_early_inline_edge_p, check_callers,
2686 update_caller_keys, update_callee_keys, recursive_inlining,
2687 add_new_edges_to_heap, speculation_useful_p,
2688 inline_small_functions,
2689 inline_small_functions, flatten_function,
2690 inline_to_all_callers_1): Update.
2692 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2694 * profile-count.c (profile_count::combine_with_ipa_count): Handle
2697 2018-01-30 Richard Biener <rguenther@suse.de>
2699 PR tree-optimization/83008
2700 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
2701 invariant and constant vector uses in stmts when they need
2704 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2707 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
2708 * configure: Regenerate.
2710 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2712 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
2714 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
2715 Use gen_rtx_REG rather than gen_lowpart.
2717 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2719 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
2720 rather than 0 when creating partial subregs.
2722 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2724 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
2727 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
2730 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
2731 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
2732 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
2733 flags. This restores the settings used before the 2017-07-24.
2734 Turning off pre increment/decrement/modify allows IVOPTS to
2735 optimize DF/SF loops where the index is an int.
2737 2018-01-29 Richard Biener <rguenther@suse.de>
2738 Kelvin Nilsen <kelvin@gcc.gnu.org>
2741 * tree-vect-stmts.c (vectorizable_call): Don't call
2742 targetm.vectorize_builtin_md_vectorized_function if callee is
2745 2018-01-22 Carl Love <cel@us.ibm.com>
2747 * doc/extend.tex: Fix typo in second arg in
2748 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
2750 2018-01-29 Richard Biener <rguenther@suse.de>
2752 PR tree-optimization/84086
2753 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
2754 (flush_ssaname_freelist): When SSA names were released reset
2755 the SCEV hash table.
2757 2018-01-29 Richard Biener <rguenther@suse.de>
2759 PR tree-optimization/84057
2760 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
2761 removed paths when removing edges.
2763 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
2765 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
2766 -mfunction-return=@var{choice}.
2768 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
2771 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
2773 (layout::print_source_line): Likewise.
2774 (test_get_line_width_without_trailing_whitespace): Add test cases.
2776 2018-01-27 Jakub Jelinek <jakub@redhat.com>
2779 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
2782 2018-01-26 Jim Wilson <jimw@sifive.com>
2784 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
2786 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
2789 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2791 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
2792 and CMP + SUB-immediate -> SUBS.
2794 2018-01-26 Martin Sebor <msebor@redhat.com>
2796 PR tree-optimization/83896
2797 * tree-ssa-strlen.c (get_string_len): Rename...
2798 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
2799 Avoid assuming length is constant.
2800 (handle_char_store): Use HOST_WIDE_INT for string length.
2802 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
2805 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
2806 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
2808 2018-01-26 Richard Biener <rguenther@suse.de>
2810 PR rtl-optimization/84003
2811 * dse.c (record_store): Only record redundant stores when
2812 the earlier store aliases at least all accesses the later one does.
2814 2018-01-26 Jakub Jelinek <jakub@redhat.com>
2816 PR rtl-optimization/83985
2817 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
2818 REG_CFA_RESTORE insns.
2819 (delete_unmarked_insns): Don't ignore separate shrink wrapping
2820 REG_CFA_RESTORE insns here.
2823 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
2824 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
2826 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2828 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
2829 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
2830 (arc_init): Likewise.
2831 (arc_override_options): Likewise.
2832 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
2834 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
2836 * config/arc/arc.h (TARGET_DBNZ): Define.
2837 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
2838 properly set the tune attribute.
2839 (dbnz): Use TARGET_DBNZ guard.
2840 * config/arc/arc.opt (mtune): Add core3 option.
2842 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2844 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
2845 recognize new pic like addresses.
2846 (arc_delegitimize_address): Clean up.
2848 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2850 * config/arc/arc-arches.def: Option mrf16 valid for all
2852 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
2853 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
2854 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
2855 * config/arc/arc-tables.opt: Regenerate.
2856 * config/arc/arc.c (arc_conditional_register_usage): Handle
2857 reduced register file case.
2858 (arc_file_start): Set must have build attributes.
2859 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
2861 * config/arc/arc.opt (mrf16): Add new option.
2862 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
2863 * config/arc/genmultilib.awk: Handle new mrf16 option.
2864 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
2865 * config/arc/t-multilib: Regenerate.
2866 * doc/invoke.texi (ARC Options): Document mrf16 option.
2868 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2870 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
2871 * config/arc/arc.c (arc_handle_secure_attribute): New function.
2872 (arc_attribute_table): Add 'secure_call' attribute.
2873 (arc_print_operand): Print secure call operand.
2874 (arc_function_ok_for_sibcall): Don't optimize tail calls when
2876 (arc_is_secure_call_p): New function. * config/arc/arc.md
2877 (call_i): Add support for sjli instruction.
2878 (call_value_i): Likewise.
2879 * config/arc/constraints.md (Csc): New constraint.
2881 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2882 John Eric Martin <John.Martin@emmicro-us.com>
2884 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
2885 * config/arc/arc.c (_arc_jli_section): New struct.
2886 (arc_jli_section): New type.
2887 (rc_jli_sections): New static variable.
2888 (arc_handle_jli_attribute): New function.
2889 (arc_attribute_table): Add jli_always and jli_fixed attribute.
2890 (arc_file_end): New function.
2891 (TARGET_ASM_FILE_END): Define.
2892 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
2893 (arc_add_jli_section): New function.
2894 (jli_call_scan): Likewise.
2895 (arc_reorg): Call jli_call_scan.
2896 (arc_output_addsi): Remove 'S' from printing asm operand.
2897 (arc_is_jli_call_p): New function.
2898 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
2900 (movhi_insn): Likewise.
2901 (movsi_insn): Likewise.
2902 (movsi_set_cc_insn): Likewise.
2903 (loadqi_update): Likewise.
2904 (load_zeroextendqisi_update): Likewise.
2905 (load_signextendqisi_update): Likewise.
2906 (loadhi_update): Likewise.
2907 (load_zeroextendhisi_update): Likewise.
2908 (load_signextendhisi_update): Likewise.
2909 (loadsi_update): Likewise.
2910 (loadsf_update): Likewise.
2911 (movsicc_insn): Likewise.
2912 (bset_insn): Likewise.
2913 (bxor_insn): Likewise.
2914 (bclr_insn): Likewise.
2915 (bmsk_insn): Likewise.
2916 (bicsi3_insn): Likewise.
2917 (cmpsi_cc_c_insn): Likewise.
2918 (movsi_ne): Likewise.
2919 (movsi_cond_exec): Likewise.
2920 (clrsbsi2): Likewise.
2931 (call_i): Remove 'S' asm letter, add jli instruction.
2932 (call_value_i): Likewise.
2933 * config/arc/arc.op (mjli-always): New option.
2934 * config/arc/constraints.md (Cji): New constraint.
2935 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
2937 (subsf3_fpx): Likewise.
2938 (mulsf3_fpx): Likewise.
2939 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
2941 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
2943 * doc/invoke.texi (ARC): Document mjli-always option.
2945 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
2947 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
2948 avoid addition with 0 and use incw and decw where possible.
2950 2018-01-26 Richard Biener <rguenther@suse.de>
2952 PR tree-optimization/81082
2953 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
2954 association if it requires casting to unsigned.
2955 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
2956 from fold_plusminus_mult_expr to catch important cases late when
2957 range info is available.
2959 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2961 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
2962 * configure.ac (hidden_linkonce): New test.
2963 * configure: Regenerate.
2964 * config.in: Regenerate.
2966 2018-01-26 Julia Koval <julia.koval@intel.com>
2968 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
2969 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
2970 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
2971 _mm_mask_bitshuffle_epi64_mask): Fix type.
2972 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
2973 USI_FTYPE_V4DI_V4DI_USI): Remove.
2974 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
2975 __builtin_ia32_vpshufbitqmb256_mask,
2976 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
2977 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
2978 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
2980 2018-01-26 Alan Modra <amodra@gmail.com>
2983 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
2984 UNSPEC_VBPERMQ. Sort other unspecs.
2986 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
2988 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
2990 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
2993 * predict.c (drop_profile): Do not push/pop cfun; update also
2995 (handle_missing_profiles): Fix logic looking for zero profiles.
2997 2018-01-25 Jakub Jelinek <jakub@redhat.com>
3000 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
3001 on functions with #pragma omp declare simd or functions with simd
3003 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
3004 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
3005 Remove trailing \n from warning_at calls.
3007 2018-01-25 Tom de Vries <tom@codesourcery.com>
3010 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3011 for neutered workers.
3013 2018-01-24 Joseph Myers <joseph@codesourcery.com>
3016 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
3017 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
3019 2018-01-24 Jeff Law <law@redhat.com>
3022 * i386.c (get_probe_interval): Move to earlier point.
3023 (ix86_compute_frame_layout): If -fstack-clash-protection and
3024 the frame is larger than the probe interval, then use pushes
3025 to save registers rather than reg->mem moves.
3026 (ix86_expand_prologue): Remove conditional for int_registers_saved
3029 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
3032 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
3033 min/max for never referenced object.
3035 2018-01-24 Jakub Jelinek <jakub@redhat.com>
3038 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
3040 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
3041 attributes from DECL_ATTRIBUTES (decl) without affecting
3042 DECL_ATTRIBUTES (current_function_decl).
3043 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
3044 functions with non-NULL DECL_ABSTRACT_ORIGIN.
3046 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
3048 PR tree-optimization/83979
3049 * fold-const.c (fold_comparison): Use constant_boolean_node
3050 instead of boolean_{true,false}_node.
3052 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
3054 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
3057 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3059 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3060 Simplify the clause that sets the length attribute.
3061 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3062 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
3063 clause that sets the length attribute.
3064 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3066 2018-01-24 Tom de Vries <tom@codesourcery.com>
3069 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
3070 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
3071 Add strict parameter.
3072 (prevent_branch_around_nothing): Insert dummy insn between branch to
3073 label and label with no ptx insn inbetween.
3074 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
3076 2018-01-24 Tom de Vries <tom@codesourcery.com>
3079 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3080 for neutered threads in warp.
3081 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
3083 2018-01-24 Richard Biener <rguenther@suse.de>
3085 PR tree-optimization/83176
3086 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
3089 2018-01-24 Richard Biener <rguenther@suse.de>
3091 PR tree-optimization/82819
3092 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
3093 code generating pluses that are no-ops in the target precision.
3095 2018-01-24 Richard Biener <rguenther@suse.de>
3098 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
3100 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3102 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
3103 to merge probabilities.
3104 * predict.c (probably_never_executed): Also mark as cold functions
3105 with global 0 profile and guessed local profile.
3106 * profile-count.c (profile_probability::combine_with_count): New
3108 * profile-count.h (profile_probability::operator*,
3109 profile_probability::operator*=, profile_probability::operator/,
3110 profile_probability::operator/=): Reduce precision to adjusted
3111 and set value to guessed on contradictory divisions.
3112 (profile_probability::combine_with_freq): Remove.
3113 (profile_probability::combine_wiht_count): Declare.
3114 (profile_count::force_nonzero):: Set to adjusted.
3115 (profile_count::probability_in):: Set quality to adjusted.
3116 * tree-ssa-tail-merge.c (replace_block_by): Use
3119 2018-01-23 Andrew Waterman <andrew@sifive.com>
3120 Jim Wilson <jimw@sifive.com>
3122 * config/riscv/riscv.c (riscv_stack_boundary): New.
3123 (riscv_option_override): Set riscv_stack_boundary. Handle
3124 riscv_preferred_stack_boundary_arg.
3125 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
3126 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
3127 (STACK_BOUNDARY): Set to riscv_stack_boundary.
3128 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
3129 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
3130 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
3132 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
3135 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
3136 of struct ix86_frame.
3137 (ix86_expand_epilogue): Likewise. Add a local variable for
3138 the reg_save_offset field in struct ix86_frame.
3140 2018-01-23 Bin Cheng <bin.cheng@arm.com>
3142 PR tree-optimization/82604
3143 * tree-loop-distribution.c (enum partition_kind): New enum item
3144 PKIND_PARTIAL_MEMSET.
3145 (partition_builtin_p): Support above new enum item.
3146 (generate_code_for_partition): Ditto.
3147 (compute_access_range): Differentiate cases that equality can be
3148 proven at all loops, the innermost loops or no loops.
3149 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
3150 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
3151 (finalize_partitions, distribute_loop): Don't fuse partition of
3152 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
3153 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
3156 2018-01-23 Martin Liska <mliska@suse.cz>
3158 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
3159 order to ignore the predictor.
3160 (PRED_POLYMORPHIC_CALL): Likewise.
3161 (PRED_RECURSIVE_CALL): Likewise.
3163 2018-01-23 Martin Liska <mliska@suse.cz>
3165 * tree-profile.c (tree_profiling): Print function header to
3166 aware reader which function we are working on.
3167 * value-prof.c (gimple_find_values_to_profile): Do not print
3168 not interesting value histograms.
3170 2018-01-23 Martin Liska <mliska@suse.cz>
3172 * profile-count.h (enum profile_quality): Add
3173 profile_uninitialized as the first value. Do not number values
3174 as they are zero based.
3175 (profile_count::verify): Update sanity check.
3176 (profile_probability::verify): Likewise.
3178 2018-01-23 Nathan Sidwell <nathan@acm.org>
3180 * doc/invoke.texi (ffor-scope): Deprecate.
3182 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3184 PR tree-optimization/83510
3185 * domwalk.c (set_all_edges_as_executable): New function.
3186 (dom_walker::dom_walker): Convert bool param
3187 "skip_unreachable_blocks" to enum reachability. Move setup of
3188 edge flags to set_all_edges_as_executable and only do it when
3189 reachability is REACHABLE_BLOCKS.
3190 * domwalk.h (enum dom_walker::reachability): New enum.
3191 (dom_walker::dom_walker): Convert bool param
3192 "skip_unreachable_blocks" to enum reachability.
3193 (set_all_edges_as_executable): New decl.
3194 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
3195 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
3197 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
3198 but converting true to REACHABLE_BLOCKS.
3199 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
3201 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
3202 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
3203 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
3205 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
3206 if check_all_array_refs will be called.
3208 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3210 * tree.c (selftest::test_location_wrappers): Add more test
3213 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3215 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
3216 (selftest::test_bit_in_range): Likewise.
3218 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3221 * doc/sourcebuild.texi (vect_float): Say that the selector
3222 only describes the situation when -funsafe-math-optimizations is on.
3223 (vect_float_strict): Document.
3225 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3227 PR tree-optimization/83965
3228 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
3229 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
3230 instead of checking only for a reduction.
3231 (vect_recog_widen_sum_pattern): Likewise.
3233 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3235 * predict.c (probably_never_executed): Only use precise profile info.
3236 (compute_function_frequency): Skip after inlining hack since we now
3237 have quality checking.
3239 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3241 * profile-count.h (profile_probability::very_unlikely,
3242 profile_probability::unlikely, profile_probability::even): Set
3243 precision to guessed.
3245 2018-01-23 Richard Biener <rguenther@suse.de>
3247 PR tree-optimization/83963
3248 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
3249 Properly terminate dominator walk when crossing the exit edge not
3250 when visiting its source block.
3252 2018-01-23 Jakub Jelinek <jakub@redhat.com>
3255 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
3256 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
3258 2018-01-22 Jakub Jelinek <jakub@redhat.com>
3260 PR tree-optimization/83957
3261 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
3262 semicolon after for body surrounded by braces.
3264 PR tree-optimization/83081
3265 * profile-count.h (profile_probability::split): New method.
3266 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
3267 Use profile_probability::split.
3268 (do_compare_rtx_and_jump): Fix adjustment of probabilities
3269 when splitting a single conditional jump into 2.
3271 2018-01-22 David Malcolm <dmalcolm@redhat.com>
3273 PR tree-optimization/69452
3274 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
3277 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3279 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
3280 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
3281 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
3283 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3285 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
3286 * config/rl78/rl78.md (movdi): New define_expand.
3287 * config/rl78/rl78.c (rl78_split_movdi): New function.
3289 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
3292 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
3294 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
3295 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
3296 128-bit to produce an UNSPEC move to get the double word with the
3297 signbit and then a shift directly to do signbit.
3298 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
3299 implementation with a new version that just does either a direct
3300 move or a regular move. Move memory interface to separate insns.
3301 Move insns so they are next to the expander.
3302 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
3303 with signbit move. Split big and little endian case.
3304 (signbit<mode>2_dm_mem_le): Likewise.
3305 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
3306 (signbit<mode>2_dm2): Likewise.
3308 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3310 * config/rl78/rl78.md (anddi3): New define_expand.
3312 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3314 * config/rl78/rl78.md (umindi3): New define_expand.
3316 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3318 * config/rl78/rl78.md (smindi3): New define_expand.
3320 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3322 * config/rl78/rl78.md (smaxdi3): New define_expand.
3324 2018-01-22 Carl Love <cel@us.ibm.com>
3326 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
3327 LVX_V1TI): Add macro expansion.
3328 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
3329 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
3330 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
3331 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
3332 Change check to determine if the instruction is a byte reversing
3333 entry. Fix typo in comment.
3334 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
3335 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
3336 Add def_builtin calls for new builtins.
3337 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
3338 Add define_insn expansion.
3340 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3342 * config/rl78/rl78.md (umaxdi3): New define_expand.
3344 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3346 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
3347 for non-QImode registers.
3349 2018-01-22 Richard Biener <rguenther@suse.de>
3351 PR tree-optimization/83963
3352 * graphite-scop-detection.c (scop_detection::get_sese): Delay
3353 including the loop exit block.
3354 (scop_detection::merge_sese): Likewise.
3355 (scop_detection::add_scop): Do it here instead.
3357 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3359 * doc/sourcebuild.texi (arm_softfloat): Document.
3361 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
3364 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
3365 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
3366 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
3368 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3369 David Edelsohn <dje.gcc@gmail.com>
3372 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3373 Change "crset eq" to "crset 2".
3374 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3375 (*call_indirect_aix<mode>_nospec): Likewise.
3376 (*call_value_indirect_aix<mode>_nospec): Likewise.
3377 (*call_indirect_elfv2<mode>_nospec): Likewise.
3378 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
3379 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
3380 change assembly output from . to $.
3381 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3382 (indirect_jump<mode>_nospec): Change assembly output from . to $.
3383 (*tablejump<mode>_internal1_nospec): Likewise.
3385 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
3388 * config/sh/sh_optimize_sett_clrt.cc:
3389 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
3391 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
3393 PR tree-optimization/83940
3394 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
3395 offset_dt to vect_constant_def rather than vect_unknown_def_type.
3396 (vect_check_load_store_mask): Add a mask_dt_out parameter and
3397 use it to pass back the definition type.
3398 (vect_check_store_rhs): Likewise rhs_dt_out.
3399 (vect_build_gather_load_calls): Add a mask_dt argument and use
3400 it instead of a call to vect_is_simple_use.
3401 (vectorizable_store): Update calls to vect_check_load_store_mask
3402 and vect_check_store_rhs. Use the dt returned by the latter instead
3403 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
3404 instead of calls to vect_is_simple_use. Pass the scalar rather
3405 than the vector operand to vect_is_simple_use when handling
3406 second and subsequent copies of an rhs value.
3407 (vectorizable_load): Update calls to vect_check_load_store_mask
3408 and vect_build_gather_load_calls. Use the cached mask_dt and
3409 gs_info.offset_dt instead of calls to vect_is_simple_use.
3411 2018-01-20 Jakub Jelinek <jakub@redhat.com>
3414 * tree-emutls.c: Include gimplify.h.
3415 (lower_emutls_2): New function.
3416 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
3417 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
3418 it before further processing.
3421 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
3422 UINTVAL (trueop1) instead of INTVAL (op1).
3424 2018-01-19 Jakub Jelinek <jakub@redhat.com>
3428 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
3429 INCOMING_FRAME_SP_OFFSET if not defined.
3430 (scan_trace): Add ENTRY argument. If true and
3431 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
3432 emit a note to adjust the CFA offset.
3433 (create_cfi_notes): Adjust scan_trace callers.
3434 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
3435 INCOMING_FRAME_SP_OFFSET in the CIE.
3436 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
3437 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
3439 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
3440 * doc/tm.texi: Regenerated.
3442 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3444 PR rtl-optimization/83147
3445 * lra-constraints.c (remove_inheritance_pseudos): Use
3446 lra_substitute_pseudo_within_insn.
3448 2018-01-19 Tom de Vries <tom@codesourcery.com>
3449 Cesar Philippidis <cesar@codesourcery.com>
3452 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
3454 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
3457 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
3458 spaces for function labels.
3460 2018-01-19 Martin Liska <mliska@suse.cz>
3462 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
3463 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
3464 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
3465 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
3466 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
3467 (PRED_CONST_RETURN): Change from 69 to 65.
3468 (PRED_NULL_RETURN): Change from 91 to 71.
3469 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
3470 (PRED_LOOP_GUARD): Change from 66 to 73.
3472 2018-01-19 Martin Liska <mliska@suse.cz>
3474 * predict.c (predict_insn_def): Add new assert.
3475 (struct branch_predictor): Change type to signed integer.
3476 (test_prediction_value_range): Amend test to cover
3478 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
3479 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
3480 (PRED_LOOP_ITERATIONS_MAX): Likewise.
3481 (PRED_LOOP_IV_COMPARE): Likewise.
3482 * predict.h (PROB_UNINITIALIZED): Define new constant.
3484 2018-01-19 Martin Liska <mliska@suse.cz>
3486 * predict.c (dump_prediction): Add new format for
3487 analyze_brprob.py script which is enabled with -details
3489 * profile-count.h (precise_p): New function.
3491 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3493 PR tree-optimization/83922
3494 * tree-vect-loop.c (vect_verify_full_masking): Return false if
3495 there are no statements that need masking.
3496 (vect_active_double_reduction_p): New function.
3497 (vect_analyze_loop_operations): Use it when handling phis that
3498 are not in the loop header.
3500 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3502 PR tree-optimization/83914
3503 * tree-vect-loop.c (vectorizable_induction): Don't convert
3504 init_expr or apply the peeling adjustment for inductions
3505 that are nested within the vectorized loop.
3507 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3509 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
3512 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3516 * function.h (gimplify_parameters): Add gimple_seq * argument.
3517 * function.c: Include gimple.h and options.h.
3518 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
3519 for the added local temporaries if needed.
3520 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
3521 if there are any parameter cleanups, wrap whole body into a
3522 try/finally with the cleanups.
3524 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
3527 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
3528 Use GET_MODE_CLASS for scalar floating point.
3530 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3534 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
3535 Fix call of call_cgraph_insertion_hooks.
3537 2018-01-18 Martin Sebor <msebor@redhat.com>
3539 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
3541 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3544 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
3547 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
3550 * common.opt: (-ffile-prefix-map): New option.
3551 * opts.c (common_handle_option): Defer it.
3552 * opts-global.c (handle_common_deferred_options): Handle it.
3553 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
3554 * file-prefix-map.h: New file.
3555 (remap_debug_filename, add_debug_prefix_map): ...here.
3556 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
3557 * final.c (debug_prefix_map, add_debug_prefix_map
3558 remap_debug_filename): Move to...
3559 * file-prefix-map.c: New file.
3560 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
3561 generalize, get rid of alloca(), use strrchr() instead of strchr().
3562 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
3563 Implement in terms of add_prefix_map().
3564 (remap_macro_filename, remap_debug_filename): Implement in term of
3566 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
3567 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
3568 * dbxout.c: Include file-prefix-map.h.
3569 * varasm.c: Likewise.
3570 * vmsdbgout.c: Likewise.
3571 * xcoffout.c: Likewise.
3572 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
3573 * doc/cppopts.texi (-fmacro-prefix-map): Document.
3574 * doc/invoke.texi (-ffile-prefix-map): Document.
3575 (-fdebug-prefix-map): Update description.
3577 2018-01-18 Martin Liska <mliska@suse.cz>
3579 * config/i386/i386.c (indirect_thunk_name): Document that also
3581 (output_indirect_thunk): Document why both instructions
3582 (pause and lfence) are generated.
3584 2018-01-18 Richard Biener <rguenther@suse.de>
3586 PR tree-optimization/83887
3587 * graphite-scop-detection.c
3588 (scop_detection::get_nearest_dom_with_single_entry): Remove.
3589 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
3590 (scop_detection::merge_sese): Re-implement with a flood-fill
3591 algorithm that properly finds a SESE region if it exists.
3593 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3596 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
3597 pointer_diff optimizations use view_convert instead of convert.
3599 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3601 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3602 Generate different code for -mno-speculate-indirect-jumps.
3603 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3604 (*call_indirect_aix<mode>): Disable for
3605 -mno-speculate-indirect-jumps.
3606 (*call_indirect_aix<mode>_nospec): New define_insn.
3607 (*call_value_indirect_aix<mode>): Disable for
3608 -mno-speculate-indirect-jumps.
3609 (*call_value_indirect_aix<mode>_nospec): New define_insn.
3610 (*sibcall_nonlocal_sysv<mode>): Generate different code for
3611 -mno-speculate-indirect-jumps.
3612 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3614 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3616 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
3617 long double type, set the flags for noting the default long double
3618 type, even if we don't pass or return a long double type.
3620 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
3623 * ipa-inline.c (flatten_function): Do not overwrite final inlining
3626 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
3628 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
3629 support for merge[hl].
3630 (fold_mergehl_helper): New helper function.
3631 (tree-vector-builder.h): New #include for tree_vector_builder usage.
3632 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
3633 (altivec_vmrglw_direct): Add xxmrglw insn.
3635 2018-01-17 Andrew Waterman <andrew@sifive.com>
3637 * config/riscv/riscv.c (riscv_conditional_register_usage): If
3638 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
3640 2018-01-17 David Malcolm <dmalcolm@redhat.com>
3643 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
3644 call the lto_location_cache before reading the
3645 DECL_SOURCE_LOCATION of the types.
3647 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3648 Richard Sandiford <richard.sandiford@linaro.org>
3650 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
3651 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
3652 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
3653 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
3654 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
3656 * config/aarch64/constraints.md (aarch64_movti_operand):
3658 * config/aarch64/predicates.md (Uti): Add new constraint.
3660 2018-01-17 Carl Love <cel@us.ibm.com>
3662 * config/rs6000/vsx.md (define_expand xl_len_r,
3663 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
3664 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
3666 (define_expand, define_insn): Move the shift left from the
3667 define_insn to the define_expand for lxvl and stxvl instructions.
3668 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
3669 and XL_LEN_R definitions to PURE.
3671 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
3673 * config/i386/i386.c (indirect_thunk_name): Declare regno
3674 as unsigned int. Compare regno with INVALID_REGNUM.
3675 (output_indirect_thunk): Ditto.
3676 (output_indirect_thunk_function): Ditto.
3677 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
3678 in the call to output_indirect_thunk_function.
3680 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
3683 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
3684 rather than the size of inner_type to determine the stack slot size
3685 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
3687 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
3690 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
3693 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
3695 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
3696 endian Linux systems to optionally enable multilibs for selecting
3697 the long double type if the user configured an explicit type.
3698 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
3699 have no long double multilibs if not defined.
3700 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
3701 warn if the user used -mabi={ieee,ibm}longdouble and we built
3702 multilibs for long double.
3703 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
3704 appropriate multilib option.
3705 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
3707 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
3708 for building long double multilibs.
3709 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
3711 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
3713 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
3716 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
3718 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
3721 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
3724 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
3727 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3729 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
3730 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
3732 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
3734 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
3735 different rtl trees depending on TARGET_64BIT.
3736 (rs6000_gen_lvx): Likewise.
3738 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3740 * config/visium/visium.md (nop): Tweak comment.
3741 (hazard_nop): Likewise.
3743 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3745 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
3746 -mspeculate-indirect-jumps.
3747 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
3748 for -mno-speculate-indirect-jumps.
3749 (*call_indirect_elfv2<mode>_nospec): New define_insn.
3750 (*call_value_indirect_elfv2<mode>): Disable for
3751 -mno-speculate-indirect-jumps.
3752 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
3753 (indirect_jump): Emit different RTL for
3754 -mno-speculate-indirect-jumps.
3755 (*indirect_jump<mode>): Disable for
3756 -mno-speculate-indirect-jumps.
3757 (*indirect_jump<mode>_nospec): New define_insn.
3758 (tablejump): Emit different RTL for
3759 -mno-speculate-indirect-jumps.
3760 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
3761 (tablejumpsi_nospec): New define_expand.
3762 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
3763 (tablejumpdi_nospec): New define_expand.
3764 (*tablejump<mode>_internal1): Disable for
3765 -mno-speculate-indirect-jumps.
3766 (*tablejump<mode>_internal1_nospec): New define_insn.
3767 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
3770 2018-01-16 Artyom Skrobov tyomitch@gmail.com
3772 * caller-save.c (insert_save): Drop unnecessary parameter. All
3775 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3776 Richard Biener <rguenth@suse.de>
3779 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
3780 return early, inline manually is_gimple_sizepos. Make sure if we
3781 call gimplify_expr we don't end up with a gimple constant.
3782 * tree.c (variably_modified_type_p): Don't return true for
3783 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
3784 * gimplify.h (is_gimple_sizepos): Remove.
3786 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3788 PR tree-optimization/83857
3789 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
3790 vectorizable_live_operation for pure SLP statements.
3791 (vectorizable_live_operation): Handle PHIs.
3793 2018-01-16 Richard Biener <rguenther@suse.de>
3795 PR tree-optimization/83867
3796 * tree-vect-stmts.c (vect_transform_stmt): Precompute
3797 nested_in_vect_loop_p since the scalar stmt may get invalidated.
3799 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3802 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
3803 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
3804 If off is not INTEGER_CST, issue a may not be aligned warning
3805 rather than isn't aligned. Use isn%'t rather than isn't.
3806 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
3808 <case MULT_EXPR>: Improve the case when bottom and one of the
3809 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
3810 operand, in that case check if the other operand is multiple of
3811 bottom divided by the INTEGER_CST operand.
3813 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3816 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
3817 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
3818 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
3819 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
3820 * config/pa/pa.c (pa_function_arg_advance): Likewise.
3821 (pa_function_arg, pa_arg_partial_bytes): Likewise.
3822 (pa_function_arg_size): New function.
3824 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3826 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
3827 in a separate statement.
3829 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3831 PR tree-optimization/83847
3832 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
3833 group gathers and scatters.
3835 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3837 PR rtl-optimization/86620
3838 * params.def (max-sched-ready-insns): Bump minimum value to 1.
3840 PR rtl-optimization/83213
3841 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
3842 to last if both are JUMP_INSNs.
3844 PR tree-optimization/83843
3845 * gimple-ssa-store-merging.c
3846 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
3847 store_immediate_info for bswap/nop orig_stores.
3849 2018-01-15 Andrew Waterman <andrew@sifive.com>
3851 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
3853 <UDIV>: Increase cost if !TARGET_DIV.
3855 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
3857 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
3858 (define_attr "cr_logical_3op"): New.
3859 (cceq_ior_compare): Adjust.
3860 (cceq_ior_compare_complement): Adjust.
3861 (*cceq_rev_compare): Adjust.
3862 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
3863 (is_cracked_insn): Adjust.
3864 (insn_must_be_first_in_group): Adjust.
3865 * config/rs6000/40x.md: Adjust.
3866 * config/rs6000/440.md: Adjust.
3867 * config/rs6000/476.md: Adjust.
3868 * config/rs6000/601.md: Adjust.
3869 * config/rs6000/603.md: Adjust.
3870 * config/rs6000/6xx.md: Adjust.
3871 * config/rs6000/7450.md: Adjust.
3872 * config/rs6000/7xx.md: Adjust.
3873 * config/rs6000/8540.md: Adjust.
3874 * config/rs6000/cell.md: Adjust.
3875 * config/rs6000/e300c2c3.md: Adjust.
3876 * config/rs6000/e500mc.md: Adjust.
3877 * config/rs6000/e500mc64.md: Adjust.
3878 * config/rs6000/e5500.md: Adjust.
3879 * config/rs6000/e6500.md: Adjust.
3880 * config/rs6000/mpc.md: Adjust.
3881 * config/rs6000/power4.md: Adjust.
3882 * config/rs6000/power5.md: Adjust.
3883 * config/rs6000/power6.md: Adjust.
3884 * config/rs6000/power7.md: Adjust.
3885 * config/rs6000/power8.md: Adjust.
3886 * config/rs6000/power9.md: Adjust.
3887 * config/rs6000/rs64.md: Adjust.
3888 * config/rs6000/titan.md: Adjust.
3890 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3892 * config/i386/predicates.md (indirect_branch_operand): Rewrite
3893 ix86_indirect_branch_register logic.
3895 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3897 * config/i386/constraints.md (Bs): Update
3898 ix86_indirect_branch_register check. Don't check
3899 ix86_indirect_branch_register with GOT_memory_operand.
3901 * config/i386/predicates.md (GOT_memory_operand): Don't check
3902 ix86_indirect_branch_register here.
3903 (GOT32_symbol_operand): Likewise.
3905 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3907 * config/i386/predicates.md (constant_call_address_operand):
3908 Rewrite ix86_indirect_branch_register logic.
3909 (sibcall_insn_operand): Likewise.
3911 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3913 * config/i386/constraints.md (Bs): Replace
3914 ix86_indirect_branch_thunk_register with
3915 ix86_indirect_branch_register.
3917 * config/i386/i386.md (indirect_jump): Likewise.
3918 (tablejump): Likewise.
3919 (*sibcall_memory): Likewise.
3920 (*sibcall_value_memory): Likewise.
3921 Peepholes of indirect call and jump via memory: Likewise.
3922 * config/i386/i386.opt: Likewise.
3923 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3924 (GOT_memory_operand): Likewise.
3925 (call_insn_operand): Likewise.
3926 (sibcall_insn_operand): Likewise.
3927 (GOT32_symbol_operand): Likewise.
3929 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3932 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
3933 type rather than type addr's type points to.
3934 (expand_omp_atomic_mutex): Likewise.
3935 (expand_omp_atomic): Likewise.
3937 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3940 * config/i386/i386.c (output_indirect_thunk_function): Use
3941 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
3942 for __x86_return_thunk.
3944 2018-01-15 Richard Biener <rguenther@suse.de>
3947 * expmed.c (extract_bit_field_1): Fix typo.
3949 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3952 * config/arm/iterators.md (VF): New mode iterator.
3953 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
3954 Remove integer-related logic from pattern.
3955 (neon_vabd<mode>_3): Likewise.
3957 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3960 * common.opt (fstrict-overflow): No longer an alias.
3961 (fwrapv-pointer): New option.
3962 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
3963 also for pointer types based on flag_wrapv_pointer.
3964 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
3965 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
3966 opts->x_flag_wrapv got set.
3967 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
3968 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
3969 POINTER_TYPE_OVERFLOW_UNDEFINED.
3970 * match.pd: Likewise in address comparison pattern.
3971 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
3973 2018-01-15 Richard Biener <rguenther@suse.de>
3976 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
3977 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
3978 Reset type names to their identifier if their TYPE_DECL doesn't
3979 have linkage (and thus is used for ODR and devirt).
3980 (save_debug_info_for_decl): Remove.
3981 (save_debug_info_for_type): Likewise.
3982 (add_tree_to_fld_list): Adjust.
3983 * tree-pretty-print.c (dump_generic_node): Make dumping of
3984 type names more robust.
3986 2018-01-15 Richard Biener <rguenther@suse.de>
3988 * BASE-VER: Bump to 8.0.1.
3990 2018-01-14 Martin Sebor <msebor@redhat.com>
3993 * builtins.c (check_access): Avoid warning when the no-warning bit
3996 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
3998 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
3999 * ira-color (allocno_hard_regs_compare): Likewise.
4001 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
4004 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
4005 Use .pushsection/.popsection.
4007 2018-01-14 Martin Sebor <msebor@redhat.com>
4010 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
4012 2018-01-14 Jakub Jelinek <jakub@redhat.com>
4014 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
4015 entry from extra_headers.
4016 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
4017 extra_headers, make the list bitwise identical to the i?86-*-* one.
4019 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4021 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
4022 -mcmodel=large with -mindirect-branch=thunk,
4023 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
4024 -mfunction-return=thunk-extern.
4025 * doc/invoke.texi: Document -mcmodel=large is incompatible with
4026 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
4027 -mfunction-return=thunk and -mfunction-return=thunk-extern.
4029 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4031 * config/i386/i386.c (print_reg): Print the name of the full
4032 integer register without '%'.
4033 (ix86_print_operand): Handle 'V'.
4034 * doc/extend.texi: Document 'V' modifier.
4036 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4038 * config/i386/constraints.md (Bs): Disallow memory operand for
4039 -mindirect-branch-register.
4041 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4042 (GOT_memory_operand): Likewise.
4043 (call_insn_operand): Likewise.
4044 (sibcall_insn_operand): Likewise.
4045 (GOT32_symbol_operand): Likewise.
4046 * config/i386/i386.md (indirect_jump): Call convert_memory_address
4047 for -mindirect-branch-register.
4048 (tablejump): Likewise.
4049 (*sibcall_memory): Likewise.
4050 (*sibcall_value_memory): Likewise.
4051 Disallow peepholes of indirect call and jump via memory for
4052 -mindirect-branch-register.
4053 (*call_pop): Replace m with Bw.
4054 (*call_value_pop): Likewise.
4055 (*sibcall_pop_memory): Replace m with Bs.
4056 * config/i386/i386.opt (mindirect-branch-register): New option.
4057 * doc/invoke.texi: Document -mindirect-branch-register option.
4059 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4061 * config/i386/i386-protos.h (ix86_output_function_return): New.
4062 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
4063 set function_return_type.
4064 (indirect_thunk_name): Add ret_p to indicate thunk for function
4066 (output_indirect_thunk_function): Pass false to
4067 indirect_thunk_name.
4068 (ix86_output_indirect_branch_via_reg): Likewise.
4069 (ix86_output_indirect_branch_via_push): Likewise.
4070 (output_indirect_thunk_function): Create alias for function
4071 return thunk if regno < 0.
4072 (ix86_output_function_return): New function.
4073 (ix86_handle_fndecl_attribute): Handle function_return.
4074 (ix86_attribute_table): Add function_return.
4075 * config/i386/i386.h (machine_function): Add
4076 function_return_type.
4077 * config/i386/i386.md (simple_return_internal): Use
4078 ix86_output_function_return.
4079 (simple_return_internal_long): Likewise.
4080 * config/i386/i386.opt (mfunction-return=): New option.
4081 (indirect_branch): Mention -mfunction-return=.
4082 * doc/extend.texi: Document function_return function attribute.
4083 * doc/invoke.texi: Document -mfunction-return= option.
4085 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4087 * config/i386/i386-opts.h (indirect_branch): New.
4088 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
4089 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
4090 with local indirect jump when converting indirect call and jump.
4091 (ix86_set_indirect_branch_type): New.
4092 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
4093 (indirectlabelno): New.
4094 (indirect_thunk_needed): Likewise.
4095 (indirect_thunk_bnd_needed): Likewise.
4096 (indirect_thunks_used): Likewise.
4097 (indirect_thunks_bnd_used): Likewise.
4098 (INDIRECT_LABEL): Likewise.
4099 (indirect_thunk_name): Likewise.
4100 (output_indirect_thunk): Likewise.
4101 (output_indirect_thunk_function): Likewise.
4102 (ix86_output_indirect_branch_via_reg): Likewise.
4103 (ix86_output_indirect_branch_via_push): Likewise.
4104 (ix86_output_indirect_branch): Likewise.
4105 (ix86_output_indirect_jmp): Likewise.
4106 (ix86_code_end): Call output_indirect_thunk_function if needed.
4107 (ix86_output_call_insn): Call ix86_output_indirect_branch if
4109 (ix86_handle_fndecl_attribute): Handle indirect_branch.
4110 (ix86_attribute_table): Add indirect_branch.
4111 * config/i386/i386.h (machine_function): Add indirect_branch_type
4112 and has_local_indirect_jump.
4113 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
4115 (tablejump): Likewise.
4116 (*indirect_jump): Use ix86_output_indirect_jmp.
4117 (*tablejump_1): Likewise.
4118 (simple_return_indirect_internal): Likewise.
4119 * config/i386/i386.opt (mindirect-branch=): New option.
4120 (indirect_branch): New.
4123 (thunk-inline): Likewise.
4124 (thunk-extern): Likewise.
4125 * doc/extend.texi: Document indirect_branch function attribute.
4126 * doc/invoke.texi: Document -mindirect-branch= option.
4128 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
4131 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
4133 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
4135 * ipa-inline.c (want_inline_small_function_p): Return false if
4136 inlining has already failed with CIF_FINAL_ERROR.
4137 (update_caller_keys): Call want_inline_small_function_p before
4139 (update_callee_keys): Likewise.
4141 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4143 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
4145 (rs6000_quadword_masked_address_p): Likewise.
4146 (quad_aligned_load_p): Likewise.
4147 (quad_aligned_store_p): Likewise.
4148 (const_load_sequence_p): Add comment to describe the outer-most loop.
4149 (mimic_memory_attributes_and_flags): New function.
4150 (rs6000_gen_stvx): Likewise.
4151 (replace_swapped_aligned_store): Likewise.
4152 (rs6000_gen_lvx): Likewise.
4153 (replace_swapped_aligned_load): Likewise.
4154 (replace_swapped_load_constant): Capitalize argument name in
4155 comment describing this function.
4156 (rs6000_analyze_swaps): Add a third pass to search for vector loads
4157 and stores that access quad-word aligned addresses and replace
4158 with stvx or lvx instructions when appropriate.
4159 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
4160 New function prototype.
4161 (rs6000_quadword_masked_address_p): Likewise.
4162 (rs6000_gen_lvx): Likewise.
4163 (rs6000_gen_stvx): Likewise.
4164 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
4165 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
4166 when memory address is aligned.
4167 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
4168 this split to select lvx instruction when memory address is aligned.
4169 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
4170 instruction when memory address is aligned.
4171 (*vsx_le_perm_load_v16qi): Likewise.
4172 (four unnamed splitters): Modify to select the stvx instruction
4173 when memory is aligned.
4175 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
4177 * predict.c (determine_unlikely_bbs): Handle correctly BBs
4178 which appears in the queue multiple times.
4180 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4181 Alan Hayward <alan.hayward@arm.com>
4182 David Sherwood <david.sherwood@arm.com>
4184 * tree-vectorizer.h (vec_lower_bound): New structure.
4185 (_loop_vec_info): Add check_nonzero and lower_bounds.
4186 (LOOP_VINFO_CHECK_NONZERO): New macro.
4187 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
4188 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
4189 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
4190 fields. Make seg_len the distance travelled, not including the
4192 (dr_direction_indicator): Declare.
4193 (dr_zero_step_indicator): Likewise.
4194 (dr_known_forward_stride_p): Likewise.
4195 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
4197 (runtime_alias_check_p): Allow runtime alias checks with
4199 (operator ==): Compare access_size and align.
4200 (prune_runtime_alias_test_list): Rework for new distinction between
4201 the access_size and seg_len.
4202 (create_intersect_range_checks_index): Likewise. Cope with polynomial
4204 (get_segment_min_max): New function.
4205 (create_intersect_range_checks): Use it.
4206 (dr_step_indicator): New function.
4207 (dr_direction_indicator): Likewise.
4208 (dr_zero_step_indicator): Likewise.
4209 (dr_known_forward_stride_p): Likewise.
4210 * tree-loop-distribution.c (data_ref_segment_size): Return
4211 DR_STEP * (niters - 1).
4212 (compute_alias_check_pairs): Update call to the dr_with_seg_len
4214 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
4215 (vect_preserves_scalar_order_p): New function, split out from...
4216 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
4217 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
4218 (vect_vfa_access_size): New function.
4219 (vect_vfa_align): Likewise.
4220 (vect_compile_time_alias): Take access_size_a and access_b arguments.
4221 (dump_lower_bound): New function.
4222 (vect_check_lower_bound): Likewise.
4223 (vect_small_gap_p): Likewise.
4224 (vectorizable_with_step_bound_p): Likewise.
4225 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
4226 depencies if the vectorization factor is 1. Convert the checks
4227 for nonzero steps into checks on the bounds of DR_STEP. Try using
4228 a bunds check for variable steps if the minimum required step is
4229 relatively small. Update calls to the dr_with_seg_len
4230 constructor and to vect_compile_time_alias.
4231 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
4233 (vect_loop_versioning): Call it.
4234 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
4236 (vect_estimate_min_profitable_iters): Account for any bounds checks.
4238 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4239 Alan Hayward <alan.hayward@arm.com>
4240 David Sherwood <david.sherwood@arm.com>
4242 * doc/sourcebuild.texi (vect_scatter_store): Document.
4243 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
4245 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
4247 * genopinit.c (main): Add supports_vec_scatter_store and
4248 supports_vec_scatter_store_cached to target_optabs.
4249 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
4250 IFN_MASK_SCATTER_STORE.
4251 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
4253 * internal-fn.h (internal_store_fn_p): Declare.
4254 (internal_fn_stored_value_index): Likewise.
4255 * internal-fn.c (scatter_store_direct): New macro.
4256 (expand_scatter_store_optab_fn): New function.
4257 (direct_scatter_store_optab_supported_p): New macro.
4258 (internal_store_fn_p): New function.
4259 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
4260 IFN_MASK_SCATTER_STORE.
4261 (internal_fn_mask_index): Likewise.
4262 (internal_fn_stored_value_index): New function.
4263 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
4265 * optabs-query.h (supports_vec_scatter_store_p): Declare.
4266 * optabs-query.c (supports_vec_scatter_store_p): New function.
4267 * tree-vectorizer.h (vect_get_store_rhs): Declare.
4268 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
4269 true for scatter stores.
4270 (vect_gather_scatter_fn_p): Handle scatter stores too.
4271 (vect_check_gather_scatter): Consider using scatter stores if
4272 supports_vec_scatter_store_p.
4273 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
4275 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4276 internal_fn_stored_value_index.
4277 (check_load_store_masking): Handle scatter stores too.
4278 (vect_get_store_rhs): Make public.
4279 (vectorizable_call): Use internal_store_fn_p.
4280 (vectorizable_store): Handle scatter store internal functions.
4281 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
4282 when deciding whether the end of the group has been reached.
4283 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
4284 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
4285 (mask_scatter_store<mode>): New insns.
4287 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4288 Alan Hayward <alan.hayward@arm.com>
4289 David Sherwood <david.sherwood@arm.com>
4291 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
4292 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
4293 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
4295 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
4296 Use vect_truncate_gather_scatter_offset if we can't treat the
4297 operation as a normal gather load or scatter store.
4298 (get_group_load_store_type): Take the gather_scatter_info
4299 as argument. Try using a gather load or scatter store for
4300 single-element groups.
4301 (get_load_store_type): Update calls to get_group_load_store_type
4302 and vect_use_strided_gather_scatters_p.
4304 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4305 Alan Hayward <alan.hayward@arm.com>
4306 David Sherwood <david.sherwood@arm.com>
4308 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
4309 optional tree argument.
4310 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
4312 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
4313 but continue to use the current value as a fallback.
4314 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
4315 to compare the updates.
4316 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
4317 (get_load_store_type): Use it when handling a strided access.
4318 (vect_get_strided_load_store_ops): New function.
4319 (vect_get_data_ptr_increment): Likewise.
4320 (vectorizable_load): Handle strided gather loads. Always pass
4321 a step to vect_create_data_ref_ptr and bump_vector_ptr.
4323 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4324 Alan Hayward <alan.hayward@arm.com>
4325 David Sherwood <david.sherwood@arm.com>
4327 * doc/md.texi (gather_load@var{m}): Document.
4328 (mask_gather_load@var{m}): Likewise.
4329 * genopinit.c (main): Add supports_vec_gather_load and
4330 supports_vec_gather_load_cached to target_optabs.
4331 * optabs-tree.c (init_tree_optimization_optabs): Use
4332 ggc_cleared_alloc to allocate target_optabs.
4333 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
4334 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
4336 * internal-fn.h (internal_load_fn_p): Declare.
4337 (internal_gather_scatter_fn_p): Likewise.
4338 (internal_fn_mask_index): Likewise.
4339 (internal_gather_scatter_fn_supported_p): Likewise.
4340 * internal-fn.c (gather_load_direct): New macro.
4341 (expand_gather_load_optab_fn): New function.
4342 (direct_gather_load_optab_supported_p): New macro.
4343 (direct_internal_fn_optab): New function.
4344 (internal_load_fn_p): Likewise.
4345 (internal_gather_scatter_fn_p): Likewise.
4346 (internal_fn_mask_index): Likewise.
4347 (internal_gather_scatter_fn_supported_p): Likewise.
4348 * optabs-query.c (supports_at_least_one_mode_p): New function.
4349 (supports_vec_gather_load_p): Likewise.
4350 * optabs-query.h (supports_vec_gather_load_p): Declare.
4351 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
4352 and memory_type field.
4353 (NUM_PATTERNS): Bump to 15.
4354 * tree-vect-data-refs.c: Include internal-fn.h.
4355 (vect_gather_scatter_fn_p): New function.
4356 (vect_describe_gather_scatter_call): Likewise.
4357 (vect_check_gather_scatter): Try using internal functions for
4358 gather loads. Recognize existing calls to a gather load function.
4359 (vect_analyze_data_refs): Consider using gather loads if
4360 supports_vec_gather_load_p.
4361 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
4362 (vect_get_gather_scatter_offset_type): Likewise.
4363 (vect_convert_mask_for_vectype): Likewise.
4364 (vect_add_conversion_to_patterm): Likewise.
4365 (vect_try_gather_scatter_pattern): Likewise.
4366 (vect_recog_gather_scatter_pattern): New pattern recognizer.
4367 (vect_vect_recog_func_ptrs): Add it.
4368 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4369 internal_fn_mask_index and internal_gather_scatter_fn_p.
4370 (check_load_store_masking): Take the gather_scatter_info as an
4371 argument and handle gather loads.
4372 (vect_get_gather_scatter_ops): New function.
4373 (vectorizable_call): Check internal_load_fn_p.
4374 (vectorizable_load): Likewise. Handle gather load internal
4376 (vectorizable_store): Update call to check_load_store_masking.
4377 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
4378 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
4379 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
4380 (aarch64_gather_scale_operand_d): New predicates.
4381 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
4382 (mask_gather_load<mode>): New insns.
4384 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4385 Alan Hayward <alan.hayward@arm.com>
4386 David Sherwood <david.sherwood@arm.com>
4388 * optabs.def (fold_left_plus_optab): New optab.
4389 * doc/md.texi (fold_left_plus_@var{m}): Document.
4390 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
4391 * internal-fn.c (fold_left_direct): Define.
4392 (expand_fold_left_optab_fn): Likewise.
4393 (direct_fold_left_optab_supported_p): Likewise.
4394 * fold-const-call.c (fold_const_fold_left): New function.
4395 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
4396 * tree-parloops.c (valid_reduction_p): New function.
4397 (gather_scalar_reductions): Use it.
4398 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
4399 (vect_finish_replace_stmt): Declare.
4400 * tree-vect-loop.c (fold_left_reduction_fn): New function.
4401 (needs_fold_left_reduction_p): New function, split out from...
4402 (vect_is_simple_reduction): ...here. Accept reductions that
4403 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
4404 (vect_force_simple_reduction): Also store the reduction type in
4405 the assignment's STMT_VINFO_REDUC_TYPE.
4406 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
4407 (merge_with_identity): New function.
4408 (vect_expand_fold_left): Likewise.
4409 (vectorize_fold_left_reduction): Likewise.
4410 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
4411 scalar phi in place for it. Check for target support and reject
4412 cases that would reassociate the operation. Defer the transform
4413 phase to vectorize_fold_left_reduction.
4414 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
4415 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
4416 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
4418 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4420 * tree-if-conv.c (predicate_mem_writes): Remove redundant
4421 call to ifc_temp_var.
4423 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4424 Alan Hayward <alan.hayward@arm.com>
4425 David Sherwood <david.sherwood@arm.com>
4427 * target.def (legitimize_address_displacement): Take the original
4428 offset as a poly_int.
4429 * targhooks.h (default_legitimize_address_displacement): Update
4431 * targhooks.c (default_legitimize_address_displacement): Likewise.
4432 * doc/tm.texi: Regenerate.
4433 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
4434 as an argument, moving assert of ad->disp == ad->disp_term to...
4435 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
4436 Try calling targetm.legitimize_address_displacement before expanding
4437 the address rather than afterwards, and adjust for the new interface.
4438 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
4439 Match the new hook interface. Handle SVE addresses.
4440 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
4443 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4445 * Makefile.in (OBJS): Add early-remat.o.
4446 * target.def (select_early_remat_modes): New hook.
4447 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
4448 * doc/tm.texi: Regenerate.
4449 * targhooks.h (default_select_early_remat_modes): Declare.
4450 * targhooks.c (default_select_early_remat_modes): New function.
4451 * timevar.def (TV_EARLY_REMAT): New timevar.
4452 * passes.def (pass_early_remat): New pass.
4453 * tree-pass.h (make_pass_early_remat): Declare.
4454 * early-remat.c: New file.
4455 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
4457 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
4459 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4460 Alan Hayward <alan.hayward@arm.com>
4461 David Sherwood <david.sherwood@arm.com>
4463 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
4464 vfm1 with a bound_epilog parameter.
4465 (vect_do_peeling): Update calls accordingly, and move the prologue
4466 call earlier in the function. Treat the base bound_epilog as 0 for
4467 fully-masked loops and retain vf - 1 for other loops. Add 1 to
4468 this base when peeling for gaps.
4469 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
4470 with fully-masked loops.
4471 (vect_estimate_min_profitable_iters): Handle the single peeled
4472 iteration in that case.
4474 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4475 Alan Hayward <alan.hayward@arm.com>
4476 David Sherwood <david.sherwood@arm.com>
4478 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
4479 single-element interleaving even if the size is not a power of 2.
4480 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
4481 accesses for single-element interleaving if the group size is
4484 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4485 Alan Hayward <alan.hayward@arm.com>
4486 David Sherwood <david.sherwood@arm.com>
4488 * doc/md.texi (fold_extract_last_@var{m}): Document.
4489 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
4490 * optabs.def (fold_extract_last_optab): New optab.
4491 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
4492 * internal-fn.c (fold_extract_direct): New macro.
4493 (expand_fold_extract_optab_fn): Likewise.
4494 (direct_fold_extract_optab_supported_p): Likewise.
4495 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
4496 * tree-vect-loop.c (vect_model_reduction_cost): Handle
4497 EXTRACT_LAST_REDUCTION.
4498 (get_initial_def_for_reduction): Do not create an initial vector
4499 for EXTRACT_LAST_REDUCTION reductions.
4500 (vectorizable_reduction): Leave the scalar phi in place for
4501 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
4502 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
4503 epilogue code for EXTRACT_LAST_REDUCTION and defer the
4504 transform phase to vectorizable_condition.
4505 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
4507 (vect_finish_stmt_generation): ...here.
4508 (vect_finish_replace_stmt): New function.
4509 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
4510 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
4512 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
4514 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4515 Alan Hayward <alan.hayward@arm.com>
4516 David Sherwood <david.sherwood@arm.com>
4518 * doc/md.texi (extract_last_@var{m}): Document.
4519 * optabs.def (extract_last_optab): New optab.
4520 * internal-fn.def (EXTRACT_LAST): New internal function.
4521 * internal-fn.c (cond_unary_direct): New macro.
4522 (expand_cond_unary_optab_fn): Likewise.
4523 (direct_cond_unary_optab_supported_p): Likewise.
4524 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
4525 loops using EXTRACT_LAST.
4526 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
4527 (extract_last_<mode>): ...this optab.
4528 (vec_extract<mode><Vel>): Update accordingly.
4530 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4531 Alan Hayward <alan.hayward@arm.com>
4532 David Sherwood <david.sherwood@arm.com>
4534 * target.def (empty_mask_is_expensive): New hook.
4535 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
4536 * doc/tm.texi: Regenerate.
4537 * targhooks.h (default_empty_mask_is_expensive): Declare.
4538 * targhooks.c (default_empty_mask_is_expensive): New function.
4539 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
4540 if the target says that empty masks are expensive.
4541 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
4543 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
4545 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4546 Alan Hayward <alan.hayward@arm.com>
4547 David Sherwood <david.sherwood@arm.com>
4549 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
4550 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
4551 (vect_use_loop_mask_for_alignment_p): New function.
4552 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
4553 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
4554 niters_skip argument. Make sure that the first niters_skip elements
4555 of the first iteration are inactive.
4556 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
4557 Update call to vect_set_loop_masks_directly.
4558 (get_misalign_in_elems): New function, split out from...
4559 (vect_gen_prolog_loop_niters): ...here.
4560 (vect_update_init_of_dr): Take a code argument that specifies whether
4561 the adjustment should be added or subtracted.
4562 (vect_update_init_of_drs): Likewise.
4563 (vect_prepare_for_masked_peels): New function.
4564 (vect_do_peeling): Skip prologue peeling if we're using a mask
4565 instead. Update call to vect_update_inits_of_drs.
4566 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4568 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
4569 alignment. Do not include the number of peeled iterations in
4570 the minimum threshold in that case.
4571 (vectorizable_induction): Adjust the start value down by
4572 LOOP_VINFO_MASK_SKIP_NITERS iterations.
4573 (vect_transform_loop): Call vect_prepare_for_masked_peels.
4574 Take the number of skipped iterations into account when calculating
4576 * tree-vect-stmts.c (vect_gen_while_not): New function.
4578 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4579 Alan Hayward <alan.hayward@arm.com>
4580 David Sherwood <david.sherwood@arm.com>
4582 * doc/sourcebuild.texi (vect_fully_masked): Document.
4583 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
4585 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
4587 (vect_analyze_loop_2): ...here. Don't check the vectorization
4588 factor against the number of loop iterations if the loop is
4591 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4592 Alan Hayward <alan.hayward@arm.com>
4593 David Sherwood <david.sherwood@arm.com>
4595 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
4596 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
4597 (dump_groups): Update accordingly.
4598 (iv_use::mem_type): New member variable.
4599 (address_p): New function.
4600 (record_use): Add a mem_type argument and initialize the new
4602 (record_group_use): Add a mem_type argument. Use address_p.
4603 Remove obsolete null checks of base_object. Update call to record_use.
4604 (find_interesting_uses_op): Update call to record_group_use.
4605 (find_interesting_uses_cond): Likewise.
4606 (find_interesting_uses_address): Likewise.
4607 (get_mem_type_for_internal_fn): New function.
4608 (find_address_like_use): Likewise.
4609 (find_interesting_uses_stmt): Try find_address_like_use before
4610 calling find_interesting_uses_op.
4611 (addr_offset_valid_p): Use the iv mem_type field as the type
4612 of the addressed memory.
4613 (add_autoinc_candidates): Likewise.
4614 (get_address_cost): Likewise.
4615 (split_small_address_groups_p): Use address_p.
4616 (split_address_groups): Likewise.
4617 (add_iv_candidate_for_use): Likewise.
4618 (autoinc_possible_for_pair): Likewise.
4619 (rewrite_groups): Likewise.
4620 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
4621 (determine_group_iv_cost): Update after split of USE_ADDRESS.
4622 (get_alias_ptr_type_for_ptr_address): New function.
4623 (rewrite_use_address): Rewrite address uses in calls that were
4624 identified by find_address_like_use.
4626 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4627 Alan Hayward <alan.hayward@arm.com>
4628 David Sherwood <david.sherwood@arm.com>
4630 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
4632 * gimple-expr.h (is_gimple_addressable: Likewise.
4633 * gimple-expr.c (is_gimple_address): Likewise.
4634 * internal-fn.c (expand_call_mem_ref): New function.
4635 (expand_mask_load_optab_fn): Use it.
4636 (expand_mask_store_optab_fn): Likewise.
4638 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4639 Alan Hayward <alan.hayward@arm.com>
4640 David Sherwood <david.sherwood@arm.com>
4642 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
4643 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
4644 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
4645 (cond_umax@var{mode}): Document.
4646 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
4647 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
4648 (cond_umin_optab, cond_umax_optab): New optabs.
4649 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
4650 (COND_IOR, COND_XOR): New internal functions.
4651 * internal-fn.h (get_conditional_internal_fn): Declare.
4652 * internal-fn.c (cond_binary_direct): New macro.
4653 (expand_cond_binary_optab_fn): Likewise.
4654 (direct_cond_binary_optab_supported_p): Likewise.
4655 (get_conditional_internal_fn): New function.
4656 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
4657 Cope with reduction statements that are vectorized as calls rather
4659 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
4660 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
4661 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
4662 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4663 (UNSPEC_COND_EOR): New unspecs.
4664 (optab): Add mappings for them.
4665 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
4666 (sve_int_op, sve_fp_op): New int attributes.
4668 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4669 Alan Hayward <alan.hayward@arm.com>
4670 David Sherwood <david.sherwood@arm.com>
4672 * optabs.def (while_ult_optab): New optab.
4673 * doc/md.texi (while_ult@var{m}@var{n}): Document.
4674 * internal-fn.def (WHILE_ULT): New internal function.
4675 * internal-fn.h (direct_internal_fn_supported_p): New override
4676 that takes two types as argument.
4677 * internal-fn.c (while_direct): New macro.
4678 (expand_while_optab_fn): New function.
4679 (convert_optab_supported_p): Likewise.
4680 (direct_while_optab_supported_p): New macro.
4681 * wide-int.h (wi::udiv_ceil): New function.
4682 * tree-vectorizer.h (rgroup_masks): New structure.
4683 (vec_loop_masks): New typedef.
4684 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
4686 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
4687 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
4688 (vect_max_vf): New function.
4689 (slpeel_make_loop_iterate_ntimes): Delete.
4690 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
4691 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
4692 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
4693 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
4694 internal-fn.h, stor-layout.h and optabs-query.h.
4695 (vect_set_loop_mask): New function.
4696 (add_preheader_seq): Likewise.
4697 (add_header_seq): Likewise.
4698 (interleave_supported_p): Likewise.
4699 (vect_maybe_permute_loop_masks): Likewise.
4700 (vect_set_loop_masks_directly): Likewise.
4701 (vect_set_loop_condition_masked): Likewise.
4702 (vect_set_loop_condition_unmasked): New function, split out from
4703 slpeel_make_loop_iterate_ntimes.
4704 (slpeel_make_loop_iterate_ntimes): Rename to..
4705 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
4706 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
4707 (vect_do_peeling): Update call accordingly.
4708 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
4710 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4711 mask_compare_type, can_fully_mask_p and fully_masked_p.
4712 (release_vec_loop_masks): New function.
4713 (_loop_vec_info): Use it to free the loop masks.
4714 (can_produce_all_loop_masks_p): New function.
4715 (vect_get_max_nscalars_per_iter): Likewise.
4716 (vect_verify_full_masking): Likewise.
4717 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
4718 retries, and free the mask rgroups before retrying. Check loop-wide
4719 reasons for disallowing fully-masked loops. Make the final decision
4720 about whether use a fully-masked loop or not.
4721 (vect_estimate_min_profitable_iters): Do not assume that peeling
4722 for the number of iterations will be needed for fully-masked loops.
4723 (vectorizable_reduction): Disable fully-masked loops.
4724 (vectorizable_live_operation): Likewise.
4725 (vect_halve_mask_nunits): New function.
4726 (vect_double_mask_nunits): Likewise.
4727 (vect_record_loop_mask): Likewise.
4728 (vect_get_loop_mask): Likewise.
4729 (vect_transform_loop): Handle the case in which the final loop
4730 iteration might handle a partial vector. Call vect_set_loop_condition
4731 instead of slpeel_make_loop_iterate_ntimes.
4732 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
4733 (check_load_store_masking): New function.
4734 (prepare_load_store_mask): Likewise.
4735 (vectorizable_store): Handle fully-masked loops.
4736 (vectorizable_load): Likewise.
4737 (supportable_widening_operation): Use vect_halve_mask_nunits for
4739 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
4740 (vect_gen_while): New function.
4741 * config/aarch64/aarch64.md (umax<mode>3): New expander.
4742 (aarch64_uqdec<mode>): New insn.
4744 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4745 Alan Hayward <alan.hayward@arm.com>
4746 David Sherwood <david.sherwood@arm.com>
4748 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
4749 (reduc_xor_scal_optab): New optabs.
4750 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
4751 (reduc_xor_scal_@var{m}): Document.
4752 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
4753 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
4755 * fold-const-call.c (fold_const_call): Handle them.
4756 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
4757 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
4758 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
4759 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
4760 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
4761 (UNSPEC_XORV): New unspecs.
4762 (optab): Add entries for them.
4763 (BITWISEV): New int iterator.
4764 (bit_reduc_op): New int attributes.
4766 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4767 Alan Hayward <alan.hayward@arm.com>
4768 David Sherwood <david.sherwood@arm.com>
4770 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
4771 * internal-fn.def (VEC_SHL_INSERT): New internal function.
4772 * optabs.def (vec_shl_insert_optab): New optab.
4773 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
4774 (duplicate_and_interleave): Likewise.
4775 * tree-vect-loop.c: Include internal-fn.h.
4776 (neutral_op_for_slp_reduction): New function, split out from
4777 get_initial_defs_for_reduction.
4778 (get_initial_def_for_reduction): Handle option 2 for variable-length
4779 vectors by loading the neutral value into a vector and then shifting
4780 the initial value into element 0.
4781 (get_initial_defs_for_reduction): Replace the code argument with
4782 the neutral value calculated by neutral_op_for_slp_reduction.
4783 Use gimple_build_vector for constant-length vectors.
4784 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
4785 but the first group_size elements have a neutral value.
4786 Use duplicate_and_interleave otherwise.
4787 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
4788 Update call to get_initial_defs_for_reduction. Handle SLP
4789 reductions for variable-length vectors by creating one vector
4790 result for each scalar result, with the elements associated
4791 with other scalar results stubbed out with the neutral value.
4792 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
4793 Require IFN_VEC_SHL_INSERT for double reductions on
4794 variable-length vectors, or SLP reductions that have
4795 a neutral value. Require can_duplicate_and_interleave_p
4796 support for variable-length unchained SLP reductions if there
4797 is no neutral value, such as for MIN/MAX reductions. Also require
4798 the number of vector elements to be a multiple of the number of
4799 SLP statements when doing variable-length unchained SLP reductions.
4800 Update call to vect_create_epilog_for_reduction.
4801 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
4802 and remove initial values.
4803 (duplicate_and_interleave): Make public.
4804 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
4805 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
4807 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4808 Alan Hayward <alan.hayward@arm.com>
4809 David Sherwood <david.sherwood@arm.com>
4811 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
4812 (can_duplicate_and_interleave_p): New function.
4813 (vect_get_and_check_slp_defs): Take the vector of statements
4814 rather than just the current one. Remove excess parentheses.
4815 Restriction rejectinon of vect_constant_def and vect_external_def
4816 for variable-length vectors to boolean types, or types for which
4817 can_duplicate_and_interleave_p is false.
4818 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
4819 (duplicate_and_interleave): New function.
4820 (vect_get_constant_vectors): Use gimple_build_vector for
4821 constant-length vectors and suitable variable-length constant
4822 vectors. Use duplicate_and_interleave for other variable-length
4823 vectors. Don't defer the update when inserting new statements.
4825 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4826 Alan Hayward <alan.hayward@arm.com>
4827 David Sherwood <david.sherwood@arm.com>
4829 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
4830 min_profitable_iters doesn't go negative.
4832 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4833 Alan Hayward <alan.hayward@arm.com>
4834 David Sherwood <david.sherwood@arm.com>
4836 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
4837 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
4838 * optabs.def (vec_mask_load_lanes_optab): New optab.
4839 (vec_mask_store_lanes_optab): Likewise.
4840 * internal-fn.def (MASK_LOAD_LANES): New internal function.
4841 (MASK_STORE_LANES): Likewise.
4842 * internal-fn.c (mask_load_lanes_direct): New macro.
4843 (mask_store_lanes_direct): Likewise.
4844 (expand_mask_load_optab_fn): Handle masked operations.
4845 (expand_mask_load_lanes_optab_fn): New macro.
4846 (expand_mask_store_optab_fn): Handle masked operations.
4847 (expand_mask_store_lanes_optab_fn): New macro.
4848 (direct_mask_load_lanes_optab_supported_p): Likewise.
4849 (direct_mask_store_lanes_optab_supported_p): Likewise.
4850 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
4852 (vect_load_lanes_supported): Likewise.
4853 * tree-vect-data-refs.c (strip_conversion): New function.
4854 (can_group_stmts_p): Likewise.
4855 (vect_analyze_data_ref_accesses): Use it instead of checking
4856 for a pair of assignments.
4857 (vect_store_lanes_supported): Take a masked_p parameter.
4858 (vect_load_lanes_supported): Likewise.
4859 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
4860 vect_store_lanes_supported and vect_load_lanes_supported.
4861 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4862 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
4863 parameter. Don't allow gaps for masked accesses.
4864 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
4865 and vect_load_lanes_supported.
4866 (get_load_store_type): Take a masked_p parameter and update
4867 call to get_group_load_store_type.
4868 (vectorizable_store): Update call to get_load_store_type.
4869 Handle IFN_MASK_STORE_LANES.
4870 (vectorizable_load): Update call to get_load_store_type.
4871 Handle IFN_MASK_LOAD_LANES.
4873 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4874 Alan Hayward <alan.hayward@arm.com>
4875 David Sherwood <david.sherwood@arm.com>
4877 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
4879 * config/aarch64/aarch64-protos.h
4880 (aarch64_sve_struct_memory_operand_p): Declare.
4881 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
4882 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
4883 (VPRED, vpred): Handle SVE structure modes.
4884 * config/aarch64/constraints.md (Utx): New constraint.
4885 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
4886 (aarch64_sve_struct_nonimmediate_operand): New predicates.
4887 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
4888 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
4889 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
4890 structure modes. Split into pieces after RA.
4891 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
4892 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
4894 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
4895 SVE structure modes.
4896 (aarch64_classify_address): Likewise.
4897 (sizetochar): Move earlier in file.
4898 (aarch64_print_operand): Handle SVE register lists.
4899 (aarch64_array_mode): New function.
4900 (aarch64_sve_struct_memory_operand_p): Likewise.
4901 (TARGET_ARRAY_MODE): Redefine.
4903 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4904 Alan Hayward <alan.hayward@arm.com>
4905 David Sherwood <david.sherwood@arm.com>
4907 * target.def (array_mode): New target hook.
4908 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
4909 * doc/tm.texi: Regenerate.
4910 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
4911 * hooks.c (hook_optmode_mode_uhwi_none): New function.
4912 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
4914 * stor-layout.c (mode_for_array): Likewise. Support polynomial
4917 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4918 Alan Hayward <alan.hayward@arm.com>
4919 David Sherwood <david.sherwood@arm.com>
4921 * fold-const.c (fold_binary_loc): Check the argument types
4922 rather than the result type when testing for a vector operation.
4924 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4926 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
4927 * doc/tm.texi: Regenerate.
4929 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4930 Alan Hayward <alan.hayward@arm.com>
4931 David Sherwood <david.sherwood@arm.com>
4933 * doc/invoke.texi (-msve-vector-bits=): Document new option.
4934 (sve): Document new AArch64 extension.
4935 * doc/md.texi (w): Extend the description of the AArch64
4936 constraint to include SVE vectors.
4937 (Upl, Upa): Document new AArch64 predicate constraints.
4938 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
4940 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
4941 (msve-vector-bits=): New option.
4942 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
4943 SVE when these are disabled.
4944 (sve): New extension.
4945 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
4946 modes. Adjust their number of units based on aarch64_sve_vg.
4947 (MAX_BITSIZE_MODE_ANY_MODE): Define.
4948 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
4949 aarch64_addr_query_type.
4950 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
4951 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
4952 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
4953 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
4954 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
4955 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
4956 (aarch64_simd_imm_zero_p): Delete.
4957 (aarch64_check_zero_based_sve_index_immediate): Declare.
4958 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4959 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4960 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4961 (aarch64_sve_float_mul_immediate_p): Likewise.
4962 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4964 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
4965 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
4966 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
4967 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
4968 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
4969 (aarch64_regmode_natural_size): Likewise.
4970 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
4971 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
4973 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
4974 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
4975 for VG and the SVE predicate registers.
4976 (V_ALIASES): Add a "z"-prefixed alias.
4977 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
4978 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
4979 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
4980 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
4981 (REG_CLASS_NAMES): Add entries for them.
4982 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
4983 and the predicate registers.
4984 (aarch64_sve_vg): Declare.
4985 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
4986 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
4987 (REGMODE_NATURAL_SIZE): Define.
4988 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
4990 * config/aarch64/aarch64.c: Include cfgrtl.h.
4991 (simd_immediate_info): Add a constructor for series vectors,
4992 and an associated step field.
4993 (aarch64_sve_vg): New variable.
4994 (aarch64_dbx_register_number): Handle VG and the predicate registers.
4995 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
4996 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
4997 (VEC_ANY_DATA, VEC_STRUCT): New constants.
4998 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
4999 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
5000 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
5001 (aarch64_get_mask_mode): New functions.
5002 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
5003 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
5004 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
5005 predicate modes and predicate registers. Explicitly restrict
5006 GPRs to modes of 16 bytes or smaller. Only allow FP registers
5007 to store a vector mode if it is recognized by
5008 aarch64_classify_vector_mode.
5009 (aarch64_regmode_natural_size): New function.
5010 (aarch64_hard_regno_caller_save_mode): Return the original mode
5012 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
5013 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
5014 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
5015 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
5017 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
5018 does not overlap dest if the function is frame-related. Handle
5020 (aarch64_split_add_offset): New function.
5021 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
5022 them aarch64_add_offset.
5023 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
5024 and update call to aarch64_sub_sp.
5025 (aarch64_add_cfa_expression): New function.
5026 (aarch64_expand_prologue): Pass extra temporary registers to the
5027 functions above. Handle the case in which we need to emit new
5028 DW_CFA_expressions for registers that were originally saved
5029 relative to the stack pointer, but now have to be expressed
5030 relative to the frame pointer.
5031 (aarch64_output_mi_thunk): Pass extra temporary registers to the
5033 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
5034 IP0 and IP1 values for SVE frames.
5035 (aarch64_expand_vec_series): New function.
5036 (aarch64_expand_sve_widened_duplicate): Likewise.
5037 (aarch64_expand_sve_const_vector): Likewise.
5038 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
5039 Handle SVE constants. Use emit_move_insn to move a force_const_mem
5040 into the register, rather than emitting a SET directly.
5041 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
5042 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
5043 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
5044 (offset_9bit_signed_scaled_p): New functions.
5045 (aarch64_replicate_bitmask_imm): New function.
5046 (aarch64_bitmask_imm): Use it.
5047 (aarch64_cannot_force_const_mem): Reject expressions involving
5048 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
5049 (aarch64_classify_index): Handle SVE indices, by requiring
5050 a plain register index with a scale that matches the element size.
5051 (aarch64_classify_address): Handle SVE addresses. Assert that
5052 the mode of the address is VOIDmode or an integer mode.
5053 Update call to aarch64_classify_symbol.
5054 (aarch64_classify_symbolic_expression): Update call to
5055 aarch64_classify_symbol.
5056 (aarch64_const_vec_all_in_range_p): New function.
5057 (aarch64_print_vector_float_operand): Likewise.
5058 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
5059 "vN" for FP registers with SVE modes. Handle (const ...) vectors
5060 and the FP immediates 1.0 and 0.5.
5061 (aarch64_print_address_internal): Handle SVE addresses.
5062 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
5063 (aarch64_regno_regclass): Handle predicate registers.
5064 (aarch64_secondary_reload): Handle big-endian reloads of SVE
5066 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
5067 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
5068 (aarch64_convert_sve_vector_bits): New function.
5069 (aarch64_override_options): Use it to handle -msve-vector-bits=.
5070 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5072 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
5073 Handle SVE vector and predicate modes. Accept VL-based constants
5074 that need only one temporary register, and VL offsets that require
5075 no temporary registers.
5076 (aarch64_conditional_register_usage): Mark the predicate registers
5077 as fixed if SVE isn't available.
5078 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
5079 Return true for SVE vector and predicate modes.
5080 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
5081 rather than an unsigned int. Handle SVE modes.
5082 (aarch64_preferred_simd_mode): Update call accordingly. Handle
5084 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
5086 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5087 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5088 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5089 (aarch64_sve_float_mul_immediate_p): New functions.
5090 (aarch64_sve_valid_immediate): New function.
5091 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
5092 Explicitly reject structure modes. Check for INDEX constants.
5093 Handle PTRUE and PFALSE constants.
5094 (aarch64_check_zero_based_sve_index_immediate): New function.
5095 (aarch64_simd_imm_zero_p): Delete.
5096 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
5097 vector modes. Accept constants in the range of CNT[BHWD].
5098 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
5099 ask for an Advanced SIMD mode.
5100 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
5101 (aarch64_simd_vector_alignment): Handle SVE predicates.
5102 (aarch64_vectorize_preferred_vector_alignment): New function.
5103 (aarch64_simd_vector_alignment_reachable): Use it instead of
5105 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
5106 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
5108 (MAX_VECT_LEN): Delete.
5109 (expand_vec_perm_d): Add a vec_flags field.
5110 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
5111 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
5112 (aarch64_evpc_ext): Don't apply a big-endian lane correction
5114 (aarch64_evpc_rev): Rename to...
5115 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
5116 (aarch64_evpc_rev_global): New function.
5117 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
5118 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
5120 (aarch64_evpc_sve_tbl): New function.
5121 (aarch64_expand_vec_perm_const_1): Update after rename of
5122 aarch64_evpc_rev. Handle SVE permutes too, trying
5123 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
5124 than aarch64_evpc_tbl.
5125 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
5126 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
5127 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
5128 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
5129 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
5130 (aarch64_expand_sve_vcond): New functions.
5131 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
5132 of aarch64_vector_mode_p.
5133 (aarch64_dwarf_poly_indeterminate_value): New function.
5134 (aarch64_compute_pressure_classes): Likewise.
5135 (aarch64_can_change_mode_class): Likewise.
5136 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
5137 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
5138 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
5139 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
5140 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
5141 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
5142 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
5143 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
5145 (Dn, Dl, Dr): Accept const as well as const_vector.
5146 (Dz): Likewise. Compare against CONST0_RTX.
5147 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
5148 of "vector" where appropriate.
5149 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
5150 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
5151 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
5152 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
5153 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
5154 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
5155 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
5156 (v_int_equiv): Extend to SVE modes.
5157 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
5159 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
5160 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
5161 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
5162 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
5163 (SVE_COND_FP_CMP): New int iterators.
5164 (perm_hilo): Handle the new unpack unspecs.
5165 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
5167 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
5168 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
5169 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
5170 (aarch64_equality_operator, aarch64_constant_vector_operand)
5171 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
5172 (aarch64_sve_nonimmediate_operand): Likewise.
5173 (aarch64_sve_general_operand): Likewise.
5174 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
5175 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
5176 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
5177 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
5178 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
5179 (aarch64_sve_float_arith_immediate): Likewise.
5180 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
5181 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
5182 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
5183 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
5184 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
5185 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
5186 (aarch64_sve_float_arith_operand): Likewise.
5187 (aarch64_sve_float_arith_with_sub_operand): Likewise.
5188 (aarch64_sve_float_mul_operand): Likewise.
5189 (aarch64_sve_vec_perm_operand): Likewise.
5190 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
5191 (aarch64_mov_operand): Accept const_poly_int and const_vector.
5192 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
5193 as well as const_vector.
5194 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
5195 in file. Use CONST0_RTX and CONSTM1_RTX.
5196 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
5197 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
5198 Use aarch64_simd_imm_zero.
5199 * config/aarch64/aarch64-sve.md: New file.
5200 * config/aarch64/aarch64.md: Include it.
5201 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
5202 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
5203 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
5204 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
5205 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
5206 (sve): New attribute.
5207 (enabled): Disable instructions with the sve attribute unless
5209 (movqi, movhi): Pass CONST_POLY_INT operaneds through
5210 aarch64_expand_mov_immediate.
5211 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
5212 CNT[BHSD] immediates.
5213 (movti): Split CONST_POLY_INT moves into two halves.
5214 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
5215 Split additions that need a temporary here if the destination
5216 is the stack pointer.
5217 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
5218 (*add<mode>3_poly_1): New instruction.
5219 (set_clobber_cc): New expander.
5221 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5223 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
5224 parameter and use it instead of GET_MODE_SIZE (innermode). Use
5225 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
5226 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
5227 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
5228 Change innermode from fixed_mode_size to machine_mode.
5229 (simplify_subreg): Update call accordingly. Handle a constant-sized
5230 subreg of a variable-length CONST_VECTOR.
5232 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5233 Alan Hayward <alan.hayward@arm.com>
5234 David Sherwood <david.sherwood@arm.com>
5236 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
5237 (add_offset_to_base): New function, split out from...
5238 (create_mem_ref): ...here. When handling a scale other than 1,
5239 check first whether the address is valid without the offset.
5240 Add it into the base if so, leaving the index and scale as-is.
5242 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5245 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
5246 fold_for_warn before checking if arg2 is INTEGER_CST.
5248 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
5250 * config/rs6000/predicates.md (load_multiple_operation): Delete.
5251 (store_multiple_operation): Delete.
5252 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
5253 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
5254 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
5255 guarded by TARGET_STRING.
5256 (rs6000_output_load_multiple): Delete.
5257 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
5258 OPTION_MASK_STRING / TARGET_STRING handling.
5259 (print_operand) <'N', 'O'>: Add comment that these are unused now.
5260 (const rs6000_opt_masks) <"string">: Change mask to 0.
5261 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
5262 (MASK_STRING): Delete.
5263 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
5265 (load_multiple): Delete.
5272 (store_multiple): Delete.
5279 (movmemsi_8reg): Delete.
5280 (corresponding unnamed define_insn): Delete.
5281 (movmemsi_6reg): Delete.
5282 (corresponding unnamed define_insn): Delete.
5283 (movmemsi_4reg): Delete.
5284 (corresponding unnamed define_insn): Delete.
5285 (movmemsi_2reg): Delete.
5286 (corresponding unnamed define_insn): Delete.
5287 (movmemsi_1reg): Delete.
5288 (corresponding unnamed define_insn): Delete.
5289 * config/rs6000/rs6000.opt (mno-string): New.
5290 (mstring): Replace by deprecation warning stub.
5291 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
5293 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5295 * regrename.c (regrename_do_replace): If replacing the same
5296 reg multiple times, try to reuse last created gen_raw_REG.
5299 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
5300 main to workaround a bug in GDB.
5302 2018-01-12 Tom de Vries <tom@codesourcery.com>
5305 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
5307 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
5309 PR rtl-optimization/80481
5310 * ira-color.c (get_cap_member): New function.
5311 (allocnos_conflict_by_live_ranges_p): Use it.
5312 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
5313 (setup_slot_coalesced_allocno_live_ranges): Ditto.
5315 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
5318 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
5319 (*saddl_se_1): Ditto.
5321 (*ssubl_se_1): Ditto.
5323 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5325 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
5326 rather than wi::to_widest for DR_INITs.
5327 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
5328 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
5329 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
5331 (vect_analyze_group_access_1): Note that here.
5333 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5335 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
5336 polynomial type sizes.
5338 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5340 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
5341 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
5342 (gimple_add_tmp_var): Likewise.
5344 2018-01-12 Martin Liska <mliska@suse.cz>
5346 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
5347 (gimple_alloc_sizes): Likewise.
5348 (dump_gimple_statistics): Use PRIu64 in printf format.
5349 * gimple.h: Change uint64_t to int.
5351 2018-01-12 Martin Liska <mliska@suse.cz>
5353 * tree-core.h: Use uint64_t instead of int.
5354 * tree.c (tree_node_counts): Likewise.
5355 (tree_node_sizes): Likewise.
5356 (dump_tree_statistics): Use PRIu64 in printf format.
5358 2018-01-12 Martin Liska <mliska@suse.cz>
5360 * Makefile.in: As qsort_chk is implemented in vec.c, add
5361 vec.o to linkage of gencfn-macros.
5362 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
5363 passing the info to record_node_allocation_statistics.
5364 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
5366 * ggc-common.c (struct ggc_usage): Add operator== and use
5367 it in operator< and compare function.
5368 * mem-stats.h (struct mem_usage): Likewise.
5369 * vec.c (struct vec_usage): Remove operator< and compare
5370 function. Can be simply inherited.
5372 2018-01-12 Martin Jambor <mjambor@suse.cz>
5375 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
5376 * tree-ssa-math-opts.c: Include domwalk.h.
5377 (convert_mult_to_fma_1): New function.
5378 (fma_transformation_info): New type.
5379 (fma_deferring_state): Likewise.
5380 (cancel_fma_deferring): New function.
5381 (result_of_phi): Likewise.
5382 (last_fma_candidate_feeds_initial_phi): Likewise.
5383 (convert_mult_to_fma): Added deferring logic, split actual
5384 transformation to convert_mult_to_fma_1.
5385 (math_opts_dom_walker): New type.
5386 (math_opts_dom_walker::after_dom_children): New method, body moved
5387 here from pass_optimize_widening_mul::execute, added deferring logic
5389 (pass_optimize_widening_mul::execute): Moved most of code to
5390 math_opts_dom_walker::after_dom_children.
5391 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
5392 * config/i386/i386.c (ix86_option_override_internal): Added
5393 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
5395 2018-01-12 Richard Biener <rguenther@suse.de>
5398 * dwarf2out.c (gen_variable_die): Do not reset old_die for
5399 inline instance vars.
5401 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
5404 * config/rx/rx.c (rx_is_restricted_memory_address):
5407 2018-01-12 Richard Biener <rguenther@suse.de>
5409 PR tree-optimization/80846
5410 * target.def (split_reduction): New target hook.
5411 * targhooks.c (default_split_reduction): New function.
5412 * targhooks.h (default_split_reduction): Declare.
5413 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
5414 target requests first reduce vectors by combining low and high
5416 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
5417 (get_vectype_for_scalar_type_and_size): Export.
5418 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
5419 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
5420 * doc/tm.texi: Regenerate.
5421 * config/i386/i386.c (ix86_split_reduction): Implement
5422 TARGET_VECTORIZE_SPLIT_REDUCTION.
5424 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5427 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
5428 in PIC mode except for TARGET_VXWORKS_RTP.
5429 * config/sparc/sparc.c: Include cfgrtl.h.
5430 (TARGET_INIT_PIC_REG): Define.
5431 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
5432 (sparc_pic_register_p): New predicate.
5433 (sparc_legitimate_address_p): Use it.
5434 (sparc_legitimize_pic_address): Likewise.
5435 (sparc_delegitimize_address): Likewise.
5436 (sparc_mode_dependent_address_p): Likewise.
5437 (gen_load_pcrel_sym): Remove 4th parameter.
5438 (load_got_register): Adjust call to above. Remove obsolete stuff.
5439 (sparc_expand_prologue): Do not call load_got_register here.
5440 (sparc_flat_expand_prologue): Likewise.
5441 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
5442 (sparc_use_pseudo_pic_reg): New function.
5443 (sparc_init_pic_reg): Likewise.
5444 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5445 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
5447 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
5449 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
5450 Add item for branch_cost.
5452 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5454 PR rtl-optimization/83565
5455 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
5456 not extend the result to a larger mode for rotate operations.
5457 (num_sign_bit_copies1): Likewise.
5459 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5462 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
5464 Use values-Xc.o for -pedantic.
5465 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
5467 2018-01-12 Martin Liska <mliska@suse.cz>
5470 * ipa-devirt.c (final_warning_record::grow_type_warnings):
5472 (possible_polymorphic_call_targets): Use it.
5473 (ipa_devirt): Likewise.
5475 2018-01-12 Martin Liska <mliska@suse.cz>
5477 * profile-count.h (enum profile_quality): Use 0 as invalid
5478 enum value of profile_quality.
5480 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
5482 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
5483 -mext-string options.
5485 2018-01-12 Richard Biener <rguenther@suse.de>
5487 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
5488 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
5489 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
5491 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
5493 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5495 * configure.ac (--with-long-double-format): Add support for the
5496 configuration option to change the default long double format on
5498 * config.gcc (powerpc*-linux*-*): Likewise.
5499 * configure: Regenerate.
5500 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
5501 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
5502 used without modification.
5504 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5506 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
5507 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
5508 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
5509 MISC_BUILTIN_SPEC_BARRIER.
5510 (rs6000_init_builtins): Likewise.
5511 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
5513 (speculation_barrier): New define_insn.
5514 * doc/extend.texi: Document __builtin_speculation_barrier.
5516 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5519 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
5520 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
5521 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
5523 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
5524 integral modes instead of "ss" and "sd".
5525 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
5526 vectors with 32-bit and 64-bit elements.
5527 (vecdupssescalarmodesuffix): New mode attribute.
5528 (vec_dup<mode>): Use it.
5530 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
5533 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
5534 frame if argument is passed on stack.
5536 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5539 * ree.c (combine_reaching_defs): Optimize also
5540 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
5541 reg2=any_extend(exp); reg1=reg2;, formatting fix.
5543 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5546 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
5548 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5551 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
5552 after they are computed.
5554 2018-01-11 Bin Cheng <bin.cheng@arm.com>
5556 PR tree-optimization/83695
5557 * gimple-loop-linterchange.cc
5558 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
5559 reset cached scev information after interchange.
5560 (pass_linterchange::execute): Remove call to scev_reset_htab.
5562 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5564 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
5565 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
5566 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
5567 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
5568 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
5569 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
5570 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
5571 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
5572 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
5573 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
5574 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
5575 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
5576 (V_lane_reg): Likewise.
5577 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
5579 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
5580 (vfmal_lane_low<mode>_intrinsic,
5581 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
5582 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
5583 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
5584 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
5585 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
5586 vfmsl_lane_high<mode>_intrinsic): New define_insns.
5588 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5590 * config/arm/arm-cpus.in (fp16fml): New feature.
5591 (ALL_SIMD): Add fp16fml.
5592 (armv8.2-a): Add fp16fml as an option.
5593 (armv8.3-a): Likewise.
5594 (armv8.4-a): Add fp16fml as part of fp16.
5595 * config/arm/arm.h (TARGET_FP16FML): Define.
5596 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
5598 * config/arm/arm-modes.def (V2HF): Define.
5599 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
5600 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
5601 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
5602 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
5603 vfmsl_low, vfmsl_high): New set of builtins.
5604 * config/arm/iterators.md (PLUSMINUS): New code iterator.
5605 (vfml_op): New code attribute.
5606 (VFMLHALVES): New int iterator.
5607 (VFML, VFMLSEL): New mode attributes.
5608 (V_reg): Define mapping for V2HF.
5609 (V_hi, V_lo): New mode attributes.
5610 (VF_constraint): Likewise.
5611 (vfml_half, vfml_half_selector): New int attributes.
5612 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
5614 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
5615 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
5617 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
5618 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
5619 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
5620 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
5622 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
5623 Document new effective target and option set.
5625 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5627 * config/arm/arm-cpus.in (armv8_4): New feature.
5628 (ARMv8_4a): New fgroup.
5629 (armv8.4-a): New arch.
5630 * config/arm/arm-tables.opt: Regenerate.
5631 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
5632 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
5633 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
5634 Add matching rules for -march=armv8.4-a and extensions.
5635 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
5637 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
5640 * config/rx/rx.md (BW): New mode attribute.
5641 (sync_lock_test_and_setsi): Add mode suffix to insn output.
5643 2018-01-11 Richard Biener <rguenther@suse.de>
5645 PR tree-optimization/83435
5646 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
5647 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
5648 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
5650 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5651 Alan Hayward <alan.hayward@arm.com>
5652 David Sherwood <david.sherwood@arm.com>
5654 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
5656 (aarch64_classify_address): Initialize it. Track polynomial offsets.
5657 (aarch64_print_address_internal): Use it to check for a zero offset.
5659 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5660 Alan Hayward <alan.hayward@arm.com>
5661 David Sherwood <david.sherwood@arm.com>
5663 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
5664 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
5665 Return a poly_int64 rather than a HOST_WIDE_INT.
5666 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
5667 rather than a HOST_WIDE_INT.
5668 * config/aarch64/aarch64.h (aarch64_frame): Protect with
5669 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
5670 hard_fp_offset, frame_size, initial_adjust, callee_offset and
5671 final_offset from HOST_WIDE_INT to poly_int64.
5672 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
5673 to_constant when getting the number of units in an Advanced SIMD
5675 (aarch64_builtin_vectorized_function): Check for a constant number
5677 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
5679 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
5680 attribute instead of GET_MODE_NUNITS.
5681 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
5682 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
5683 GET_MODE_SIZE for fixed-size registers.
5684 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
5685 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
5686 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
5687 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
5688 (aarch64_print_operand, aarch64_print_address_internal)
5689 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
5690 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
5691 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
5692 Handle polynomial GET_MODE_SIZE.
5693 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
5694 wider than SImode without modification.
5695 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
5696 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
5697 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
5698 passing and returning SVE modes.
5699 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
5700 rather than GEN_INT.
5701 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
5702 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
5703 (aarch64_allocate_and_probe_stack_space): Likewise.
5704 (aarch64_layout_frame): Cope with polynomial offsets.
5705 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
5706 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
5708 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
5709 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
5710 poly_int64 rather than a HOST_WIDE_INT.
5711 (aarch64_get_separate_components, aarch64_process_components)
5712 (aarch64_expand_prologue, aarch64_expand_epilogue)
5713 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
5714 (aarch64_anchor_offset): New function, split out from...
5715 (aarch64_legitimize_address): ...here.
5716 (aarch64_builtin_vectorization_cost): Handle polynomial
5717 TYPE_VECTOR_SUBPARTS.
5718 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
5720 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
5721 number of elements from the PARALLEL rather than the mode.
5722 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
5723 rather than GET_MODE_BITSIZE.
5724 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
5725 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
5726 (aarch64_expand_vec_perm_const_1): Handle polynomial
5727 d->perm.length () and d->perm elements.
5728 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
5729 Apply to_constant to d->perm elements.
5730 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
5731 polynomial CONST_VECTOR_NUNITS.
5732 (aarch64_move_pointer): Take amount as a poly_int64 rather
5734 (aarch64_progress_pointer): Avoid temporary variable.
5735 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
5736 the mode attribute instead of GET_MODE.
5738 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5739 Alan Hayward <alan.hayward@arm.com>
5740 David Sherwood <david.sherwood@arm.com>
5742 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
5743 x exists before using it.
5744 (aarch64_add_constant_internal): Rename to...
5745 (aarch64_add_offset_1): ...this. Replace regnum with separate
5746 src and dest rtxes. Handle the case in which they're different,
5747 including when the offset is zero. Replace scratchreg with an rtx.
5748 Use 2 additions if there is no spare register into which we can
5749 move a 16-bit constant.
5750 (aarch64_add_constant): Delete.
5751 (aarch64_add_offset): Replace reg with separate src and dest
5752 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
5753 Use aarch64_add_offset_1.
5754 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
5755 an rtx rather than an int. Take the delta as a poly_int64
5756 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
5757 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
5758 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
5759 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
5760 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
5762 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
5763 aarch64_add_constant.
5765 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5767 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
5768 Use scalar_float_mode.
5770 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5772 * config/aarch64/aarch64-simd.md
5773 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
5774 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
5775 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
5776 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
5777 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
5778 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
5779 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
5780 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
5781 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
5782 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
5784 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5787 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
5788 targ_options->x_arm_arch_string is non NULL.
5790 2018-01-11 Tamar Christina <tamar.christina@arm.com>
5792 * config/aarch64/aarch64.h
5793 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
5795 2018-01-11 Sudakshina Das <sudi.das@arm.com>
5798 * expmed.c (emit_store_flag_force): Swap if const op0
5799 and change VOIDmode to mode of op0.
5801 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5803 PR rtl-optimization/83761
5804 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
5805 than bytes to mode_for_size.
5807 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5810 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
5811 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
5814 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5817 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
5818 when in layout mode.
5819 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
5820 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
5823 2018-01-10 Michael Collison <michael.collison@arm.com>
5825 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
5826 * config/aarch64/aarch64-option-extension.def: Add
5827 AARCH64_OPT_EXTENSION of 'fp16fml'.
5828 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5829 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
5830 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
5831 * config/aarch64/constraints.md (Ui7): New constraint.
5832 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
5833 (VFMLA_SEL_W): Ditto.
5836 (VFMLA16_LOW): New int iterator.
5837 (VFMLA16_HIGH): Ditto.
5838 (UNSPEC_FMLAL): New unspec.
5839 (UNSPEC_FMLSL): Ditto.
5840 (UNSPEC_FMLAL2): Ditto.
5841 (UNSPEC_FMLSL2): Ditto.
5842 (f16mac): New code attribute.
5843 * config/aarch64/aarch64-simd-builtins.def
5844 (aarch64_fmlal_lowv2sf): Ditto.
5845 (aarch64_fmlsl_lowv2sf): Ditto.
5846 (aarch64_fmlalq_lowv4sf): Ditto.
5847 (aarch64_fmlslq_lowv4sf): Ditto.
5848 (aarch64_fmlal_highv2sf): Ditto.
5849 (aarch64_fmlsl_highv2sf): Ditto.
5850 (aarch64_fmlalq_highv4sf): Ditto.
5851 (aarch64_fmlslq_highv4sf): Ditto.
5852 (aarch64_fmlal_lane_lowv2sf): Ditto.
5853 (aarch64_fmlsl_lane_lowv2sf): Ditto.
5854 (aarch64_fmlal_laneq_lowv2sf): Ditto.
5855 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
5856 (aarch64_fmlalq_lane_lowv4sf): Ditto.
5857 (aarch64_fmlsl_lane_lowv4sf): Ditto.
5858 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
5859 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
5860 (aarch64_fmlal_lane_highv2sf): Ditto.
5861 (aarch64_fmlsl_lane_highv2sf): Ditto.
5862 (aarch64_fmlal_laneq_highv2sf): Ditto.
5863 (aarch64_fmlsl_laneq_highv2sf): Ditto.
5864 (aarch64_fmlalq_lane_highv4sf): Ditto.
5865 (aarch64_fmlsl_lane_highv4sf): Ditto.
5866 (aarch64_fmlalq_laneq_highv4sf): Ditto.
5867 (aarch64_fmlsl_laneq_highv4sf): Ditto.
5868 * config/aarch64/aarch64-simd.md:
5869 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
5870 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5871 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
5872 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5873 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
5874 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
5875 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
5876 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
5877 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
5878 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
5879 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
5880 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
5881 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
5882 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
5883 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
5884 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
5885 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
5886 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
5887 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
5888 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
5889 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
5890 (vfmlsl_low_u32): Ditto.
5891 (vfmlalq_low_u32): Ditto.
5892 (vfmlslq_low_u32): Ditto.
5893 (vfmlal_high_u32): Ditto.
5894 (vfmlsl_high_u32): Ditto.
5895 (vfmlalq_high_u32): Ditto.
5896 (vfmlslq_high_u32): Ditto.
5897 (vfmlal_lane_low_u32): Ditto.
5898 (vfmlsl_lane_low_u32): Ditto.
5899 (vfmlal_laneq_low_u32): Ditto.
5900 (vfmlsl_laneq_low_u32): Ditto.
5901 (vfmlalq_lane_low_u32): Ditto.
5902 (vfmlslq_lane_low_u32): Ditto.
5903 (vfmlalq_laneq_low_u32): Ditto.
5904 (vfmlslq_laneq_low_u32): Ditto.
5905 (vfmlal_lane_high_u32): Ditto.
5906 (vfmlsl_lane_high_u32): Ditto.
5907 (vfmlal_laneq_high_u32): Ditto.
5908 (vfmlsl_laneq_high_u32): Ditto.
5909 (vfmlalq_lane_high_u32): Ditto.
5910 (vfmlslq_lane_high_u32): Ditto.
5911 (vfmlalq_laneq_high_u32): Ditto.
5912 (vfmlslq_laneq_high_u32): Ditto.
5913 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
5914 (AARCH64_FL_FOR_ARCH8_4): New.
5915 (AARCH64_ISA_F16FML): New ISA flag.
5916 (TARGET_F16FML): New feature flag for fp16fml.
5917 (doc/invoke.texi): Document new fp16fml option.
5919 2018-01-10 Michael Collison <michael.collison@arm.com>
5921 * config/aarch64/aarch64-builtins.c:
5922 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
5923 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5924 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
5925 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
5926 (AARCH64_ISA_SHA3): New ISA flag.
5927 (TARGET_SHA3): New feature flag for sha3.
5928 * config/aarch64/iterators.md (sha512_op): New int attribute.
5929 (CRYPTO_SHA512): New int iterator.
5930 (UNSPEC_SHA512H): New unspec.
5931 (UNSPEC_SHA512H2): Ditto.
5932 (UNSPEC_SHA512SU0): Ditto.
5933 (UNSPEC_SHA512SU1): Ditto.
5934 * config/aarch64/aarch64-simd-builtins.def
5935 (aarch64_crypto_sha512hqv2di): New builtin.
5936 (aarch64_crypto_sha512h2qv2di): Ditto.
5937 (aarch64_crypto_sha512su0qv2di): Ditto.
5938 (aarch64_crypto_sha512su1qv2di): Ditto.
5939 (aarch64_eor3qv8hi): Ditto.
5940 (aarch64_rax1qv2di): Ditto.
5941 (aarch64_xarqv2di): Ditto.
5942 (aarch64_bcaxqv8hi): Ditto.
5943 * config/aarch64/aarch64-simd.md:
5944 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
5945 (aarch64_crypto_sha512su0qv2di): Ditto.
5946 (aarch64_crypto_sha512su1qv2di): Ditto.
5947 (aarch64_eor3qv8hi): Ditto.
5948 (aarch64_rax1qv2di): Ditto.
5949 (aarch64_xarqv2di): Ditto.
5950 (aarch64_bcaxqv8hi): Ditto.
5951 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
5952 (vsha512h2q_u64): Ditto.
5953 (vsha512su0q_u64): Ditto.
5954 (vsha512su1q_u64): Ditto.
5955 (veor3q_u16): Ditto.
5956 (vrax1q_u64): Ditto.
5958 (vbcaxq_u16): Ditto.
5959 * config/arm/types.md (crypto_sha512): New type attribute.
5960 (crypto_sha3): Ditto.
5961 (doc/invoke.texi): Document new sha3 option.
5963 2018-01-10 Michael Collison <michael.collison@arm.com>
5965 * config/aarch64/aarch64-builtins.c:
5966 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
5967 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5968 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
5969 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
5970 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
5971 (AARCH64_ISA_SM4): New ISA flag.
5972 (TARGET_SM4): New feature flag for sm4.
5973 * config/aarch64/aarch64-simd-builtins.def
5974 (aarch64_sm3ss1qv4si): Ditto.
5975 (aarch64_sm3tt1aq4si): Ditto.
5976 (aarch64_sm3tt1bq4si): Ditto.
5977 (aarch64_sm3tt2aq4si): Ditto.
5978 (aarch64_sm3tt2bq4si): Ditto.
5979 (aarch64_sm3partw1qv4si): Ditto.
5980 (aarch64_sm3partw2qv4si): Ditto.
5981 (aarch64_sm4eqv4si): Ditto.
5982 (aarch64_sm4ekeyqv4si): Ditto.
5983 * config/aarch64/aarch64-simd.md:
5984 (aarch64_sm3ss1qv4si): Ditto.
5985 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
5986 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
5987 (aarch64_sm4eqv4si): Ditto.
5988 (aarch64_sm4ekeyqv4si): Ditto.
5989 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
5990 (sm3part_op): Ditto.
5991 (CRYPTO_SM3TT): Ditto.
5992 (CRYPTO_SM3PART): Ditto.
5993 (UNSPEC_SM3SS1): New unspec.
5994 (UNSPEC_SM3TT1A): Ditto.
5995 (UNSPEC_SM3TT1B): Ditto.
5996 (UNSPEC_SM3TT2A): Ditto.
5997 (UNSPEC_SM3TT2B): Ditto.
5998 (UNSPEC_SM3PARTW1): Ditto.
5999 (UNSPEC_SM3PARTW2): Ditto.
6000 (UNSPEC_SM4E): Ditto.
6001 (UNSPEC_SM4EKEY): Ditto.
6002 * config/aarch64/constraints.md (Ui2): New constraint.
6003 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
6004 * config/arm/types.md (crypto_sm3): New type attribute.
6005 (crypto_sm4): Ditto.
6006 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
6007 (vsm3tt1aq_u32): Ditto.
6008 (vsm3tt1bq_u32): Ditto.
6009 (vsm3tt2aq_u32): Ditto.
6010 (vsm3tt2bq_u32): Ditto.
6011 (vsm3partw1q_u32): Ditto.
6012 (vsm3partw2q_u32): Ditto.
6013 (vsm4eq_u32): Ditto.
6014 (vsm4ekeyq_u32): Ditto.
6015 (doc/invoke.texi): Document new sm4 option.
6017 2018-01-10 Michael Collison <michael.collison@arm.com>
6019 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
6020 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
6021 (AARCH64_FL_FOR_ARCH8_4): New.
6022 (AARCH64_FL_V8_4): New flag.
6023 (doc/invoke.texi): Document new armv8.4-a option.
6025 2018-01-10 Michael Collison <michael.collison@arm.com>
6027 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6028 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
6029 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
6030 * config/aarch64/aarch64-option-extension.def: Add
6031 AARCH64_OPT_EXTENSION of 'sha2'.
6032 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
6033 (crypto): Disable sha2 and aes if crypto disabled.
6034 (crypto): Enable aes and sha2 if enabled.
6035 (simd): Disable sha2 and aes if simd disabled.
6036 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
6038 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
6039 (TARGET_SHA2): New feature flag for sha2.
6040 (TARGET_AES): New feature flag for aes.
6041 * config/aarch64/aarch64-simd.md:
6042 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
6043 conditional on TARGET_AES.
6044 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
6045 (aarch64_crypto_sha1hsi): Make pattern conditional
6047 (aarch64_crypto_sha1hv4si): Ditto.
6048 (aarch64_be_crypto_sha1hv4si): Ditto.
6049 (aarch64_crypto_sha1su1v4si): Ditto.
6050 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
6051 (aarch64_crypto_sha1su0v4si): Ditto.
6052 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
6053 (aarch64_crypto_sha256su0v4si): Ditto.
6054 (aarch64_crypto_sha256su1v4si): Ditto.
6055 (doc/invoke.texi): Document new aes and sha2 options.
6057 2018-01-10 Martin Sebor <msebor@redhat.com>
6059 PR tree-optimization/83781
6060 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
6063 2018-01-11 Martin Sebor <msebor@gmail.com>
6064 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6066 PR tree-optimization/83501
6067 PR tree-optimization/81703
6069 * tree-ssa-strlen.c (get_string_cst): Rename...
6070 (get_string_len): ...to this. Handle global constants.
6071 (handle_char_store): Adjust.
6073 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
6074 Jim Wilson <jimw@sifive.com>
6076 * config/riscv/riscv-protos.h (riscv_output_return): New.
6077 * config/riscv/riscv.c (struct machine_function): New naked_p field.
6078 (riscv_attribute_table, riscv_output_return),
6079 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
6080 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
6081 (riscv_compute_frame_info): Only compute frame->mask if not a naked
6083 (riscv_expand_prologue): Add early return for naked function.
6084 (riscv_expand_epilogue): Likewise.
6085 (riscv_function_ok_for_sibcall): Return false for naked function.
6086 (riscv_set_current_function): New.
6087 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
6088 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
6089 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
6090 * doc/extend.texi (RISC-V Function Attributes): New.
6092 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6094 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
6095 check for 128-bit long double before checking TCmode.
6096 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
6097 128-bit long doubles before checking TFmode or TCmode.
6098 (FLOAT128_IBM_P): Likewise.
6100 2018-01-10 Martin Sebor <msebor@redhat.com>
6102 PR tree-optimization/83671
6103 * builtins.c (c_strlen): Unconditionally return zero for the empty
6105 Use -Warray-bounds for warnings.
6106 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
6107 for non-constant array indices with COMPONENT_REF, arrays of
6108 arrays, and pointers to arrays.
6109 (gimple_fold_builtin_strlen): Determine and set length range for
6110 non-constant character arrays.
6112 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
6115 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
6118 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
6120 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
6122 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6125 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
6126 VECTOR_MEM_ALTIVEC_OR_VSX_P.
6127 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
6128 indexed_or_indirect_operand predicate.
6129 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
6130 (*vsx_le_perm_load_v8hi): Likewise.
6131 (*vsx_le_perm_load_v16qi): Likewise.
6132 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
6133 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
6134 (*vsx_le_perm_store_v8hi): Likewise.
6135 (*vsx_le_perm_store_v16qi): Likewise.
6136 (eight unnamed splitters): Likewise.
6138 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6140 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
6141 * config/rs6000/emmintrin.h: Likewise.
6142 * config/rs6000/mmintrin.h: Likewise.
6143 * config/rs6000/xmmintrin.h: Likewise.
6145 2018-01-10 David Malcolm <dmalcolm@redhat.com>
6148 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
6150 * tree.c (tree_nop_conversion): Return true for location wrapper
6152 (maybe_wrap_with_location): New function.
6153 (selftest::check_strip_nops): New function.
6154 (selftest::test_location_wrappers): New function.
6155 (selftest::tree_c_tests): Call it.
6156 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
6157 (maybe_wrap_with_location): New decl.
6158 (EXPR_LOCATION_WRAPPER_P): New macro.
6159 (location_wrapper_p): New inline function.
6160 (tree_strip_any_location_wrapper): New inline function.
6162 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
6165 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
6166 stack_realign_offset for the largest alignment of stack slot
6168 (ix86_find_max_used_stack_alignment): New function.
6169 (ix86_finalize_stack_frame_flags): Use it. Set
6170 max_used_stack_alignment if we don't realign stack.
6171 * config/i386/i386.h (machine_function): Add
6172 max_used_stack_alignment.
6174 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
6176 * config/arm/arm.opt (-mbranch-cost): New option.
6177 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
6180 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
6183 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
6184 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
6186 2018-01-10 Richard Biener <rguenther@suse.de>
6189 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
6190 early out so it also covers the case where we have a non-NULL
6193 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6195 PR tree-optimization/83753
6196 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
6197 for non-strided grouped accesses if the number of elements is 1.
6199 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6202 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
6203 * i386.h (TARGET_USE_GATHER): Define.
6204 * x86-tune.def (X86_TUNE_USE_GATHER): New.
6206 2018-01-10 Martin Liska <mliska@suse.cz>
6209 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
6210 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
6212 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
6213 CLEANUP_NO_PARTITIONING is not set.
6215 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6217 * doc/rtl.texi: Remove documentation of (const ...) wrappers
6218 for vectors, as a partial revert of r254296.
6219 * rtl.h (const_vec_p): Delete.
6220 (const_vec_duplicate_p): Don't test for vector CONSTs.
6221 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
6222 * expmed.c (make_tree): Likewise.
6225 * common.md (E, F): Use CONSTANT_P instead of checking for
6227 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
6228 checking for CONST_VECTOR.
6230 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6233 * predict.c (force_edge_cold): Handle in more sane way edges
6236 2018-01-09 Carl Love <cel@us.ibm.com>
6238 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
6240 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
6241 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
6242 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
6243 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
6244 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
6245 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
6246 * config/rs6000/rs6000-protos.h: Add extern defition for
6247 rs6000_generate_float2_double_code.
6248 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
6250 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
6251 (float2_v2df): Add define_expand.
6253 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
6256 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
6257 op_mode in the force_to_mode call.
6259 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6261 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
6262 instead of checking each element individually.
6263 (aarch64_evpc_uzp): Likewise.
6264 (aarch64_evpc_zip): Likewise.
6265 (aarch64_evpc_ext): Likewise.
6266 (aarch64_evpc_rev): Likewise.
6267 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
6268 instead of checking each element individually. Return true without
6270 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
6271 whether all selected elements come from the same input, instead of
6272 checking each element individually. Remove calls to gen_rtx_REG,
6273 start_sequence and end_sequence and instead assert that no rtl is
6276 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6278 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
6279 order of HIGH and CONST checks.
6281 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6283 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
6284 if the destination isn't an SSA_NAME.
6286 2018-01-09 Richard Biener <rguenther@suse.de>
6288 PR tree-optimization/83668
6289 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
6291 (canonicalize_loop_form): ... here, renamed from ...
6292 (canonicalize_loop_closed_ssa_form): ... this and amended to
6293 swap successor edges for loop exit blocks to make us use
6294 the RPO order we need for initial schedule generation.
6296 2018-01-09 Joseph Myers <joseph@codesourcery.com>
6298 PR tree-optimization/64811
6299 * match.pd: When optimizing comparisons with Inf, avoid
6300 introducing or losing exceptions from comparisons with NaN.
6302 2018-01-09 Martin Liska <mliska@suse.cz>
6305 * asan.c (shadow_mem_size): Add gcc_assert.
6307 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
6309 Don't save registers in main().
6312 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
6313 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
6314 * config/avr/avr.c (avr_set_current_function): Don't error if
6315 naked, OS_task or OS_main are specified at the same time.
6316 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
6318 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
6320 * common/config/avr/avr-common.c (avr_option_optimization_table):
6321 Switch on -mmain-is-OS_task for optimizing compilations.
6323 2018-01-09 Richard Biener <rguenther@suse.de>
6325 PR tree-optimization/83572
6326 * graphite.c: Include cfganal.h.
6327 (graphite_transform_loops): Connect infinite loops to exit
6328 and remove fake edges at the end.
6330 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6332 * ipa-inline.c (edge_badness): Revert accidental checkin.
6334 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6337 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
6338 symbols; not inline clones.
6340 2018-01-09 Jakub Jelinek <jakub@redhat.com>
6343 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
6344 hard registers. Formatting fixes.
6346 PR preprocessor/83722
6347 * gcc.c (try_generate_repro): Pass
6348 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
6349 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
6352 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
6353 Kito Cheng <kito.cheng@gmail.com>
6355 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
6356 (riscv_leaf_function_p): Delete.
6357 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
6359 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6361 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
6363 (do_ifelse): New function.
6364 (do_isel): New function.
6365 (do_sub3): New function.
6366 (do_add3): New function.
6367 (do_load_mask_compare): New function.
6368 (do_overlap_load_compare): New function.
6369 (expand_compare_loop): New function.
6370 (expand_block_compare): Call expand_compare_loop() when appropriate.
6371 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
6373 (-mblock-compare-inline-loop-limit): New option.
6375 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6378 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
6379 Reverse order of second and third operands in first alternative.
6380 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
6381 of first and second elements in UNSPEC_VPERMR vector.
6382 (altivec_expand_vec_perm_le): Likewise.
6384 2018-01-08 Jeff Law <law@redhat.com>
6386 PR rtl-optimizatin/81308
6387 * tree-switch-conversion.c (cfg_altered): New file scoped static.
6388 (process_switch): If group_case_labels makes a change, then set
6390 (pass_convert_switch::execute): If a switch is converted, then
6391 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
6393 PR rtl-optimization/81308
6394 * recog.c (split_all_insns): Conditionally cleanup the CFG after
6397 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
6399 PR target/83663 - Revert r255946
6400 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
6401 generation for cases where splatting a value is not useful.
6402 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
6403 across a vec_duplicate and a paradoxical subreg forming a vector
6404 mode to a vec_concat.
6406 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6408 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
6409 -march=armv8.3-a variants.
6410 * config/arm/t-multilib: Likewise.
6411 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
6413 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6415 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
6417 (cceq_ior_compare_complement): Give it a name so I can use it, and
6418 change boolean_or_operator predicate to boolean_operator so it can
6419 be used to generate a crand.
6420 (eqne): New code iterator.
6421 (bd/bd_neg): New code_attrs.
6422 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
6423 a single define_insn.
6424 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
6425 decrement (bdnzt/bdnzf/bdzt/bdzf).
6426 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
6427 with the new names of the branch decrement patterns, and added the
6428 names of the branch decrement conditional patterns.
6430 2018-01-08 Richard Biener <rguenther@suse.de>
6432 PR tree-optimization/83563
6433 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
6436 2018-01-08 Richard Biener <rguenther@suse.de>
6439 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
6441 2018-01-08 Richard Biener <rguenther@suse.de>
6443 PR tree-optimization/83685
6444 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
6445 references to abnormals.
6447 2018-01-08 Richard Biener <rguenther@suse.de>
6450 * dwarf2out.c (output_indirect_strings): Handle empty
6451 skeleton_debug_str_hash.
6452 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
6454 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6456 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
6457 (emit_store_direct): Likewise.
6458 (arc_trampoline_adjust_address): Likewise.
6459 (arc_asm_trampoline_template): New function.
6460 (arc_initialize_trampoline): Use asm_trampoline_template.
6461 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
6462 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
6463 * config/arc/arc.md (flush_icache): Delete pattern.
6465 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6467 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
6468 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
6471 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6474 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
6475 by not USED_FOR_TARGET.
6476 (make_pass_resolve_sw_modes): Likewise.
6478 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6480 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
6483 2018-01-08 Richard Biener <rguenther@suse.de>
6486 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
6488 2018-01-08 Richard Biener <rguenther@suse.de>
6491 * match.pd ((t * 2) / 2) -> t): Add missing :c.
6493 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
6496 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
6497 basic blocks with a small number of successors.
6498 (convert_control_dep_chain_into_preds): Improve handling of
6500 (dump_predicates): Split apart into...
6501 (dump_pred_chain): ...here...
6502 (dump_pred_info): ...and here.
6503 (can_one_predicate_be_invalidated_p): Add debugging printfs.
6504 (can_chain_union_be_invalidated_p): Improve check for invalidation
6506 (uninit_uses_cannot_happen): Avoid unnecessary if
6507 convert_control_dep_chain_into_preds yielded nothing.
6509 2018-01-06 Martin Sebor <msebor@redhat.com>
6511 PR tree-optimization/83640
6512 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
6513 subtracting negative offset from size.
6514 (builtin_access::overlap): Adjust offset bounds of the access to fall
6515 within the size of the object if possible.
6517 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
6519 PR rtl-optimization/83699
6520 * expmed.c (extract_bit_field_1): Restrict the vector usage of
6521 extract_bit_field_as_subreg to cases in which the extracted
6522 value is also a vector.
6524 * lra-constraints.c (process_alt_operands): Test for the equivalence
6525 substitutions when detecting a possible reload cycle.
6527 2018-01-06 Jakub Jelinek <jakub@redhat.com>
6530 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
6531 by default if flag_selective_schedling{,2}. Formatting fixes.
6533 PR rtl-optimization/83682
6534 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
6535 if it has non-VECTOR_MODE element mode.
6536 (vec_duplicate_p): Likewise.
6539 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
6540 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
6542 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6545 * config/i386/i386-builtin.def
6546 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
6547 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
6548 Require also OPTION_MASK_ISA_AVX512F in addition to
6549 OPTION_MASK_ISA_GFNI.
6550 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
6551 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
6552 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
6553 to OPTION_MASK_ISA_GFNI.
6554 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
6555 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
6556 OPTION_MASK_ISA_AVX512BW.
6557 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
6558 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
6559 addition to OPTION_MASK_ISA_GFNI.
6560 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
6561 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
6562 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
6563 to OPTION_MASK_ISA_GFNI.
6564 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
6565 a requirement for all ISAs rather than any of them with a few
6567 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
6569 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
6570 bitmasks to be enabled with 3 exceptions, instead of requiring any
6571 enabled ISA with lots of exceptions.
6572 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
6573 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
6574 Change avx512bw in isa attribute to avx512f.
6575 * config/i386/sgxintrin.h: Add license boilerplate.
6576 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
6577 to __AVX512F__ and __AVX512VL to __AVX512VL__.
6578 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
6579 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
6581 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
6582 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
6583 temporarily sse2 rather than sse if not enabled already.
6586 * config/i386/sse.md (VI248_VLBW): Rename to ...
6587 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
6588 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
6589 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
6590 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
6591 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
6592 mode iterator instead of VI248_VLBW.
6594 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
6596 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
6597 (record_modified): Skip clobbers; add debug output.
6598 (param_change_prob): Use sreal frequencies.
6600 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6602 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
6603 punt for user-aligned variables.
6605 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6607 * tree-chrec.c (chrec_contains_symbols): Return true for
6610 2018-01-05 Sudakshina Das <sudi.das@arm.com>
6613 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
6614 of (x|y) == x for BICS pattern.
6616 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6618 PR tree-optimization/83605
6619 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
6620 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
6623 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
6625 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
6626 * config/epiphany/rtems.h: New file.
6628 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6629 Uros Bizjak <ubizjak@gmail.com>
6632 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
6633 QIreg_operand instead of register_operand predicate.
6634 * config/i386/i386.c (ix86_rop_should_change_byte_p,
6635 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
6636 comments instead of -fmitigate[-_]rop.
6638 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6641 * cgraphunit.c (symbol_table::compile): Switch to text_section
6642 before calling assembly_start debug hook.
6643 * run-rtl-passes.c (run_rtl_passes): Likewise.
6646 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6648 * tree-vrp.c (extract_range_from_binary_expr_1): Check
6649 range_int_cst_p rather than !symbolic_range_p before calling
6650 extract_range_from_multiplicative_op_1.
6652 2018-01-04 Jeff Law <law@redhat.com>
6654 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
6655 redundant test in assertion.
6657 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6659 * doc/rtl.texi: Document machine_mode wrapper classes.
6661 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6663 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
6666 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6668 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
6669 the VEC_PERM_EXPR fold to fail.
6671 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6674 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
6675 to switched_sections.
6677 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6680 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
6683 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
6686 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
6687 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
6689 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6692 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
6693 is BLKmode and bitpos not zero or mode change is needed.
6695 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6698 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
6701 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
6704 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
6705 instead of MULT rtx. Update all corresponding splitters.
6707 (*ssub<modesuffix>): Ditto.
6709 (*cmp_sadd_di): Update split patterns.
6710 (*cmp_sadd_si): Ditto.
6711 (*cmp_sadd_sidi): Ditto.
6712 (*cmp_ssub_di): Ditto.
6713 (*cmp_ssub_si): Ditto.
6714 (*cmp_ssub_sidi): Ditto.
6715 * config/alpha/predicates.md (const23_operand): New predicate.
6716 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
6717 Look for ASHIFT, not MULT inner operand.
6718 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
6720 2018-01-04 Martin Liska <mliska@suse.cz>
6722 PR gcov-profile/83669
6723 * gcov.c (output_intermediate_file): Add version to intermediate
6725 * doc/gcov.texi: Document new field 'version' in intermediate
6726 file format. Fix location of '-k' option of gcov command.
6728 2018-01-04 Martin Liska <mliska@suse.cz>
6731 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
6733 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6735 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
6737 2018-01-03 Martin Sebor <msebor@redhat.com>
6739 PR tree-optimization/83655
6740 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
6741 checking calls with invalid arguments.
6743 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6745 * tree-vect-stmts.c (vect_get_store_rhs): New function.
6746 (vectorizable_mask_load_store): Delete.
6747 (vectorizable_call): Return false for masked loads and stores.
6748 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
6749 instead of gimple_assign_rhs1.
6750 (vectorizable_load): Handle IFN_MASK_LOAD.
6751 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
6753 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6755 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
6757 (vectorizable_mask_load_store): ...here.
6758 (vectorizable_load): ...and here.
6760 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6762 * tree-vect-stmts.c (vect_build_all_ones_mask)
6763 (vect_build_zero_merge_argument): New functions, split out from...
6764 (vectorizable_load): ...here.
6766 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6768 * tree-vect-stmts.c (vect_check_store_rhs): New function,
6770 (vectorizable_mask_load_store): ...here.
6771 (vectorizable_store): ...and here.
6773 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6775 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
6777 (vectorizable_mask_load_store): ...here.
6779 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6781 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
6782 (vect_model_store_cost): Take a vec_load_store_type instead of a
6784 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
6785 (vect_model_store_cost): Take a vec_load_store_type instead of a
6787 (vectorizable_mask_load_store): Update accordingly.
6788 (vectorizable_store): Likewise.
6789 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
6791 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6793 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
6794 IFN_MASK_LOAD calls here rather than...
6795 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
6797 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6798 Alan Hayward <alan.hayward@arm.com>
6799 David Sherwood <david.sherwood@arm.com>
6801 * expmed.c (extract_bit_field_1): For vector extracts,
6802 fall back to extract_bit_field_as_subreg if vec_extract
6805 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6806 Alan Hayward <alan.hayward@arm.com>
6807 David Sherwood <david.sherwood@arm.com>
6809 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
6810 they are variable or constant sized.
6811 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
6812 slots for constant-sized data.
6814 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6815 Alan Hayward <alan.hayward@arm.com>
6816 David Sherwood <david.sherwood@arm.com>
6818 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
6819 handling COND_EXPRs with boolean comparisons, try to find a better
6820 basis for the mask type than the boolean itself.
6822 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6824 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
6825 is calculated and how it can be overridden.
6826 * genmodes.c (max_bitsize_mode_any_mode): New variable.
6827 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
6829 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
6832 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6833 Alan Hayward <alan.hayward@arm.com>
6834 David Sherwood <david.sherwood@arm.com>
6836 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
6837 Remove the mode argument.
6838 (aarch64_simd_valid_immediate): Remove the mode and inverse
6840 * config/aarch64/iterators.md (bitsize): New iterator.
6841 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
6842 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
6843 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
6844 aarch64_simd_valid_immediate.
6845 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
6846 (aarch64_reg_or_bic_imm): Likewise.
6847 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
6848 with an insn_type enum and msl with a modifier_type enum.
6849 Replace element_width with a scalar_mode. Change the shift
6850 to unsigned int. Add constructors for scalar_float_mode and
6851 scalar_int_mode elements.
6852 (aarch64_vect_float_const_representable_p): Delete.
6853 (aarch64_can_const_movi_rtx_p)
6854 (aarch64_simd_scalar_immediate_valid_for_move)
6855 (aarch64_simd_make_constant): Update call to
6856 aarch64_simd_valid_immediate.
6857 (aarch64_advsimd_valid_immediate_hs): New function.
6858 (aarch64_advsimd_valid_immediate): Likewise.
6859 (aarch64_simd_valid_immediate): Remove mode and inverse
6860 arguments. Rewrite to use the above. Use const_vec_duplicate_p
6861 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
6862 and aarch64_float_const_representable_p on the result.
6863 (aarch64_output_simd_mov_immediate): Remove mode argument.
6864 Update call to aarch64_simd_valid_immediate and use of
6865 simd_immediate_info.
6866 (aarch64_output_scalar_simd_mov_immediate): Update call
6869 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6870 Alan Hayward <alan.hayward@arm.com>
6871 David Sherwood <david.sherwood@arm.com>
6873 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
6874 (mode_nunits): Likewise CONST_MODE_NUNITS.
6875 * machmode.def (ADJUST_NUNITS): Document.
6876 * genmodes.c (mode_data::need_nunits_adj): New field.
6877 (blank_mode): Update accordingly.
6878 (adj_nunits): New variable.
6879 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
6881 (emit_mode_size_inline): Set need_bytesize_adj for all modes
6882 listed in adj_nunits.
6883 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
6884 listed in adj_nunits. Don't emit case statements for such modes.
6885 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
6886 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
6887 nothing if adj_nunits is nonnull.
6888 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
6889 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
6890 (emit_mode_fbit): Update use of print_maybe_const_decl.
6891 (emit_move_size): Likewise. Treat the array as non-const
6893 (emit_mode_adjustments): Handle adj_nunits.
6895 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6897 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
6898 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
6899 (VECTOR_MODES): Use it.
6900 (make_vector_modes): Take the prefix as an argument.
6902 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6903 Alan Hayward <alan.hayward@arm.com>
6904 David Sherwood <david.sherwood@arm.com>
6906 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
6907 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
6908 for MODE_VECTOR_BOOL.
6909 * machmode.def (VECTOR_BOOL_MODE): Document.
6910 * genmodes.c (VECTOR_BOOL_MODE): New macro.
6911 (make_vector_bool_mode): New function.
6912 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
6914 * lto-streamer-in.c (lto_input_mode_table): Likewise.
6915 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
6917 * stor-layout.c (int_mode_for_mode): Likewise.
6918 * tree.c (build_vector_type_for_mode): Likewise.
6919 * varasm.c (output_constant_pool_2): Likewise.
6920 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
6921 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
6922 for MODE_VECTOR_BOOL.
6923 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
6924 of mode class checks.
6925 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
6926 instead of a list of mode class checks.
6927 (expand_vector_scalar_condition): Likewise.
6928 (type_for_widest_vector_mode): Handle BImode as an inner mode.
6930 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6931 Alan Hayward <alan.hayward@arm.com>
6932 David Sherwood <david.sherwood@arm.com>
6934 * machmode.h (mode_size): Change from unsigned short to
6936 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
6937 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6938 or if measurement_type is not polynomial.
6939 (fixed_size_mode::includes_p): Check for constant-sized modes.
6940 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
6941 return a poly_uint16 rather than an unsigned short.
6942 (emit_mode_size): Change the type of mode_size from unsigned short
6943 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
6944 (emit_mode_adjustments): Cope with polynomial vector sizes.
6945 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6947 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6949 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
6950 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
6951 * caller-save.c (setup_save_areas): Likewise.
6952 (replace_reg_with_saved_mem): Likewise.
6953 * calls.c (emit_library_call_value_1): Likewise.
6954 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
6955 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
6956 (gen_lowpart_for_combine): Likewise.
6957 * convert.c (convert_to_integer_1): Likewise.
6958 * cse.c (equiv_constant, cse_insn): Likewise.
6959 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
6960 (cselib_subst_to_values): Likewise.
6961 * dce.c (word_dce_process_block): Likewise.
6962 * df-problems.c (df_word_lr_mark_ref): Likewise.
6963 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
6964 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
6965 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
6966 (rtl_for_decl_location): Likewise.
6967 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
6968 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
6969 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
6970 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
6971 (expand_expr_real_1): Likewise.
6972 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
6973 (pad_below): Likewise.
6974 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6975 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
6976 * ira.c (get_subreg_tracking_sizes): Likewise.
6977 * ira-build.c (ira_create_allocno_objects): Likewise.
6978 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
6979 (ira_sort_regnos_for_alter_reg): Likewise.
6980 * ira-costs.c (record_operand_costs): Likewise.
6981 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
6982 (resolve_simple_move): Likewise.
6983 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
6984 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
6985 (lra_constraints): Likewise.
6986 (CONST_POOL_OK_P): Reject variable-sized modes.
6987 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
6988 (add_pseudo_to_slot, lra_spill): Likewise.
6989 * omp-low.c (omp_clause_aligned_alignment): Likewise.
6990 * optabs-query.c (get_best_extraction_insn): Likewise.
6991 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
6992 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
6993 (expand_mult_highpart, valid_multiword_target_p): Likewise.
6994 * recog.c (offsettable_address_addr_space_p): Likewise.
6995 * regcprop.c (maybe_mode_change): Likewise.
6996 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
6997 * regrename.c (build_def_use): Likewise.
6998 * regstat.c (dump_reg_info): Likewise.
6999 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
7000 (find_reloads, find_reloads_subreg_address): Likewise.
7001 * reload1.c (eliminate_regs_1): Likewise.
7002 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
7003 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
7004 (simplify_binary_operation_1, simplify_subreg): Likewise.
7005 * targhooks.c (default_function_arg_padding): Likewise.
7006 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
7007 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
7008 (verify_gimple_assign_ternary): Likewise.
7009 * tree-inline.c (estimate_move_cost): Likewise.
7010 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7011 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
7012 (get_address_cost_ainc): Likewise.
7013 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
7014 (vect_supportable_dr_alignment): Likewise.
7015 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7016 (vectorizable_reduction): Likewise.
7017 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
7018 (vectorizable_operation, vectorizable_load): Likewise.
7019 * tree.c (build_same_sized_truth_vector_type): Likewise.
7020 * valtrack.c (cleanup_auto_inc_dec): Likewise.
7021 * var-tracking.c (emit_note_insn_var_location): Likewise.
7022 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
7023 (ADDR_VEC_ALIGN): Likewise.
7025 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7026 Alan Hayward <alan.hayward@arm.com>
7027 David Sherwood <david.sherwood@arm.com>
7029 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
7031 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7032 or if measurement_type is polynomial.
7033 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
7034 * combine.c (make_extraction): Likewise.
7035 * dse.c (find_shift_sequence): Likewise.
7036 * dwarf2out.c (mem_loc_descriptor): Likewise.
7037 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
7038 (extract_bit_field, extract_low_bits): Likewise.
7039 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
7040 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
7041 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
7042 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
7043 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7044 * reload.c (find_reloads): Likewise.
7045 * reload1.c (alter_reg): Likewise.
7046 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
7047 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
7048 * tree-if-conv.c (predicate_mem_writes): Likewise.
7049 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
7050 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
7051 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
7052 * valtrack.c (dead_debug_insert_temp): Likewise.
7053 * varasm.c (mergeable_constant_section): Likewise.
7054 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
7056 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7057 Alan Hayward <alan.hayward@arm.com>
7058 David Sherwood <david.sherwood@arm.com>
7060 * expr.c (expand_assignment): Cope with polynomial mode sizes
7061 when assigning to a CONCAT.
7063 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7064 Alan Hayward <alan.hayward@arm.com>
7065 David Sherwood <david.sherwood@arm.com>
7067 * machmode.h (mode_precision): Change from unsigned short to
7069 (mode_to_precision): Return a poly_uint16 rather than an unsigned
7071 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
7072 or if measurement_type is not polynomial.
7073 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
7074 in which the mode is already known to be a scalar_int_mode.
7075 * genmodes.c (emit_mode_precision): Change the type of mode_precision
7076 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
7078 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7079 for GET_MODE_PRECISION.
7080 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7081 for GET_MODE_PRECISION.
7082 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
7084 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
7085 (expand_field_assignment, make_extraction): Likewise.
7086 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
7087 (get_last_value): Likewise.
7088 * convert.c (convert_to_integer_1): Likewise.
7089 * cse.c (cse_insn): Likewise.
7090 * expr.c (expand_expr_real_1): Likewise.
7091 * lra-constraints.c (simplify_operand_subreg): Likewise.
7092 * optabs-query.c (can_atomic_load_p): Likewise.
7093 * optabs.c (expand_atomic_load): Likewise.
7094 (expand_atomic_store): Likewise.
7095 * ree.c (combine_reaching_defs): Likewise.
7096 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
7097 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
7098 * tree.h (type_has_mode_precision_p): Likewise.
7099 * ubsan.c (instrument_si_overflow): Likewise.
7101 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7102 Alan Hayward <alan.hayward@arm.com>
7103 David Sherwood <david.sherwood@arm.com>
7105 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
7106 polynomial numbers of units.
7107 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
7108 (valid_vector_subparts_p): New function.
7109 (build_vector_type): Remove temporary shim and take the number
7110 of units as a poly_uint64 rather than an int.
7111 (build_opaque_vector_type): Take the number of units as a
7112 poly_uint64 rather than an int.
7113 * tree.c (build_vector_from_ctor): Handle polynomial
7114 TYPE_VECTOR_SUBPARTS.
7115 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
7116 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
7117 (build_vector_from_val): If the number of units is variable,
7118 use build_vec_duplicate_cst for constant operands and
7119 VEC_DUPLICATE_EXPR otherwise.
7120 (make_vector_type): Remove temporary is_constant ().
7121 (build_vector_type, build_opaque_vector_type): Take the number of
7122 units as a poly_uint64 rather than an int.
7123 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
7125 * cfgexpand.c (expand_debug_expr): Likewise.
7126 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
7127 (store_constructor, expand_expr_real_1): Likewise.
7128 (const_scalar_mask_from_tree): Likewise.
7129 * fold-const-call.c (fold_const_reduction): Likewise.
7130 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
7131 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
7132 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
7133 (fold_relational_const): Likewise.
7134 (native_interpret_vector): Likewise. Change the size from an
7135 int to an unsigned int.
7136 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
7137 TYPE_VECTOR_SUBPARTS.
7138 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
7139 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
7140 duplicating a non-constant operand into a variable-length vector.
7141 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
7142 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
7143 * ipa-icf.c (sem_variable::equals): Likewise.
7144 * match.pd: Likewise.
7145 * omp-simd-clone.c (simd_clone_subparts): Likewise.
7146 * print-tree.c (print_node): Likewise.
7147 * stor-layout.c (layout_type): Likewise.
7148 * targhooks.c (default_builtin_vectorization_cost): Likewise.
7149 * tree-cfg.c (verify_gimple_comparison): Likewise.
7150 (verify_gimple_assign_binary): Likewise.
7151 (verify_gimple_assign_ternary): Likewise.
7152 (verify_gimple_assign_single): Likewise.
7153 * tree-pretty-print.c (dump_generic_node): Likewise.
7154 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7155 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
7156 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
7157 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
7158 (vect_shift_permute_load_chain): Likewise.
7159 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
7160 (expand_vector_condition, optimize_vector_constructor): Likewise.
7161 (lower_vec_perm, get_compute_type): Likewise.
7162 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7163 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
7164 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
7165 (vect_recog_mask_conversion_pattern): Likewise.
7166 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
7167 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
7168 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7169 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
7170 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
7171 (vectorizable_shift, vectorizable_operation, vectorizable_store)
7172 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
7173 (supportable_widening_operation): Likewise.
7174 (supportable_narrowing_operation): Likewise.
7175 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
7177 * varasm.c (output_constant): Likewise.
7179 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7180 Alan Hayward <alan.hayward@arm.com>
7181 David Sherwood <david.sherwood@arm.com>
7183 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
7184 so that both the length == 3 and length != 3 cases set up their
7185 own permute vectors. Add comments explaining why we know the
7186 number of elements is constant.
7187 (vect_permute_load_chain): Likewise.
7189 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7190 Alan Hayward <alan.hayward@arm.com>
7191 David Sherwood <david.sherwood@arm.com>
7193 * machmode.h (mode_nunits): Change from unsigned char to
7195 (ONLY_FIXED_SIZE_MODES): New macro.
7196 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
7197 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
7198 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
7200 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
7201 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
7202 or if measurement_type is not polynomial.
7203 * genmodes.c (ZERO_COEFFS): New macro.
7204 (emit_mode_nunits_inline): Make mode_nunits_inline return a
7206 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
7207 Use ZERO_COEFFS when emitting initializers.
7208 * data-streamer.h (bp_pack_poly_value): New function.
7209 (bp_unpack_poly_value): Likewise.
7210 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7211 for GET_MODE_NUNITS.
7212 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7213 for GET_MODE_NUNITS.
7214 * tree.c (make_vector_type): Remove temporary shim and make
7215 the real function take the number of units as a poly_uint64
7217 (build_vector_type_for_mode): Handle polynomial nunits.
7218 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
7219 * emit-rtl.c (const_vec_series_p_1): Likewise.
7220 (gen_rtx_CONST_VECTOR): Likewise.
7221 * fold-const.c (test_vec_duplicate_folding): Likewise.
7222 * genrecog.c (validate_pattern): Likewise.
7223 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
7224 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7225 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
7226 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
7227 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
7228 * rtlanal.c (subreg_get_info): Likewise.
7229 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7230 (vect_grouped_load_supported): Likewise.
7231 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
7232 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
7233 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
7234 (simplify_const_unary_operation, simplify_binary_operation_1)
7235 (simplify_const_binary_operation, simplify_ternary_operation)
7236 (test_vector_ops_duplicate, test_vector_ops): Likewise.
7237 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
7238 instead of CONST_VECTOR_NUNITS.
7239 * varasm.c (output_constant_pool_2): Likewise.
7240 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
7241 explicit-encoded elements in the XVEC for variable-length vectors.
7243 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7245 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
7247 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7248 Alan Hayward <alan.hayward@arm.com>
7249 David Sherwood <david.sherwood@arm.com>
7251 * coretypes.h (fixed_size_mode): Declare.
7252 (fixed_size_mode_pod): New typedef.
7253 * builtins.h (target_builtins::x_apply_args_mode)
7254 (target_builtins::x_apply_result_mode): Change type to
7255 fixed_size_mode_pod.
7256 * builtins.c (apply_args_size, apply_result_size, result_vector)
7257 (expand_builtin_apply_args_1, expand_builtin_apply)
7258 (expand_builtin_return): Update accordingly.
7260 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7262 * cse.c (hash_rtx_cb): Hash only the encoded elements.
7263 * cselib.c (cselib_hash_rtx): Likewise.
7264 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
7265 CONST_VECTOR encoding.
7267 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7268 Jeff Law <law@redhat.com>
7271 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
7272 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
7273 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
7274 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
7277 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
7278 explicitly probe *sp in a noreturn function if there were any callee
7279 register saves or frame pointer is needed.
7281 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7284 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
7285 BLKmode for ternary, binary or unary expressions.
7288 * var-tracking.c (delete_vta_debug_insn): New inline function.
7289 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
7290 insns from get_insns () to NULL instead of each bb separately.
7291 Use delete_vta_debug_insn. No longer static.
7292 (vt_debug_insns_local, variable_tracking_main_1): Adjust
7293 delete_vta_debug_insns callers.
7294 * rtl.h (delete_vta_debug_insns): Declare.
7295 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
7296 instead of variable_tracking_main.
7298 2018-01-03 Martin Sebor <msebor@redhat.com>
7300 PR tree-optimization/83603
7301 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
7302 arguments past the endof the argument list in functions declared
7303 without a prototype.
7304 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
7305 Avoid checking when arguments are null.
7307 2018-01-03 Martin Sebor <msebor@redhat.com>
7310 * doc/extend.texi (attribute const): Fix a typo.
7311 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
7312 issuing -Wsuggest-attribute for void functions.
7314 2018-01-03 Martin Sebor <msebor@redhat.com>
7316 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
7317 offset_int::from instead of wide_int::to_shwi.
7318 (maybe_diag_overlap): Remove assertion.
7319 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
7320 * gimple-ssa-sprintf.c (format_directive): Same.
7321 (parse_directive): Same.
7322 (sprintf_dom_walker::compute_format_length): Same.
7323 (try_substitute_return_value): Same.
7325 2018-01-03 Jeff Law <law@redhat.com>
7328 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
7329 non-constant residual for zero at runtime and avoid probing in
7330 that case. Reorganize code for trailing problem to mirror handling
7333 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7335 PR tree-optimization/83501
7336 * tree-ssa-strlen.c (get_string_cst): New.
7337 (handle_char_store): Call get_string_cst.
7339 2018-01-03 Martin Liska <mliska@suse.cz>
7341 PR tree-optimization/83593
7342 * tree-ssa-strlen.c: Include tree-cfg.h.
7343 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
7344 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
7345 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
7347 (strlen_dom_walker::before_dom_children): Call
7348 gimple_purge_dead_eh_edges. Dump tranformation with details
7350 (strlen_dom_walker::before_dom_children): Update call by adding
7351 new argument cleanup_eh.
7352 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
7354 2018-01-03 Martin Liska <mliska@suse.cz>
7357 * cif-code.def (VARIADIC_THUNK): New enum value.
7358 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
7361 2018-01-03 Jan Beulich <jbeulich@suse.com>
7363 * sse.md (mov<mode>_internal): Tighten condition for when to use
7364 vmovdqu<ssescalarsize> for TI and OI modes.
7366 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7368 Update copyright years.
7370 2018-01-03 Martin Liska <mliska@suse.cz>
7373 * ipa-visibility.c (function_and_variable_visibility): Skip
7374 functions with noipa attribure.
7376 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7378 * gcc.c (process_command): Update copyright notice dates.
7379 * gcov-dump.c (print_version): Ditto.
7380 * gcov.c (print_version): Ditto.
7381 * gcov-tool.c (print_version): Ditto.
7382 * gengtype.c (create_file): Ditto.
7383 * doc/cpp.texi: Bump @copying's copyright year.
7384 * doc/cppinternals.texi: Ditto.
7385 * doc/gcc.texi: Ditto.
7386 * doc/gccint.texi: Ditto.
7387 * doc/gcov.texi: Ditto.
7388 * doc/install.texi: Ditto.
7389 * doc/invoke.texi: Ditto.
7391 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7393 * vector-builder.h (vector_builder::m_full_nelts): Change from
7394 unsigned int to poly_uint64.
7395 (vector_builder::full_nelts): Update prototype accordingly.
7396 (vector_builder::new_vector): Likewise.
7397 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
7398 (vector_builder::operator ==): Likewise.
7399 (vector_builder::finalize): Likewise.
7400 * int-vector-builder.h (int_vector_builder::int_vector_builder):
7401 Take the number of elements as a poly_uint64 rather than an
7403 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
7404 from unsigned int to poly_uint64.
7405 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
7406 (vec_perm_indices::new_vector): Likewise.
7407 (vec_perm_indices::length): Likewise.
7408 (vec_perm_indices::nelts_per_input): Likewise.
7409 (vec_perm_indices::input_nelts): Likewise.
7410 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
7411 number of elements per input as a poly_uint64 rather than an
7412 unsigned int. Use the original encoding for variable-length
7413 vectors, rather than clamping each individual element.
7414 For the second and subsequent elements in each pattern,
7415 clamp the step and base before clamping their sum.
7416 (vec_perm_indices::series_p): Handle polynomial element counts.
7417 (vec_perm_indices::all_in_range_p): Likewise.
7418 (vec_perm_indices_to_tree): Likewise.
7419 (vec_perm_indices_to_rtx): Likewise.
7420 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
7421 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
7422 (tree_vector_builder::new_binary_operation): Handle polynomial
7423 element counts. Return false if we need to know the number
7424 of elements at compile time.
7425 * fold-const.c (fold_vec_perm): Punt if the number of elements
7426 isn't known at compile time.
7428 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7430 * vec-perm-indices.h (vec_perm_builder): Change element type
7431 from HOST_WIDE_INT to poly_int64.
7432 (vec_perm_indices::element_type): Update accordingly.
7433 (vec_perm_indices::clamp): Handle polynomial element_types.
7434 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7435 (vec_perm_indices::all_in_range_p): Likewise.
7436 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
7438 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
7439 polynomial vec_perm_indices element types.
7440 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
7441 * fold-const.c (fold_vec_perm): Likewise.
7442 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
7443 * tree-vect-generic.c (lower_vec_perm): Likewise.
7444 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7445 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
7446 element type to HOST_WIDE_INT.
7448 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7449 Alan Hayward <alan.hayward@arm.com>
7450 David Sherwood <david.sherwood@arm.com>
7452 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
7453 rather than an int. Use plus_constant.
7454 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
7455 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
7457 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7458 Alan Hayward <alan.hayward@arm.com>
7459 David Sherwood <david.sherwood@arm.com>
7461 * calls.c (emit_call_1, expand_call): Change struct_value_size from
7462 a HOST_WIDE_INT to a poly_int64.
7464 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7465 Alan Hayward <alan.hayward@arm.com>
7466 David Sherwood <david.sherwood@arm.com>
7468 * calls.c (load_register_parameters): Cope with polynomial
7469 mode sizes. Require a constant size for BLKmode parameters
7470 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
7471 forces a parameter to be padded at the lsb end in order to
7472 fill a complete number of words, require the parameter size
7473 to be ordered wrt UNITS_PER_WORD.
7475 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7476 Alan Hayward <alan.hayward@arm.com>
7477 David Sherwood <david.sherwood@arm.com>
7479 * reload1.c (spill_stack_slot_width): Change element type
7480 from unsigned int to poly_uint64_pod.
7481 (alter_reg): Treat mode sizes as polynomial.
7483 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7484 Alan Hayward <alan.hayward@arm.com>
7485 David Sherwood <david.sherwood@arm.com>
7487 * reload.c (complex_word_subreg_p): New function.
7488 (reload_inner_reg_of_subreg, push_reload): Use it.
7490 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7491 Alan Hayward <alan.hayward@arm.com>
7492 David Sherwood <david.sherwood@arm.com>
7494 * lra-constraints.c (process_alt_operands): Reject matched
7495 operands whose sizes aren't ordered.
7496 (match_reload): Refer to this check here.
7498 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7499 Alan Hayward <alan.hayward@arm.com>
7500 David Sherwood <david.sherwood@arm.com>
7502 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
7503 that the mode size is in the set {1, 2, 4, 8, 16}.
7505 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7506 Alan Hayward <alan.hayward@arm.com>
7507 David Sherwood <david.sherwood@arm.com>
7509 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
7510 Use plus_constant instead of gen_rtx_PLUS.
7512 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7513 Alan Hayward <alan.hayward@arm.com>
7514 David Sherwood <david.sherwood@arm.com>
7516 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
7517 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
7518 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
7519 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
7520 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
7521 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
7522 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
7523 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
7524 * config/i386/i386.c (ix86_push_rounding): ...this new function.
7525 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
7527 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
7528 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
7529 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
7530 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
7531 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
7532 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
7533 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
7534 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
7535 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
7536 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
7538 * expr.c (emit_move_resolve_push): Treat the input and result
7539 of PUSH_ROUNDING as a poly_int64.
7540 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
7541 (emit_push_insn): Likewise.
7542 * lra-eliminations.c (mark_not_eliminable): Likewise.
7543 * recog.c (push_operand): Likewise.
7544 * reload1.c (elimination_effects): Likewise.
7545 * rtlanal.c (nonzero_bits1): Likewise.
7546 * calls.c (store_one_arg): Likewise. Require the padding to be
7547 known at compile time.
7549 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7550 Alan Hayward <alan.hayward@arm.com>
7551 David Sherwood <david.sherwood@arm.com>
7553 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
7554 Use plus_constant instead of gen_rtx_PLUS.
7556 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7557 Alan Hayward <alan.hayward@arm.com>
7558 David Sherwood <david.sherwood@arm.com>
7560 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
7563 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7564 Alan Hayward <alan.hayward@arm.com>
7565 David Sherwood <david.sherwood@arm.com>
7567 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
7568 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
7569 via stack temporaries. Treat the mode size as polynomial too.
7571 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7572 Alan Hayward <alan.hayward@arm.com>
7573 David Sherwood <david.sherwood@arm.com>
7575 * expr.c (expand_expr_real_2): When handling conversions involving
7576 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
7577 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
7578 as a poly_uint64 too.
7580 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7581 Alan Hayward <alan.hayward@arm.com>
7582 David Sherwood <david.sherwood@arm.com>
7584 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
7586 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7587 Alan Hayward <alan.hayward@arm.com>
7588 David Sherwood <david.sherwood@arm.com>
7590 * combine.c (can_change_dest_mode): Handle polynomial
7591 REGMODE_NATURAL_SIZE.
7592 * expmed.c (store_bit_field_1): Likewise.
7593 * expr.c (store_constructor): Likewise.
7594 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
7595 and polynomial REGMODE_NATURAL_SIZE.
7596 (gen_lowpart_common): Likewise.
7597 * reginfo.c (record_subregs_of_mode): Likewise.
7598 * rtlanal.c (read_modify_subreg_p): Likewise.
7600 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7601 Alan Hayward <alan.hayward@arm.com>
7602 David Sherwood <david.sherwood@arm.com>
7604 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
7605 numbers of elements.
7607 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7608 Alan Hayward <alan.hayward@arm.com>
7609 David Sherwood <david.sherwood@arm.com>
7611 * match.pd: Cope with polynomial numbers of vector elements.
7613 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7614 Alan Hayward <alan.hayward@arm.com>
7615 David Sherwood <david.sherwood@arm.com>
7617 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
7618 in a POINTER_PLUS_EXPR.
7620 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7621 Alan Hayward <alan.hayward@arm.com>
7622 David Sherwood <david.sherwood@arm.com>
7624 * omp-simd-clone.c (simd_clone_subparts): New function.
7625 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
7626 (ipa_simd_modify_function_body): Likewise.
7628 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7629 Alan Hayward <alan.hayward@arm.com>
7630 David Sherwood <david.sherwood@arm.com>
7632 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
7633 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
7634 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
7635 (expand_vector_condition, vector_element): Likewise.
7636 (subparts_gt): New function.
7637 (get_compute_type): Use subparts_gt.
7638 (count_type_subparts): Delete.
7639 (expand_vector_operations_1): Use subparts_gt instead of
7640 count_type_subparts.
7642 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7643 Alan Hayward <alan.hayward@arm.com>
7644 David Sherwood <david.sherwood@arm.com>
7646 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
7647 (vect_compile_time_alias): ...this new function. Do the calculation
7648 on poly_ints rather than trees.
7649 (vect_prune_runtime_alias_test_list): Update call accordingly.
7651 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7652 Alan Hayward <alan.hayward@arm.com>
7653 David Sherwood <david.sherwood@arm.com>
7655 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
7657 (vect_schedule_slp_instance): Likewise.
7659 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7660 Alan Hayward <alan.hayward@arm.com>
7661 David Sherwood <david.sherwood@arm.com>
7663 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
7664 constant and extern definitions for variable-length vectors.
7665 (vect_get_constant_vectors): Note that the number of units
7666 is known to be constant.
7668 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7669 Alan Hayward <alan.hayward@arm.com>
7670 David Sherwood <david.sherwood@arm.com>
7672 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
7673 of units as polynomial. Choose between WIDE and NARROW based
7676 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7677 Alan Hayward <alan.hayward@arm.com>
7678 David Sherwood <david.sherwood@arm.com>
7680 * tree-vect-stmts.c (simd_clone_subparts): New function.
7681 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
7683 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7684 Alan Hayward <alan.hayward@arm.com>
7685 David Sherwood <david.sherwood@arm.com>
7687 * tree-vect-stmts.c (vectorizable_call): Treat the number of
7688 vectors as polynomial. Use build_index_vector for
7691 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7692 Alan Hayward <alan.hayward@arm.com>
7693 David Sherwood <david.sherwood@arm.com>
7695 * tree-vect-stmts.c (get_load_store_type): Treat the number of
7696 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
7697 for variable-length vectors.
7698 (vectorizable_mask_load_store): Treat the number of units as
7699 polynomial, asserting that it is constant if the condition has
7700 already been enforced.
7701 (vectorizable_store, vectorizable_load): Likewise.
7703 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7704 Alan Hayward <alan.hayward@arm.com>
7705 David Sherwood <david.sherwood@arm.com>
7707 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
7708 of units as polynomial. Punt if we can't tell at compile time
7709 which vector contains the final result.
7711 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7712 Alan Hayward <alan.hayward@arm.com>
7713 David Sherwood <david.sherwood@arm.com>
7715 * tree-vect-loop.c (vectorizable_induction): Treat the number
7716 of units as polynomial. Punt on SLP inductions. Use an integer
7717 VEC_SERIES_EXPR for variable-length integer reductions. Use a
7718 cast of such a series for variable-length floating-point
7721 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7722 Alan Hayward <alan.hayward@arm.com>
7723 David Sherwood <david.sherwood@arm.com>
7725 * tree.h (build_index_vector): Declare.
7726 * tree.c (build_index_vector): New function.
7727 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
7728 of units as polynomial, forcibly converting it to a constant if
7729 vectorizable_reduction has already enforced the condition.
7730 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
7731 to create a {1,2,3,...} vector.
7732 (vectorizable_reduction): Treat the number of units as polynomial.
7733 Choose vectype_in based on the largest scalar element size rather
7734 than the smallest number of units. Enforce the restrictions
7737 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7738 Alan Hayward <alan.hayward@arm.com>
7739 David Sherwood <david.sherwood@arm.com>
7741 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
7742 number of units as polynomial.
7744 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7745 Alan Hayward <alan.hayward@arm.com>
7746 David Sherwood <david.sherwood@arm.com>
7748 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
7749 * target.def (autovectorize_vector_sizes): Return the vector sizes
7750 by pointer, using vector_sizes rather than a bitmask.
7751 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
7752 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
7753 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
7755 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
7756 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
7757 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
7758 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
7759 * omp-general.c (omp_max_vf): Likewise.
7760 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7761 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
7762 * tree-vect-loop.c (vect_analyze_loop): Likewise.
7763 * tree-vect-slp.c (vect_slp_bb): Likewise.
7764 * doc/tm.texi: Regenerate.
7765 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
7767 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
7768 the vector size as a poly_uint64 rather than an unsigned int.
7769 (current_vector_size): Change from an unsigned int to a poly_uint64.
7770 (get_vectype_for_scalar_type): Update accordingly.
7771 * tree.h (build_truth_vector_type): Take the size and number of
7772 units as a poly_uint64 rather than an unsigned int.
7773 (build_vector_type): Add a temporary overload that takes
7774 the number of units as a poly_uint64 rather than an unsigned int.
7775 * tree.c (make_vector_type): Likewise.
7776 (build_truth_vector_type): Take the number of units as a poly_uint64
7777 rather than an unsigned int.
7779 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7780 Alan Hayward <alan.hayward@arm.com>
7781 David Sherwood <david.sherwood@arm.com>
7783 * target.def (get_mask_mode): Take the number of units and length
7784 as poly_uint64s rather than unsigned ints.
7785 * targhooks.h (default_get_mask_mode): Update accordingly.
7786 * targhooks.c (default_get_mask_mode): Likewise.
7787 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
7788 * doc/tm.texi: Regenerate.
7790 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7791 Alan Hayward <alan.hayward@arm.com>
7792 David Sherwood <david.sherwood@arm.com>
7794 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
7795 * omp-general.c (omp_max_vf): Likewise.
7796 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
7797 (expand_omp_simd): Handle polynomial safelen.
7798 * omp-low.c (omplow_simd_context): Add a default constructor.
7799 (omplow_simd_context::max_vf): Change from int to poly_uint64.
7800 (lower_rec_simd_input_clauses): Update accordingly.
7801 (lower_rec_input_clauses): Likewise.
7803 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7804 Alan Hayward <alan.hayward@arm.com>
7805 David Sherwood <david.sherwood@arm.com>
7807 * tree-vectorizer.h (vect_nunits_for_cost): New function.
7808 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
7809 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
7810 (vect_analyze_slp_cost): Likewise.
7811 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
7812 (vect_model_load_cost): Likewise.
7814 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7815 Alan Hayward <alan.hayward@arm.com>
7816 David Sherwood <david.sherwood@arm.com>
7818 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
7819 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
7820 from an unsigned int * to a poly_uint64_pod *.
7821 (calculate_unrolling_factor): New function.
7822 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
7824 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7825 Alan Hayward <alan.hayward@arm.com>
7826 David Sherwood <david.sherwood@arm.com>
7828 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
7829 from an unsigned int to a poly_uint64.
7830 (_loop_vec_info::slp_unrolling_factor): Likewise.
7831 (_loop_vec_info::vectorization_factor): Change from an int
7833 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
7834 (vect_get_num_vectors): New function.
7835 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
7836 (vect_get_num_copies): Use vect_get_num_vectors.
7837 (vect_analyze_data_ref_dependences): Change max_vf from an int *
7838 to an unsigned int *.
7839 (vect_analyze_data_refs): Change min_vf from an int * to a
7841 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7842 than an unsigned HOST_WIDE_INT.
7843 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
7844 (vect_analyze_data_ref_dependence): Change max_vf from an int *
7845 to an unsigned int *.
7846 (vect_analyze_data_ref_dependences): Likewise.
7847 (vect_compute_data_ref_alignment): Handle polynomial vf.
7848 (vect_enhance_data_refs_alignment): Likewise.
7849 (vect_prune_runtime_alias_test_list): Likewise.
7850 (vect_shift_permute_load_chain): Likewise.
7851 (vect_supportable_dr_alignment): Likewise.
7852 (dependence_distance_ge_vf): Take the vectorization factor as a
7853 poly_uint64 rather than an unsigned HOST_WIDE_INT.
7854 (vect_analyze_data_refs): Change min_vf from an int * to a
7856 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
7857 vfm1 as a poly_uint64 rather than an int. Make the same change
7858 for the returned bound_scalar.
7859 (vect_gen_vector_loop_niters): Handle polynomial vf.
7860 (vect_do_peeling): Likewise. Update call to
7861 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
7862 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
7864 * tree-vect-loop.c (vect_determine_vectorization_factor)
7865 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
7866 (vect_get_known_peeling_cost): Likewise.
7867 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
7868 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
7869 (vect_transform_loop): Likewise. Use the lowest possible VF when
7870 updating the upper bounds of the loop.
7871 (vect_min_worthwhile_factor): Make static. Return an unsigned int
7873 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
7874 polynomial unroll factors.
7875 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
7876 (vect_make_slp_decision): Likewise.
7877 (vect_supported_load_permutation_p): Likewise, and polynomial
7879 (vect_analyze_slp_cost): Handle polynomial vf.
7880 (vect_slp_analyze_node_operations): Likewise.
7881 (vect_slp_analyze_bb_1): Likewise.
7882 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7883 than an unsigned HOST_WIDE_INT.
7884 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
7885 (vectorizable_load): Handle polynomial vf.
7886 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
7888 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
7890 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7891 Alan Hayward <alan.hayward@arm.com>
7892 David Sherwood <david.sherwood@arm.com>
7894 * match.pd: Handle bit operations involving three constants
7895 and try to fold one pair.
7897 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7899 * tree-vect-loop-manip.c: Include gimple-fold.h.
7900 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
7901 niters_maybe_zero parameters. Handle other cases besides a step of 1.
7902 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
7903 Add a path that uses a step of VF instead of 1, but disable it
7905 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
7906 and niters_no_overflow parameters. Update calls to
7907 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
7908 Create a new SSA name if the latter choses to use a ste other
7909 than zero, and return it via niters_vector_mult_vf_var.
7910 * tree-vect-loop.c (vect_transform_loop): Update calls to
7911 vect_do_peeling, vect_gen_vector_loop_niters and
7912 slpeel_make_loop_iterate_ntimes.
7913 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
7914 (vect_gen_vector_loop_niters): Update declarations after above changes.
7916 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
7918 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
7919 128-bit round to integer instructions.
7920 (ceil<mode>2): Likewise.
7921 (btrunc<mode>2): Likewise.
7922 (round<mode>2): Likewise.
7924 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7926 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
7927 unaligned VSX load/store on P8/P9.
7928 (expand_block_clear): Allow the use of unaligned VSX
7929 load/store on P8/P9.
7931 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7933 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
7935 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
7936 swap associated with both a load and a store.
7938 2018-01-02 Andrew Waterman <andrew@sifive.com>
7940 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
7941 * config/riscv/riscv.md (clear_cache): Use it.
7943 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
7945 * web.c: Remove out-of-date comment.
7947 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7949 * expr.c (fixup_args_size_notes): Check that any existing
7950 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
7951 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
7952 (emit_single_push_insn): ...here.
7954 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7956 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
7957 (const_vector_encoded_nelts): New function.
7958 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
7959 (const_vector_int_elt, const_vector_elt): Declare.
7960 * emit-rtl.c (const_vector_int_elt_1): New function.
7961 (const_vector_elt): Likewise.
7962 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
7963 of CONST_VECTOR_ELT.
7965 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7967 * expr.c: Include rtx-vector-builder.h.
7968 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
7969 directly on the tree encoding.
7970 (const_vector_from_tree): Likewise.
7971 * optabs.c: Include rtx-vector-builder.h.
7972 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
7973 sequence of "u" values.
7974 * vec-perm-indices.c: Include rtx-vector-builder.h.
7975 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
7976 directly on the vec_perm_indices encoding.
7978 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7980 * doc/rtl.texi (const_vector): Describe new encoding scheme.
7981 * Makefile.in (OBJS): Add rtx-vector-builder.o.
7982 * rtx-vector-builder.h: New file.
7983 * rtx-vector-builder.c: Likewise.
7984 * rtl.h (rtx_def::u2): Add a const_vector field.
7985 (CONST_VECTOR_NPATTERNS): New macro.
7986 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
7987 (CONST_VECTOR_DUPLICATE_P): Likewise.
7988 (CONST_VECTOR_STEPPED_P): Likewise.
7989 (CONST_VECTOR_ENCODED_ELT): Likewise.
7990 (const_vec_duplicate_p): Check for a duplicated vector encoding.
7991 (unwrap_const_vec_duplicate): Likewise.
7992 (const_vec_series_p): Check for a non-duplicated vector encoding.
7993 Say that the function only returns true for integer vectors.
7994 * emit-rtl.c: Include rtx-vector-builder.h.
7995 (gen_const_vec_duplicate_1): Delete.
7996 (gen_const_vector): Call gen_const_vec_duplicate instead of
7997 gen_const_vec_duplicate_1.
7998 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
7999 (gen_const_vec_duplicate): Use rtx_vector_builder.
8000 (gen_const_vec_series): Likewise.
8001 (gen_rtx_CONST_VECTOR): Likewise.
8002 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
8003 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
8004 Build a new vector rather than modifying a CONST_VECTOR in-place.
8005 (handle_special_swappables): Update call accordingly.
8006 * config/rs6000/rs6000-p8swap.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.
8011 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8013 * simplify-rtx.c (simplify_const_binary_operation): Use
8014 CONST_VECTOR_ELT instead of XVECEXP.
8016 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8018 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
8019 the selector elements to be different from the data elements
8020 if the selector is a VECTOR_CST.
8021 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
8022 ssizetype for the selector.
8024 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8026 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
8027 before testing each element individually.
8028 * tree-vect-generic.c (lower_vec_perm): Likewise.
8030 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8032 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
8033 * selftest-run-tests.c (selftest::run_tests): Call it.
8034 * vector-builder.h (vector_builder::operator ==): New function.
8035 (vector_builder::operator !=): Likewise.
8036 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
8037 (vec_perm_indices::all_from_input_p): New function.
8038 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8039 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
8040 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
8041 instead of reading the VECTOR_CST directly. Detect whether both
8042 vector inputs are the same before constructing the vec_perm_indices,
8043 and update the number of inputs argument accordingly. Use the
8044 utility functions added above. Only construct sel2 if we need to.
8046 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8048 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
8049 the broadcast of the low byte.
8050 (expand_mult_highpart): Use an explicit encoding for the permutes.
8051 * optabs-query.c (can_mult_highpart_p): Likewise.
8052 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
8053 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8054 (vectorizable_bswap): Likewise.
8055 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
8056 explicit encoding for the power-of-2 permutes.
8057 (vect_permute_store_chain): Likewise.
8058 (vect_grouped_load_supported): Likewise.
8059 (vect_permute_load_chain): Likewise.
8061 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8063 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
8064 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
8065 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
8066 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8067 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
8068 (vect_gen_perm_mask_any): Likewise.
8070 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8072 * int-vector-builder.h: New file.
8073 * vec-perm-indices.h: Include int-vector-builder.h.
8074 (vec_perm_indices): Redefine as an int_vector_builder.
8075 (auto_vec_perm_indices): Delete.
8076 (vec_perm_builder): Redefine as a stand-alone class.
8077 (vec_perm_indices::vec_perm_indices): New function.
8078 (vec_perm_indices::clamp): Likewise.
8079 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
8080 (vec_perm_indices::new_vector): New function.
8081 (vec_perm_indices::new_expanded_vector): Update for new
8082 vec_perm_indices class.
8083 (vec_perm_indices::rotate_inputs): New function.
8084 (vec_perm_indices::all_in_range_p): Operate directly on the
8085 encoded form, without computing elided elements.
8086 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
8087 encoding. Update for new vec_perm_indices class.
8088 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
8089 the given vec_perm_builder.
8090 (expand_vec_perm_var): Update vec_perm_builder constructor.
8091 (expand_mult_highpart): Use vec_perm_builder instead of
8092 auto_vec_perm_indices.
8093 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
8094 vec_perm_indices instead of auto_vec_perm_indices. Use a single
8095 or double series encoding as appropriate.
8096 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
8097 vec_perm_indices instead of auto_vec_perm_indices.
8098 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8099 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8100 (vect_permute_store_chain): Likewise.
8101 (vect_grouped_load_supported): Likewise.
8102 (vect_permute_load_chain): Likewise.
8103 (vect_shift_permute_load_chain): Likewise.
8104 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8105 (vect_transform_slp_perm_load): Likewise.
8106 (vect_schedule_slp_instance): Likewise.
8107 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8108 (vectorizable_mask_load_store): Likewise.
8109 (vectorizable_bswap): Likewise.
8110 (vectorizable_store): Likewise.
8111 (vectorizable_load): Likewise.
8112 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
8113 vec_perm_indices instead of auto_vec_perm_indices. Use
8114 tree_to_vec_perm_builder to read the vector from a tree.
8115 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
8116 vec_perm_builder instead of a vec_perm_indices.
8117 (have_whole_vector_shift): Use vec_perm_builder and
8118 vec_perm_indices instead of auto_vec_perm_indices. Leave the
8119 truncation to calc_vec_perm_mask_for_shift.
8120 (vect_create_epilog_for_reduction): Likewise.
8121 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
8122 from auto_vec_perm_indices to vec_perm_indices.
8123 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8124 instead of changing individual elements.
8125 (aarch64_vectorize_vec_perm_const): Use new_vector to install
8126 the vector in d.perm.
8127 * config/arm/arm.c (expand_vec_perm_d::perm): Change
8128 from auto_vec_perm_indices to vec_perm_indices.
8129 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8130 instead of changing individual elements.
8131 (arm_vectorize_vec_perm_const): Use new_vector to install
8132 the vector in d.perm.
8133 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
8134 Update vec_perm_builder constructor.
8135 (rs6000_expand_interleave): Likewise.
8136 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
8137 (rs6000_expand_interleave): Likewise.
8139 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8141 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
8142 to qimode could truncate the indices.
8143 * optabs.c (expand_vec_perm_var): Likewise.
8145 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8147 * Makefile.in (OBJS): Add vec-perm-indices.o.
8148 * vec-perm-indices.h: New file.
8149 * vec-perm-indices.c: Likewise.
8150 * target.h (vec_perm_indices): Replace with a forward class
8152 (auto_vec_perm_indices): Move to vec-perm-indices.h.
8153 * optabs.h: Include vec-perm-indices.h.
8154 (expand_vec_perm): Delete.
8155 (selector_fits_mode_p, expand_vec_perm_var): Declare.
8156 (expand_vec_perm_const): Declare.
8157 * target.def (vec_perm_const_ok): Replace with...
8158 (vec_perm_const): ...this new hook.
8159 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
8160 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
8161 * doc/tm.texi: Regenerate.
8162 * optabs.def (vec_perm_const): Delete.
8163 * doc/md.texi (vec_perm_const): Likewise.
8164 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
8165 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
8166 expand_vec_perm for constant permutation vectors. Assert that
8167 the mode of variable permutation vectors is the integer equivalent
8168 of the mode that is being permuted.
8169 * optabs-query.h (selector_fits_mode_p): Declare.
8170 * optabs-query.c: Include vec-perm-indices.h.
8171 (selector_fits_mode_p): New function.
8172 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
8173 is defined, instead of checking whether the vec_perm_const_optab
8174 exists. Use targetm.vectorize.vec_perm_const instead of
8175 targetm.vectorize.vec_perm_const_ok. Check whether the indices
8176 fit in the vector mode before using a variable permute.
8177 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
8178 vec_perm_indices instead of an rtx.
8179 (expand_vec_perm): Replace with...
8180 (expand_vec_perm_const): ...this new function. Take the selector
8181 as a vec_perm_indices rather than an rtx. Also take the mode of
8182 the selector. Update call to shift_amt_for_vec_perm_mask.
8183 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
8184 Use vec_perm_indices::new_expanded_vector to expand the original
8185 selector into bytes. Check whether the indices fit in the vector
8186 mode before using a variable permute.
8187 (expand_vec_perm_var): Make global.
8188 (expand_mult_highpart): Use expand_vec_perm_const.
8189 * fold-const.c: Includes vec-perm-indices.h.
8190 * tree-ssa-forwprop.c: Likewise.
8191 * tree-vect-data-refs.c: Likewise.
8192 * tree-vect-generic.c: Likewise.
8193 * tree-vect-loop.c: Likewise.
8194 * tree-vect-slp.c: Likewise.
8195 * tree-vect-stmts.c: Likewise.
8196 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
8198 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
8199 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
8200 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
8201 (aarch64_vectorize_vec_perm_const): ...this new function.
8202 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8203 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8204 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
8205 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
8206 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8207 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8208 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
8210 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
8211 check for NEON modes.
8212 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
8213 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
8214 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
8215 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
8217 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
8218 the old VEC_PERM_CONST conditions.
8219 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
8220 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
8221 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
8222 (ia64_vectorize_vec_perm_const_ok): Merge into...
8223 (ia64_vectorize_vec_perm_const): ...this new function.
8224 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
8225 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
8226 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
8227 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
8228 * config/mips/mips.c (mips_expand_vec_perm_const)
8229 (mips_vectorize_vec_perm_const_ok): Merge into...
8230 (mips_vectorize_vec_perm_const): ...this new function.
8231 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
8232 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
8233 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
8234 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
8235 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
8236 (rs6000_expand_vec_perm_const): Delete.
8237 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
8239 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8240 (altivec_expand_vec_perm_const_le): Take each operand individually.
8241 Operate on constant selectors rather than rtxes.
8242 (altivec_expand_vec_perm_const): Likewise. Update call to
8243 altivec_expand_vec_perm_const_le.
8244 (rs6000_expand_vec_perm_const): Delete.
8245 (rs6000_vectorize_vec_perm_const_ok): Delete.
8246 (rs6000_vectorize_vec_perm_const): New function.
8247 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8248 an element count and rtx array.
8249 (rs6000_expand_extract_even): Update call accordingly.
8250 (rs6000_expand_interleave): Likewise.
8251 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
8252 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
8253 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
8254 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
8255 (rs6000_expand_vec_perm_const): Delete.
8256 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8257 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8258 (altivec_expand_vec_perm_const_le): Take each operand individually.
8259 Operate on constant selectors rather than rtxes.
8260 (altivec_expand_vec_perm_const): Likewise. Update call to
8261 altivec_expand_vec_perm_const_le.
8262 (rs6000_expand_vec_perm_const): Delete.
8263 (rs6000_vectorize_vec_perm_const_ok): Delete.
8264 (rs6000_vectorize_vec_perm_const): New function. Remove stray
8265 reference to the SPE evmerge intructions.
8266 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8267 an element count and rtx array.
8268 (rs6000_expand_extract_even): Update call accordingly.
8269 (rs6000_expand_interleave): Likewise.
8270 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
8271 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
8273 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8275 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8277 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
8278 vector mode and that that mode matches the mode of the data
8280 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
8281 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
8282 directly using expand_vec_perm_1 when forcing selectors into
8284 (expand_vec_perm_var): New function, split out from expand_vec_perm.
8286 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8288 * optabs-query.h (can_vec_perm_p): Delete.
8289 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
8290 * optabs-query.c (can_vec_perm_p): Split into...
8291 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
8292 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
8293 particular selector is valid.
8294 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8295 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8296 (vect_grouped_load_supported): Likewise.
8297 (vect_shift_permute_load_chain): Likewise.
8298 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8299 (vect_transform_slp_perm_load): Likewise.
8300 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8301 (vectorizable_bswap): Likewise.
8302 (vect_gen_perm_mask_checked): Likewise.
8303 * fold-const.c (fold_ternary_loc): Likewise. Don't take
8304 implementations of variable permutation vectors into account
8305 when deciding which selector to use.
8306 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
8307 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
8308 with a false third argument.
8309 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
8310 to test whether the constant selector is valid and can_vec_perm_var_p
8311 to test whether a variable selector is valid.
8313 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8315 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
8316 * optabs-query.c (can_vec_perm_p): Likewise.
8317 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
8318 instead of vec_perm_indices.
8319 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
8320 (vect_gen_perm_mask_checked): Likewise,
8321 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
8322 (vect_gen_perm_mask_checked): Likewise,
8324 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8326 * optabs-query.h (qimode_for_vec_perm): Declare.
8327 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
8328 (qimode_for_vec_perm): ...this new function.
8329 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
8331 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8333 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
8334 does not have a conditional at the top.
8336 2018-01-02 Richard Biener <rguenther@suse.de>
8338 * ipa-inline.c (big_speedup_p): Fix expression.
8340 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8343 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
8346 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8350 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
8351 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
8352 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
8353 cond_taken_branch_cost 3->4.
8355 2018-01-01 Jakub Jelinek <jakub@redhat.com>
8357 PR tree-optimization/83581
8358 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
8359 TODO_cleanup_cfg if any changes have been made.
8362 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
8363 convert_modes if target mode has the right side, but different mode
8367 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
8368 last argument when extracting from CONCAT. If either from_real or
8369 from_imag is NULL, use expansion through memory. If result is not
8370 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
8371 the parts directly to inner mode, if even that fails, use expansion
8375 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
8376 check for bswap in mode rather than HImode and use that in expand_unop
8379 Copyright (C) 2018 Free Software Foundation, Inc.
8381 Copying and distribution of this file, with or without modification,
8382 are permitted in any medium without royalty provided the copyright
8383 notice and this notice are preserved.