1 2018-03-08 Richard Biener <rguenther@suse.de>
3 PR tree-optimization/84178
4 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
6 (version_loop_for_if_conversion): Delay update_ssa call.
7 (tree_if_conversion): Delay update_ssa until after predicate
10 2018-03-08 David Malcolm <dmalcolm@redhat.com>
12 PR tree-optimization/84178
13 * tree-if-conv.c (release_bb_predicate): Remove the
14 the assertion that the stmts have NULL use_ops.
15 Discard the statements, asserting that they haven't
16 yet been added to a BB.
18 2018-03-08 Richard Biener <rguenther@suse.de>
20 PR tree-optimization/84746
21 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
22 (phi_translate): Pass in destination ANTIC_OUT set.
23 (phi_translate_1): Likewise. For a simplified result lookup
24 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
25 (phi_translate_set): Adjust.
26 (do_pre_regular_insertion): Likewise.
27 (do_pre_partial_partial_insertion): Likewise.
29 2018-03-08 Martin Liska <mliska@suse.cz>
32 * doc/gcov.texi: Document usage of profile files.
33 * gcov-io.h: Document changes in the format.
35 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
39 * dwarf2out.c (struct dw_line_info_table): Update comments for
41 (FORCE_RESET_NEXT_VIEW): New.
42 (FORCE_RESETTING_VIEW_P): New.
43 (RESETTING_VIEW_P): Check for -1 too.
44 (ZERO_VIEW_P): Likewise.
45 (new_line_info_table): Force-reset next view.
46 (dwarf2out_begin_function): Likewise.
47 (dwarf2out_source_line): Simplify zero_view_p initialization.
48 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
49 view directly. Omit view when omitting .loc at line 0.
51 2018-03-08 Jakub Jelinek <jakub@redhat.com>
53 PR tree-optimization/84740
54 * tree-switch-conversion.c (process_switch): Call build_constructors
55 only if info.phi_count is non-zero.
57 PR tree-optimization/84739
58 * tree-tailcall.c (find_tail_calls): Check call arguments against
59 DECL_ARGUMENTS (current_function_decl) rather than
60 DECL_ARGUMENTS (func) when checking for tail recursion.
62 2018-03-07 Jakub Jelinek <jakub@redhat.com>
64 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
65 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
66 Volker Reichelt's entry and add entries for people that perform
67 GCC fuzzy testing and report numerous bugs.
69 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
72 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
73 readonly data in sdata, if that is disabled.
74 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
75 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
76 -mreadonly-in-sdata option.
78 2018-03-07 Martin Sebor <msebor@redhat.com>
80 PR tree-optimization/84468
81 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
82 basic block when looking for nul assignment.
84 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
87 * except.h (output_function_exception_table): Adjust prototype.
88 * except.c (output_function_exception_table): Remove FNNAME parameter
89 and add SECTION parameter. Ouput one part of the table at a time.
90 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
91 the first part of the exception table and emit unwind directives.
92 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
93 (i386_pe_seh_cold_init): Likewise.
94 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
95 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
96 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
97 (ix86_output_call_insn): Emit a nop in one more case for SEH.
98 * config/i386/winnt.c: Include except.h.
99 (struct seh_frame_state): Add reg_offset, after_prologue and
100 in_cold_section fields.
101 (i386_pe_seh_end_prologue): Set seh->after_prologue.
102 (i386_pe_seh_cold_init): New function.
103 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
104 to seh->in_cold_section.
105 (seh_emit_push): Record the offset of the push.
106 (seh_emit_save): Record the offet of the save.
107 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
108 Test seh->after_prologue to disregard the epilogue.
109 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
110 (i386_pe_end_cold_function): New function.
112 2018-03-07 Jakub Jelinek <jakub@redhat.com>
115 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
116 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
119 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
121 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
122 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
125 * multiple_target.c: Include tree-inline.h and intl.h.
126 (expand_target_clones): Diagnose and fail if node->definition and
127 !tree_versionable_function_p (node->decl).
129 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
131 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
133 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
134 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
135 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
137 2018-03-06 Jakub Jelinek <jakub@redhat.com>
140 * combine.c (try_combine): Use reg_or_subregno instead of handling
141 just paradoxical SUBREGs and REGs.
143 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
145 * config/arc/arc.c (arc_finalize_pic): Remove function.
146 (arc_must_save_register): We use single base PIC register, remove
147 checks to save/restore the PIC register.
148 (arc_expand_prologue): Likewise.
149 * config/arc/arc-protos.h (arc_set_default_type_attributes):
151 (arc_verify_short): Likewise.
152 (arc_attr_type): Likewise.
153 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
154 (walk_stores): Likewise.
155 (arc_address_cost): Make it static.
156 (arc_verify_short): Likewise.
157 (branch_dest): Likewise.
158 (arc_attr_type): Likewise.
159 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
160 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
161 (arc_final_prescan_insn): Remove inserting the nops due to
162 hardware hazards. It is done in reorg step.
163 (insn_length_variant_t): Remove.
164 (insn_length_parameters_t): Likewise.
165 (arc_insn_length_parameters): Likewise.
166 (arc_get_insn_variants): Likewise.
167 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
169 2018-03-06 Jakub Jelinek <jakub@redhat.com>
172 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
175 PR tree-optimization/84687
176 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
178 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
180 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
182 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
183 Rename to ppc_speculation_barrier.
184 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
185 __builtin_ppc_speculation_barrier.
187 2018-03-05 Jakub Jelinek <jakub@redhat.com>
190 * combine.c (combine_simplify_rtx): Don't try to simplify if
191 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
194 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
196 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
197 to 32 bytes when compiling for POWER9.
199 2018-03-05 Jakub Jelinek <jakub@redhat.com>
202 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
203 regparm >= 3 with no arg reg available also for calls with
204 flag_force_indirect_call. Pass decl to ix86_function_regparm.
207 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
209 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
211 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
214 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
216 2018-03-05 Richard Biener <rguenther@suse.de>
218 PR tree-optimization/84486
219 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
220 When inserting a __builtin_assume_aligned call set the LHS
221 SSA name alignment info accordingly.
223 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
225 PR tree-optimization/84114
226 * config/aarch64/aarch64.c (aarch64_reassociation_width)
227 Avoid reassociation of FLOAT_MODE addition.
229 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
231 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
232 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
233 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
234 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
235 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
236 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
237 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
239 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
240 __builtin_ia32_wbinvd): New builtins.
241 (SPECIAL_ARGS2): New.
242 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
243 (SPECIAL_ARGS2): New.
244 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
245 (ix86_valid_target_attribute_inner_p): Ditto.
246 (ix86_init_mmx_sse_builtins): Add special_args2.
247 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
248 TARGET_WBNOINVD_P): New.
249 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
250 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
251 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
252 * config/i386/immintrin.h (_wbinvd): New intrinsic.
253 * config/i386/pconfigintrin.h: New file.
254 * config/i386/wbnoinvdintrin.h: Ditto.
255 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
256 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
258 2018-03-05 Richard Biener <rguenther@suse.de>
260 PR tree-optimization/84670
261 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
263 (BB_VISITED_WITH_VISITED_SUCCS): New define.
264 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
265 (compute_antic_aux): Only assert the number of values in ANTIC_IN
266 doesn't grow if all successors (recursively) were visited at least
269 2018-03-05 Richard Biener <rguenther@suse.de>
271 PR tree-optimization/84650
272 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
273 if executed in the loop pipeline.
275 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
277 * doc/configfiles.texi (Configuration Files): Move info about
278 conditionalizing $target-protos.h to...
279 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
280 differs from $target-protos.h.
282 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
283 Chung-Ju Wu <jasonwucj@gmail.com>
285 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
286 * config/nds32/nds32-multiple.md (setmemsi): Define.
287 * config/nds32/nds32-memory-manipulation.c
288 (nds32_gen_dup_4_byte_to_word_value): New.
289 (emit_setmem_word_loop): New.
290 (emit_setmem_byte_loop): New.
291 (nds32_expand_setmem_loop): New.
292 (nds32_expand_setmem_loop_v3m): New.
293 (nds32_expand_setmem_unroll): New.
294 (nds32_expand_setmem): New.
296 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
297 Chung-Ju Wu <jasonwucj@gmail.com>
299 * config/nds32/nds32-memory-manipulation.c
300 (nds32_emit_load_store): New.
301 (nds32_emit_post_inc_load_store): New.
302 (nds32_emit_mem_move): New.
303 (nds32_emit_mem_move_block): New.
304 (nds32_expand_movmemsi_loop_unknown_size): New.
305 (nds32_expand_movmemsi_loop_known_size): New.
306 (nds32_expand_movmemsi_loop): New.
307 (nds32_expand_movmemsi_unroll): New.
308 (nds32_expand_movmemqi): Rename ...
309 (nds32_expand_movmemsi): ... to this.
310 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
311 (movmemsi): ... to this.
312 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
313 (nds32_expand_movmemsi): ... to this.
315 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
316 Monk Chiang <sh.chiang04@gmail.com>
317 Chung-Ju Wu <jasonwucj@gmail.com>
319 * config/nds32/nds32-protos.h
320 (nds32_expand_load_multiple): New arguments.
321 (nds32_expand_store_multiple): Ditto.
322 (nds32_valid_multiple_load_store): Rename ...
323 (nds32_valid_multiple_load_store_p): ... to this.
324 * config/nds32/nds32-memory-manipulation.c
325 (nds32_expand_load_multiple): Refine implementation.
326 (nds32_expand_store_multiple): Ditto.
327 * config/nds32/nds32-multiple.md
328 (load_multiple): Update nds32_expand_load_multiple interface.
329 (store_multiple): Update nds32_expand_store_multiple interface.
330 * config/nds32/nds32-predicates.c
331 (nds32_valid_multiple_load_store): Rename ...
332 (nds32_valid_multiple_load_store_p): ... to this and refine
334 * config/nds32/predicates.md
335 (nds32_load_multiple_and_update_address_operation): New predicate.
336 (nds32_store_multiple_and_update_address_operation): New predicate.
338 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
339 Chung-Ju Wu <jasonwucj@gmail.com>
341 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
342 (combo): New attribute.
343 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
345 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
347 * config/nds32/nds32.opt: Change -mcmodel= default value.
349 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
350 Monk Chiang <sh.chiang04@gmail.com>
351 Chung-Ju Wu <jasonwucj@gmail.com>
353 * config/nds32/constants.md (unspec_element): New enum.
354 * config/nds32/constraints.md (Umw): New constraint.
355 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
356 * config/nds32/nds32-intrinsic.md: Likewise.
357 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
358 (nds32_valid_smw_lwm_base_p): New.
359 (nds32_output_smw_single_word): New.
360 (nds32_output_lmw_single_word): New.
361 (nds32_expand_unaligned_load): New.
362 (nds32_expand_unaligned_store): New.
363 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
364 (nds32_output_smw_single_word): Declare.
365 (nds32_output_lmw_single_word): Declare.
366 (nds32_expand_unaligned_load): Declare.
367 (nds32_expand_unaligned_store): Declare.
368 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
369 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
370 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
371 NDS32_BUILTIN_UASTORE_DW.
372 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
375 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
376 Kito Cheng <kito.cheng@gmail.com>
377 Chung-Ju Wu <jasonwucj@gmail.com>
379 * config/nds32/nds32-intrinsic.c
380 (nds32_expand_builtin_null_ftype_reg): Delete.
381 (nds32_expand_builtin_reg_ftype_imm): Ditto.
382 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
383 (nds32_read_argument): New.
384 (nds32_legitimize_target): Ditto.
385 (nds32_legitimize_argument): Ditto.
386 (nds32_check_constant_argument): Ditto.
387 (nds32_expand_unop_builtin): Ditto.
388 (nds32_expand_unopimm_builtin): Ditto.
389 (nds32_expand_binop_builtin): Ditto.
390 (nds32_builtin_decl_impl): Ditto.
391 (builtin_description): Ditto.
392 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
393 (nds32_init_builtins_impl): Ditto.
394 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
395 (nds32_builtin_decl): New.
396 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
397 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
399 2018-03-02 Jeff Law <law@redhat.com>
401 * reorg.c (stop_search_p): Handle DEBUG_INSN.
402 (redundant_insn, fill_simple_delay_slots): Likewise.
403 (fill_slots_from_thread): Likewise.
404 * resource.c (mark_referenced_resources): Likewise.
405 (mark_set_resources, find_dead_or_set_registers): Likewise.
407 2018-03-02 Jakub Jelinek <jakub@redhat.com>
409 * substring-locations.h (format_warning_va): Formatting fix for
411 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
413 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
414 * substring-locations.c: Include intl.h.
415 (format_warning_va): Turned into small wrapper around
416 format_warning_n_va, renamed to ...
417 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
418 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
420 (format_warning_at_substring_n): New function.
421 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
422 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
423 format_warning_at_substring with just a shorter name instead of
424 const function pointer.
425 (fmtwarn_n): New function.
426 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
427 appropriate, get rid of all the fmtstr temporaries, move conditionals
428 with G_() wrapped string literals directly into fmtwarn arguments,
429 cast dir.len to (int), formatting fixes.
431 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
433 * doc/invoke.texi: Remove "Cilk Plus" references.
435 2018-03-02 Jakub Jelinek <jakub@redhat.com>
436 Richard Biener <rguenther@suse.de>
439 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
440 for error or warning attributes if CALL_FROM_THUNK_P is set.
443 2018-03-02 Jakub Jelinek <jakub@redhat.com>
446 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
447 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
450 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
451 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
453 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
454 instead of -1U in last predictors element's probability member.
456 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
459 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
460 arguments if they are comparable.
462 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
464 PR tree-optimization/84634
465 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
466 masks and masked_loop_p with a single loop_masks, making sure it's
467 null for bb vectorization.
469 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
471 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
472 (vect_analyze_data_ref_access): Use loop->safe_len rather than
473 loop->force_vectorize to check whether there is no alias.
475 2018-03-02 Jakub Jelinek <jakub@redhat.com>
478 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
480 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
482 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
483 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
484 instead of a loop around prev_real_insn.
485 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
489 * config/i386/i386.c (ix86_print_operand): Use conditional
490 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
493 2018-03-02 Richard Biener <rguenther@suse.de>
495 PR tree-optimization/84427
496 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
497 (bitmap_set_subtract_values): Rewrite to handle multiple
500 (prune_clobbered_mems): Likewise.
501 (phi_translate): Take edge instead of pred/phiblock.
502 (phi_translate_1): Likewise.
503 (phi_translate_set): Likewise. Insert all translated
504 exprs for a value into the set, keeping possibly multiple
505 expressions per value.
506 (compute_antic_aux): Adjust for phi_translate changes.
507 When intersecting union the expressions and prune those
508 not in the final value set, keeping possibly multiple
509 expressions per value. Do not use value-insertion
510 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
511 all expressions. Add verification that the value-sets
512 only shrink during iteration.
513 (compute_partial_antic_aux): Adjust for the phi_translate changes.
514 (do_pre_regular_insertion): Likewise.
515 (do_pre_partial_partial_insertion): Likewise.
517 2018-03-02 Richard Biener <rguenther@suse.de>
520 * config/darwin.c (saved_debug_info_level): New static global.
521 (darwin_asm_lto_start): Disable debug info generation for LTO out.
522 (darwin_asm_lto_end): Restore debug info generation settings.
524 2018-03-01 Martin Liska <mliska@suse.cz>
527 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
530 2018-03-01 Richard Biener <rguenther@suse.de>
533 * dwarf2out.c (gen_variable_die): Properly handle late VLA
534 type annotation with LTO when debug was disabled at compile-time.
536 2018-03-01 Matthew Fortune <mfortune@gmail.com>
538 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
540 (mips_final_postscan_insn): Likewise.
542 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
544 PR rtl-optimization/84528
545 * alias.c (init_alias_target): Add commentary.
546 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
547 a unique base value if the frame pointer is not eliminated
548 to the stack pointer.
550 2018-03-01 Tom de Vries <tom@codesourcery.com>
552 PR rtl-optimization/83327
553 * lra-int.h (hard_regs_spilled_into): Declare.
554 * lra.c (hard_regs_spilled_into): Define.
555 (init_reg_info): Init hard_regs_spilled_into.
556 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
557 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
558 (process_bb_lives): Handle hard_regs_spilled_into.
559 (lra_create_live_ranges_1): Before doing liveness propagation, clear
560 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
562 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
564 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
565 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
566 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
567 * config/rs6000/aix72.h: New file.
569 2018-02-28 Jakub Jelinek <jakub@redhat.com>
571 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
572 instead of warning_at with conditional singular and plural messages
576 * stor-layout.c (update_alignment_for_field): For
577 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
578 && !DECL_PACKED (field), do the alignment update, just use
579 only desired_align instead of MAX (type_align, desired_align)
581 (place_field): Don't do known_align < desired_align handling
582 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
583 is non-NULL, instead do it after rli->prev_field handling and
584 only if not within a bitfield word. For DECL_PACKED (field)
585 use type_align of BITS_PER_UNIT.
587 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
589 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
590 superfluous parentheses and trailing spaces.
592 2018-02-28 Richard Biener <rguenther@suse.de>
594 PR tree-optimization/84584
595 * graphite-scop-detection.c (scop_detection::add_scop): Discard
596 SCoPs with fake exit edge.
598 2018-02-28 Martin Liska <mliska@suse.cz>
601 * timevar.c (timer::print): Fix format to properly print 100%
604 2018-02-28 Richard Biener <rguenther@suse.de>
607 * genmatch.c (capture_info::walk_match): Do not mark
608 captured expressions without operands as expr_p given
609 they act more like predicates and should be subject to
610 "lost tail" side-effect preserving.
612 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
614 PR rtl-optimization/81611
615 * auto-inc-dec.c (attempt_change): Move dead note from
616 mem_insn if it's the next use of regno
617 (find_address): Take address use of reg holding
618 non-incremented value. Add parm to limit search to the named
620 (merge_in_block): Attempt to use a mem insn that is the next
621 use of the original regno.
623 2018-02-27 Martin Sebor <msebor@redhat.com>
626 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
627 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
629 2018-02-27 Martin Sebor <msebor@redhat.com>
632 * diagnostic-core.h (warning_n, error_n, inform_n): Change
633 n argument to unsigned HOST_WIDE_INT.
634 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
635 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
636 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
637 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
639 2018-02-27 Richard Biener <rguenther@suse.de>
641 PR tree-optimization/84512
642 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
643 Do not use the estimate returned from record_stmt_cost for
644 the scalar iteration cost but sum properly using add_stmt_cost.
646 2018-02-27 Richard Biener <rguenther@suse.de>
648 PR tree-optimization/84466
649 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
650 Adjust last change to less strictly validate use operands.
652 2018-02-27 Martin Liska <mliska@suse.cz>
654 PR gcov-profile/84548
655 * gcov.c (process_file): Allow partial overlap and consider it
656 also as group functions.
657 (output_lines): Properly calculate range of lines for a group.
659 2018-02-27 Martin Liska <mliska@suse.cz>
661 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
662 'ggc' suffixes. Change first column width.
663 (timer::print): Fix formatting of the column.
665 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
667 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
668 preserve inline entry blocks for the sake of debug inline
669 entry point markers alone.
670 (remove_unused_locals): Suggest in comments a better place to
671 force the preservation of inline entry blocks that are
672 otherwise unused, but do not preserve them.
674 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
676 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
678 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
681 * config/i386/constraints.md (Bs): Replace
682 ix86_indirect_branch_register with
683 TARGET_INDIRECT_BRANCH_REGISTER.
685 * config/i386/i386.md (indirect_jump): Likewise.
686 (tablejump): Likewise.
687 (*sibcall_memory): Likewise.
688 (*sibcall_value_memory): Likewise.
689 Peepholes of indirect call and jump via memory: Likewise.
690 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
691 (*sibcall_value_GOT_32): Likewise.
692 * config/i386/predicates.md (indirect_branch_operand): Likewise.
693 (GOT_memory_operand): Likewise.
694 (call_insn_operand): Likewise.
695 (sibcall_insn_operand): Likewise.
696 (GOT32_symbol_operand): Likewise.
697 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
699 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
701 PR rtl-optimization/83496
702 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
703 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
704 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
705 redundant insn, if any.
706 (relax_delay_slots): Likewise.
707 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
709 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
711 PR tree-optimization/83965
712 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
713 that grouped statements are part of a reduction chain. Return
714 true if the statement is not marked as a reduction itself but
716 (vect_recog_dot_prod_pattern): Don't check whether the statement
717 is part of a group here.
718 (vect_recog_sad_pattern): Likewise.
719 (vect_recog_widen_sum_pattern): Likewise.
721 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
724 * final.c (rest_of_clean_state): Also look for calls inside sequences.
726 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
729 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
731 (ix86_output_indirect_function_return): New prototype.
732 (ix86_split_simple_return_pop_internal): Likewise.
733 * config/i386/i386.c (indirect_return_via_cx): New.
734 (indirect_return_via_cx_bnd): Likewise.
735 (indirect_thunk_name): Handle return va CX_REG.
736 (output_indirect_thunk_function): Create alias for
737 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
738 (ix86_output_indirect_jmp): Remove the bool argument.
739 (ix86_output_indirect_function_return): New function.
740 (ix86_split_simple_return_pop_internal): Likewise.
741 * config/i386/i386.md (*indirect_jump): Don't pass false
742 to ix86_output_indirect_jmp.
743 (*tablejump_1): Likewise.
744 (simple_return_pop_internal): Change it to define_insn_and_split.
745 Call ix86_split_simple_return_pop_internal to split it for
747 (simple_return_indirect_internal): Call
748 ix86_output_indirect_function_return instead of
749 ix86_output_indirect_jmp.
751 2018-02-26 Jakub Jelinek <jakub@redhat.com>
754 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
755 memset and value initialization afterwards.
757 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
759 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
761 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
764 * common/config/aarch64/aarch64-common.c
765 (aarch_option_optimization_table[]): Switch
766 off fomit-frame-pointer
768 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
769 Chung-Ju Wu <jasonwucj@gmail.com>
771 * config/nds32/nds32-multiple.md (load_multiple): Disallow
773 (store_multiple): Ditto.
775 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
777 * config.gcc: Add --with-cpu support for nds32 target.
778 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
779 * config/nds32/nds32.opt: Add -mcpu= option.
781 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
783 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
784 isel=yes): Warn for these deprecated options.
786 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
788 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
789 ISA_2_5_MASKS_EMBEDDED.
791 2018-02-23 Jakub Jelinek <jakub@redhat.com>
793 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
794 p->max as pointers rather than using iterative_hash_expr.
796 2018-02-23 Carl Love <cel@us.ibm.com>
798 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
799 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
801 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
802 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
803 P8V_BUILTIN_VEC_VUNSIGNED2.
805 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
808 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
809 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
811 * lra-constraints.c (curr_insn_transform): Set up
812 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
814 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
815 * lra-eliminations.c (spill_pseudos): Ditto.
816 (process_insn_for_elimination): Ditto.
817 * lra-lives.c (reg_early_clobber_p): Use the new macros.
818 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
819 LRA_NON_CLOBBERED_ALT.
821 2018-02-22 Martin Sebor <msebor@redhat.com>
823 PR tree-optimization/84480
824 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
825 to maybe_diag_stxncpy_trunc. Call it.
826 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
827 from gimple_fold_builtin_strcpy. Print inlining stack.
828 (handle_builtin_stxncpy): Print inlining stack.
829 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
831 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
834 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
835 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
836 and -fcheck-pointer-bounds are used together.
837 (indirect_thunk_prefix): New enum.
838 (indirect_thunk_need_prefix): New function.
839 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
840 "_nt" instead of "_bnd" for NOTRACK prefix.
841 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
842 (output_indirect_thunk_function): Likewise.
844 (ix86_code_end): Update output_indirect_thunk_function calls.
845 (ix86_output_indirect_branch_via_reg): Replace
846 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
847 (ix86_output_indirect_branch_via_push): Likewise.
848 (ix86_output_function_return): Likewise.
849 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
850 incompatible with -fcf-protection=branch and
851 -fcheck-pointer-bounds.
853 2018-02-22 Steve Ellcey <sellcey@cavium.com>
856 * config/aarch64/aarch64.c (aarch64_print_address_internal):
857 Change gcc_assert call to output_operand_lossage.
859 2018-02-22 Steve Ellcey <sellcey@cavium.com>
861 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
863 2018-02-22 DJ Delorie <dj@redhat.com>
864 Sebastian Perta <sebastian.perta@renesas.com>
865 Oleg Endo <olegendo@gcc.gnu.org>
867 * config/rx/rx.c (rx_rtx_costs): New function.
868 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
870 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
872 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
874 2018-02-22 Martin Liska <mliska@suse.cz>
877 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
878 Add "native" as a possible value.
880 2018-02-22 Martin Liska <mliska@suse.cz>
883 * config/i386/i386.c (ix86_option_override_internal):
884 Add "native" as a possible value for -march and -mtune.
886 2018-02-22 Jakub Jelinek <jakub@redhat.com>
889 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
890 to all type variants.
892 PR tree-optimization/84503
893 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
894 width as info->bitpos + info->bitsize - start.
895 (merged_store_group::merge_overlapping): Simplify width computation.
896 (check_no_overlap): New function.
897 (imm_store_chain_info::try_coalesce_bswap): Compute expected
898 start + width and last_order of the group, fail if check_no_overlap
900 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
901 to group if check_no_overlap fails.
903 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
905 * config/rs6000/altivec.md: Delete contraint arguments to
906 define_expand, define_split, and define_peephole2, and in
907 define_insn_and_split if always unused.
908 * config/rs6000/darwin.md: Ditto.
909 * config/rs6000/dfp.md: Ditto.
910 * config/rs6000/rs6000.md: Ditto.
911 * config/rs6000/sync.md: Ditto.
912 * config/rs6000/vector.md: Ditto.
913 * config/rs6000/vsx.md: Ditto.
915 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
917 * config/rs6000/altivec.md: Write output control strings as braced
918 blocks instead of double-quoted strings.
919 * config/rs6000/darwin.md: Ditto.
920 * config/rs6000/rs6000.md: Ditto.
921 * config/rs6000/vector.md: Ditto.
922 * config/rs6000/vsx.md: Ditto.
924 2018-02-21 Jason Merrill <jason@redhat.com>
926 PR c++/84314 - ICE with templates and fastcall attribute.
927 * attribs.c (build_type_attribute_qual_variant): Remove assert.
929 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
931 * ipa-cp.c (determine_versionability): Fix comment typos.
933 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
936 * ipa-cp.c (determine_versionability): Do not version functions caling
939 2018-02-21 Martin Liska <mliska@suse.cz>
942 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
943 Add "native" as a possible value.
944 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
945 the macro when native cpu detection is available.
947 2018-02-21 Martin Liska <mliska@suse.cz>
950 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
951 Add "native" as a possible value.
952 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
953 when native cpu detection is available.
955 2018-02-21 Jakub Jelinek <jakub@redhat.com>
956 Martin Sebor <msebor@redhat.com>
958 PR tree-optimization/84478
959 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
961 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
962 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
963 support which is conservatively correct, for 2 only stay conservative
964 for maxlen. Formatting and comment capitalization fixes. Add STRICT
965 argument to the 2 argument get_range_strlen, adjust 6 arg
966 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
968 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
969 (gimple_fold_builtin_strlen): Pass true as last argument to
972 2018-02-20 Martin Sebor <msebor@redhat.com>
975 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
976 (builtin_memref::set_base_and_offset): Same. Handle inner references.
977 (builtin_memref::builtin_memref): Factor out parts into
978 set_base_and_offset and call it.
980 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
983 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
984 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
985 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
986 search at the associated MODE_INT.
988 2018-02-20 Jeff Law <law@redhat.com>
991 PR tree-optimization/81592
993 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
994 for range data rather than using global data.
995 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
996 range data rather than using global data.
997 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
998 pass it to children as needed.
999 (struct directive::fmtresult): Similarly.
1000 (struct directive::set_width): Similarly.
1001 (struct directive::set_precision): Similarly.
1002 (format_integer, format_directive, parse_directive): Similarly.
1003 (format_none): Accept unnamed vr_values parameter.
1004 (format_percent, format_floating, format_character): Similarly.
1005 (format_string, format_plain): Similarly.
1006 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
1007 the EVRP range analyzer for range data rather than using global data.
1008 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
1009 gimple-ssa-evrp-analyze.h
1010 (class sprintf_dom_walker): Add after_dom_children member function.
1011 Add evrp_range_analyzer member.
1012 (sprintf_dom_walker::before_dom_children): Call into the EVRP
1013 range analyzer as needed.
1014 (sprintf_dom_walker::after_dom_children): New member function.
1015 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
1017 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
1018 (evrp_range_analyzer::pop_to_marker): Likewise.
1020 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1022 PR tree-optimization/84419
1023 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
1024 with the required type if its current type is compatible but
1027 2018-02-20 Jakub Jelinek <jakub@redhat.com>
1030 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
1031 after vectorization.
1033 2018-02-20 Martin Liska <mliska@suse.cz>
1036 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
1037 possible values if we don't have a hint.
1039 2018-02-20 Martin Liska <mliska@suse.cz>
1043 * final.c (shorten_branches): Build align_tab array with one
1045 * opts.c (finish_options): Add alignment option limit check.
1046 (MAX_CODE_ALIGN): Likewise.
1047 (MAX_CODE_ALIGN_VALUE): Likewise.
1048 * doc/invoke.texi: Document maximum allowed option value for
1049 all -falign-* options.
1051 2018-02-19 Jakub Jelinek <jakub@redhat.com>
1054 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
1055 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
1056 * var-tracking.c (emit_note_insn_var_location): Remove all references
1057 to NOTE_INSN_CALL_ARG_LOCATION.
1058 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
1059 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
1060 Use copy_rtx_if_shared.
1061 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
1062 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
1063 (dwarf2out_var_location): Remove handling of
1064 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
1066 * final.c (final_scan_insn): Remove all references to
1067 NOTE_INSN_CALL_ARG_LOCATION.
1068 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
1069 before dumping final insns.
1070 * except.c (emit_note_eh_region_end): Remove all references to
1071 NOTE_INSN_CALL_ARG_LOCATION.
1072 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
1073 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
1074 * config/arc/arc.c (hwloop_optimize): Likewise.
1075 * config/arm/arm.c (create_fix_barrier): Likewise.
1076 * config/s390/s390.c (s390_chunkify_start): Likewise.
1077 * config/sh/sh.c (find_barrier): Likewise.
1078 * config/i386/i386.c (rest_of_insert_endbranch,
1079 ix86_seh_fixup_eh_fallthru): Likewise.
1080 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
1081 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
1082 * config/frv/frv.c (frv_function_prologue): Likewise.
1083 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
1085 (note_outside_basic_block_p): Remove all references to
1086 NOTE_INSN_CALL_ARG_LOCATION.
1087 * gengtype.c (adjust_field_rtx_def): Likewise.
1088 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
1090 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
1091 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
1094 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
1097 PR tree-optimization/84452
1098 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
1099 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
1102 2018-02-19 Martin Liska <mliska@suse.cz>
1105 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
1107 2018-02-19 Martin Liska <mliska@suse.cz>
1108 Richard Sandiford <richard.sandiford@linaro.org>
1110 PR tree-optimization/82491
1111 * gimple-fold.c (get_base_constructor): Make earlier bail out
1114 2018-02-19 Carl Love <cel@us.ibm.com>
1116 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
1117 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
1119 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
1120 P8V_BUILTIN_VEC_NEG.
1122 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
1124 * config/rl78/rl78.md (movdf): New define expand.
1126 2018-02-19 Martin Liska <mliska@suse.cz>
1129 * doc/invoke.texi: Fix typo.
1130 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
1132 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
1134 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
1135 handle rs6000_single_float and rs6000_double_float specially for
1138 2018-02-16 Jeff Law <law@redhat.com>
1140 * config/rx/rx.c (add_pop_cfi_notes): New function.;
1143 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1146 * ipa-inline.c (inline_small_functions): Fix a typo.
1148 2018-02-16 Nathan Sidwell <nathan@acm.org>
1150 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
1152 2018-02-16 Carl Love <cel@us.ibm.com>
1154 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
1155 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
1156 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
1157 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
1158 expansion to P8V_BUILTIN_VEC_FLOAT2.
1160 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
1162 PR rtl-optimization/70023
1163 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
1164 src_regno into account.
1166 2018-02-16 Carl Love <cel@us.ibm.com>
1168 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
1169 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
1170 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
1171 * config/rs6000/rs6000.c: Remove case statements for
1172 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
1173 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
1174 and P9V_BUILTIN_VEC_VINSERT4B.
1175 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
1176 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
1177 * config/rs6000/vsx.md:
1178 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
1181 2018-02-16 Carl Love <cel@us.ibm.com>
1183 * config/rs6000/altivec.h: Add builtin names vec_extract4b
1185 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
1187 * config/rs6000/rs6000-c.c: Add the definitions for
1188 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
1189 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
1190 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
1191 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
1192 definition for insert4b and define insn *insert3b_internal.
1193 * doc/extend.texi: Add documentation for vec_extract4b.
1195 2018-02-16 Nathan Sidwell <nathan@acm.org>
1197 * doc/extend.texi (Backwards Compatibility): Mention friend
1198 injection. Note for-scope is deprecated.
1199 * doc/invoke.texi (-ffriend-injection): Deprecate.
1201 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
1203 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
1204 that moved to I2, also allow destinations that are a paradoxical
1205 subreg (instead of a normal reg).
1207 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
1210 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
1213 2018-02-16 Richard Biener <rguenther@suse.de>
1215 PR tree-optimization/84037
1216 PR tree-optimization/84016
1218 * config/i386/i386.c (ix86_builtin_vectorization_cost):
1219 Adjust vec_construct for the fact we need additional higher latency
1220 128bit inserts for AVX256 and AVX512 vector builds.
1221 (ix86_add_stmt_cost): Scale vector construction cost for
1224 2018-02-16 Richard Biener <rguenther@suse.de>
1226 PR tree-optimization/84417
1227 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
1228 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
1229 (non_rewritable_lvalue_p): Likewise, use poly-ints.
1231 2018-02-16 Martin Liska <mliska@suse.cz>
1234 * internal-fn.def (ASAN_CHECK): Set proper flags.
1235 (ASAN_MARK): Likewise.
1237 2018-02-16 Julia Koval <julia.koval@intel.com>
1239 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
1240 from PTA_CANNONLAKE.
1242 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1245 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
1246 Use ++iter rather than iter++ for std::list iterators.
1247 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
1248 defer deleting them until all nodes in the forest are processed. Do
1249 free even leaf nodes. Change to_process into auto_vec.
1252 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
1253 * vec.h (vec_default_construct): Use memset instead of placement new
1254 if BROKEN_VALUE_INITIALIZATION is defined.
1255 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
1256 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
1259 PR rtl-optimization/83723
1260 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
1261 * lra.c (lra_substitute_pseudo): Likewise. If true, use
1262 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
1264 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
1266 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
1268 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
1270 PR rtl-optimization/81443
1271 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
1272 from inner REGs to paradoxical SUBREGs.
1274 2018-02-16 Richard Biener <rguenther@suse.de>
1276 PR tree-optimization/84399
1277 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1278 For operands we can analyze at their definition make sure we can
1279 analyze them at each use as well.
1281 2018-02-16 Richard Biener <rguenther@suse.de>
1283 PR tree-optimization/84190
1284 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
1285 volatile accesses if the decl isn't volatile.
1287 2018-02-15 Jason Merrill <jason@redhat.com>
1289 PR c++/84314 - ICE with templates and fastcall attribute.
1290 * attribs.c (build_type_attribute_qual_variant): Don't clobber
1291 TYPE_CANONICAL on an existing type.
1293 2018-02-15 Jakub Jelinek <jakub@redhat.com>
1295 PR tree-optimization/84383
1296 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
1297 dstoff nor call operand_equal_p if dstbase is NULL.
1299 PR tree-optimization/84334
1300 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
1301 also a CONSTANT_CLASS_P, punt.
1303 2018-02-14 Jim Wilson <jimw@sifive.com>
1305 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
1306 first SMALL_OPERAND check. New local min_second_step. Move assert
1307 to where locals are set. Add TARGET_RVC support.
1308 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
1310 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
1312 * doc/invoke.texi: Correct -Wformat-overflow code sample.
1314 2018-02-14 Martin Sebor <msebor@redhat.com>
1316 PR tree-optimization/83698
1317 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
1318 arrays constrain the offset range to their bounds.
1319 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
1320 (builtin_access::overlap): Avoid setting the size of overlap if it's
1322 (maybe_diag_overlap): Also consider arrays when deciding what values
1323 of offsets to include in diagnostics.
1325 2018-02-14 Martin Sebor <msebor@redhat.com>
1328 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
1329 that correspond to the kind of a declaration.
1331 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
1334 * config/pa/pa.md: Load address of PIC label using the linkage table
1335 if the label is nonlocal.
1337 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
1339 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
1340 warning message if user requests -maltivec=be.
1341 * doc/invoke.texi: Document deprecation of -maltivec=be.
1343 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
1346 * config/rs6000/rs6000-c.c: Update definitions for
1347 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
1348 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
1350 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1353 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
1354 add _get_ssp intrinsics. Remove argument from
1355 __builtin_ia32_rdssp[d|q].
1356 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
1357 * config/i386/i386-builtin.def: Remove argument from
1358 __builtin_ia32_rdssp[d|q].
1359 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
1360 ix86_expand_special_args_builtin for _rdssp[d|q].
1361 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
1362 Clear register before usage.
1363 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
1364 Add documentation for new _get_ssp and _inc_ssp intrinsics.
1366 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
1368 PR tree-optimization/84357
1369 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
1370 operand 1 of an ARRAY_REF too.
1372 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
1375 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
1376 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1378 (set_of_reg): New struct.
1379 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
1380 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
1381 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1383 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
1384 Split into bitclr, bitset, bitinvert patterns if appropriate.
1385 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
1386 use rx_fuse_in_memory_bitop.
1387 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
1388 to named insn, correct maximum insn length.
1390 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
1393 * machmode.def: Define a complex mode for PARTIAL_INT.
1394 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
1396 * doc/rtl.texi: Document CSPImode.
1397 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
1399 (msp430_hard_regno_nregs_with_padding): Likewise.
1401 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
1404 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
1406 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
1408 PR rtl-optimization/84169
1409 * combine.c (try_combine): New variable split_i2i3. Set it to true if
1410 we generated a parallel as new i3 and we split that to new i2 and i3
1411 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
1412 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
1413 those to i2, not i1. Partially rewrite this scan code.
1415 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1418 * stor-layout.c (place_field): For variable length fields, adjust
1419 offset_align afterwards not just based on the field's alignment,
1420 but also on the size.
1423 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
1424 of exps and logs in the use_exp2 case.
1426 2018-02-13 Jeff Law <law@redhat.com>
1428 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
1431 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
1434 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
1438 * common.opt (gas-loc-support, gas-locview-support): New.
1439 (ginline-points, ginternal-reset-location-views): New.
1440 * doc/invoke.texi: Document them. Use @itemx where intended.
1441 (gvariable-location-views): Adjust.
1442 * target.def (reset_location_view): New.
1443 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
1444 (TARGET_RESET_LOCATION_VIEW): New.
1445 * doc/tm.texi: Rebuilt.
1446 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
1447 (dwarf2out_default_as_locview_support): New.
1448 (output_asm_line_debug_info): Use option variables.
1449 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
1450 (output_loc_list): Likewise.
1451 (add_high_low_attributes): Check option variables.
1452 Don't output entry view attribute in strict mode.
1453 (gen_inlined_subroutine_die): Check option variables.
1454 (dwarf2out_inline_entry): Likewise.
1455 (init_sections_and_labels): Likewise.
1456 (dwarf2out_early_finish): Likewise.
1457 (maybe_reset_location_view): New, from...
1458 (dwarf2out_var_location): ... here. Call it.
1459 * debug.h (dwarf2out_default_as_loc_support): Declare.
1460 (dwarf2out_default_as_locview_support): Declare.
1461 * hooks.c (hook_int_rtx_insn_0): New.
1462 * hooks.h (hook_int_rtx_insn_0): Declare.
1463 * toplev.c (process_options): Take -gas-loc-support and
1464 -gas-locview-support from dwarf2out. Enable
1465 -gvariable-location-views by default only with locview
1466 assembler support. Enable -ginternal-reset-location-views by
1467 default only if the target defines the corresponding hook.
1468 Enable -ginline-points by default if location views are
1469 enabled; force it disabled if statement frontiers are
1471 * tree-inline.c (expand_call_inline): Check option variables.
1472 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
1474 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
1476 PR tree-optimization/84321
1477 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
1478 handling. Also check whether the anti-range contains any values
1479 that satisfy the mask; switch to a VR_RANGE if not.
1481 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
1484 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
1486 2018-02-13 Martin Jambor <mjambor@suse.cz>
1489 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
1490 of call statements, also set location of a load to a temporary.
1492 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
1494 * config/rl78/rl78.c (add_vector_labels): New function.
1495 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
1496 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
1497 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
1498 which checks that no arguments are passed.
1499 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
1500 * doc/extend.texi: Documentation for the new attribute.
1502 2018-02-13 Andreas Schwab <schwab@suse.de>
1504 * config/riscv/linux.h (CPP_SPEC): Define.
1506 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1509 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
1510 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
1511 OPTION_MASK_ISA_AES as first argument to def_builtin_const
1512 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
1513 instead of OPTION_MASK_ISA_PCLMUL as first argument to
1514 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
1515 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
1516 temporarily for AES and PCLMUL builtins.
1518 PR tree-optimization/84339
1519 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
1520 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
1524 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
1525 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
1526 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
1528 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
1530 * omp-simd-clone.h: New file.
1531 * omp-simd-clone.c: Include omp-simd-clone.h.
1532 (expand_simd_clones): No longer static.
1533 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
1534 cgraph.h and omp-simd-clone.h.
1535 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
1536 (vect_recog_widen_shift_pattern): Formatting fix.
1537 (vect_pattern_recog_1): Don't check optab for calls.
1540 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
1541 operands[2] into a REG before using gen_lowpart on it.
1543 2018-02-12 Jeff Law <law@redhat.com>
1546 * config/sh/sh.c (find_barrier): Consider a sibling call
1549 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
1550 successfully back substituting a reg.
1552 2018-02-12 Richard Biener <rguenther@suse.de>
1554 PR tree-optimization/84037
1555 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
1556 parameter, move visited init to caller.
1557 (vect_slp_analyze_operations): Separate cost from validity
1558 check, initialize visited once for all instances.
1559 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
1561 * tree-vect-stmts.c (vect_model_simple_cost): Make early
1563 (vect_model_promotion_demotion_cost): Likewise.
1564 (vectorizable_bswap): Guard cost modeling with !slp_node
1565 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
1567 (vectorizable_call): Likewise.
1568 (vectorizable_conversion): Likewise.
1569 (vectorizable_assignment): Likewise.
1570 (vectorizable_shift): Likewise.
1571 (vectorizable_operation): Likewise.
1572 (vectorizable_store): Likewise.
1573 (vectorizable_load): Likewise.
1574 (vectorizable_condition): Likewise.
1575 (vectorizable_comparison): Likewise.
1577 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
1580 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
1581 (ASAN_MARK): Fix fnspec to account for return value, change pointer
1582 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
1584 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1587 * params.def (inline-min-speedup): Increase from 8 to 15.
1588 (max-inline-insns-auto): Decrease from 40 to 30.
1589 * ipa-split.c (consider_split): Add some buffer for function to
1590 be considered inlining candidate.
1591 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
1594 2018-02-12 Richard Biener <rguenther@suse.de>
1596 PR tree-optimization/84037
1597 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
1598 matched stmts if we cannot swap the non-matched ones.
1600 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1602 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
1603 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
1604 _mm_maskz_scalef_round_ss): New intrinsics.
1605 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
1606 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
1607 __builtin_ia32_scalefss_round): Remove.
1608 (__builtin_ia32_scalefsd_mask_round,
1609 __builtin_ia32_scalefss_mask_round): New intrinsics.
1610 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
1611 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
1612 ((match_operand:VF_128 2 "<round_nimm_predicate>"
1613 "<round_constraint>")): Changed to ...
1614 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
1615 "<round_scalar_constraint>")): ... this.
1616 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
1617 %0, %1, %2<round_op3>}"): Changed to ...
1618 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
1619 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
1620 %2<round_scalar_mask_op3>}"): ... this.
1621 * config/i386/subst.md (round_scalar_nimm_predicate): New.
1623 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1625 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
1626 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
1627 (_mm_maskz_sqrt_round_ss): New intrinsics.
1628 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
1629 (__builtin_ia32_sqrtsd_mask_round)
1630 (__builtin_ia32_sqrtss_mask_round): New builtins.
1631 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
1632 (__builtin_ia32_sqrtss_round): Remove.
1633 (__builtin_ia32_sqrtsd_mask_round)
1634 (__builtin_ia32_sqrtss_mask_round): New builtins.
1635 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
1636 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
1637 ((match_operand:VF_128 1 "vector_operand"
1638 "xBm,<round_constraint>")): Changed to ...
1639 ((match_operand:VF_128 1 "vector_operand"
1640 "xBm,<round_scalar_constraint>")): ... this.
1641 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
1642 %0, %2, %<iptr>1<round_op3>}): Changed to ...
1643 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
1644 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
1645 %<iptr>1<round_scalar_mask_op3>}): ... this.
1646 ((set_attr "prefix" "<round_prefix>")): Changed to ...
1647 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
1649 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
1652 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
1653 Cast vec_cmpeq result to correct type.
1654 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
1655 Cast vec_cmpgt result to correct type.
1657 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
1659 * final.c (final_scan_insn_1): Renamed from...
1660 (final_scan_insn): ... this. New wrapper, to recover
1661 seen from the outermost call in recursive ones.
1662 * config/sparc/sparc.c (output_return): Drop seen from call.
1663 (output_sibcall): Likewise.
1664 * config/visium/visium.c (output_branch): Likewise.
1666 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
1668 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
1671 2018-02-10 Alan Modra <amodra@gmail.com>
1674 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
1675 Specify LR as an input.
1677 2018-02-10 Jakub Jelinek <jakub@redhat.com>
1680 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
1681 remove_member_access_dummy_vars): New functions.
1682 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
1683 lower_omp_1, execute_lower_omp): Use them.
1685 PR rtl-optimization/84308
1686 * shrink-wrap.c (spread_components): Release todo vector.
1688 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
1690 PR rtl-optimization/57193
1691 * ira-color.c (struct allocno_color_data): Add member
1692 conflict_allocno_hard_prefs.
1693 (update_conflict_allocno_hard_prefs): New.
1694 (bucket_allocno_compare_func): Add a preference based on
1695 conflict_allocno_hard_prefs.
1696 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
1697 (color_allocnos): Remove a dead code. Initiate
1698 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
1700 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1703 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
1704 constraint from =wa to wa. Avoid a subreg on the output operand,
1705 instead use a pseudo and subreg it in a move.
1706 (p9_xxbrd_<mode>): Changed to ...
1707 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
1708 (p9_xxbrd_v2df): New expander.
1709 (p9_xxbrw_<mode>): Changed to ...
1710 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
1711 (p9_xxbrw_v4sf): New expander.
1713 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1715 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
1717 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
1720 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
1721 multiply in 32-bit mode.
1722 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
1723 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
1726 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1728 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
1729 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
1730 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
1731 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
1733 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
1736 * dwarf2out.c (is_trivial_indirect_ref): New function.
1737 (dwarf2out_late_global_decl): Do not generate a location
1738 attribute for variables that have a non-trivial DECL_VALUE_EXPR
1739 and that are not defined in the current unit.
1741 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1743 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
1744 instead of a libcall for UNORDERED.
1746 2018-02-09 Tamar Christina <tamar.christina@arm.com>
1749 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
1750 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
1752 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1755 * config/s390/s390.c (s390_set_current_function): Invoke
1756 s390_indirect_branch_settings also if fndecl didn't change.
1758 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1760 * config/rs6000/rs6000.md (blockage): Set length to zero.
1762 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1764 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
1766 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1769 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
1770 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
1774 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
1775 PARALLEL incoming that failed vt_get_decl_and_offset check.
1778 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
1779 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
1781 (get_variable_section): For decls in named .bss* sections pass true as
1782 second argument to bss_initializer_p.
1784 2018-02-09 Marek Polacek <polacek@redhat.com>
1785 Jakub Jelinek <jakub@redhat.com>
1788 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
1789 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
1790 Sync some changes from cxx_fold_indirect_ref.
1792 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1794 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
1796 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
1797 (BLOCK_INLINE_ENTRY_LABEL): New.
1798 (dwarf2out_var_location): Disregard inline entry markers.
1799 (inline_entry_data): New struct.
1800 (inline_entry_data_hasher): New hashtable type.
1801 (inline_entry_data_hasher::hash): New.
1802 (inline_entry_data_hasher::equal): New.
1803 (inline_entry_data_table): New variable.
1804 (add_high_low_attributes): Add DW_AT_entry_pc and
1805 DW_AT_GNU_entry_view attributes if a pending entry is found
1806 in inline_entry_data_table. Add old entry_pc attribute only
1807 if debug nonbinding markers are disabled.
1808 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
1809 markers are enabled.
1810 (block_within_block_p, dwarf2out_inline_entry): New.
1811 (dwarf2out_finish): Check that no entries remained in
1812 inline_entry_data_table.
1813 * final.c (reemit_insn_block_notes): Handle inline entry notes.
1814 (final_scan_insn, notice_source_line): Likewise.
1815 (rest_of_clean_state): Skip inline entry markers.
1816 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
1818 * gimple.c (gimple_build_debug_inline_entry): New.
1819 * gimple.h (enum gimple_debug_subcode): Add
1820 GIMPLE_DEBUG_INLINE_ENTRY.
1821 (gimple_build_debug_inline_entry): Declare.
1822 (gimple_debug_inline_entry_p): New.
1823 (gimple_debug_nonbind_marker_p): Adjust.
1824 * insn-notes.def (INLINE_ENTRY): New.
1825 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
1826 inline entry marker notes.
1827 (print_insn): Likewise.
1828 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
1829 (INSN_DEBUG_MARKER_KIND): Likewise.
1830 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
1831 * tree-inline.c (expand_call_inline): Build and insert
1832 debug_inline_entry stmt.
1833 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
1834 inline entry blocks early, if nonbind markers are enabled.
1835 (dump_scope_block): Dump fragment info.
1836 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
1837 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
1838 (gimple_build_debug_inline_entry): New.
1839 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
1840 Enable/disable inline entry points too.
1841 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
1842 (DEBUG_INSN): Describe inline entry markers.
1844 * common.opt (gvariable-location-views): New.
1845 (gvariable-location-views=incompat5): New.
1846 * config.in: Rebuilt.
1847 * configure: Rebuilt.
1848 * configure.ac: Test assembler for view support.
1849 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
1850 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
1851 * dwarf2out.c (var_loc_view): New typedef.
1852 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
1853 (dwarf2out_locviews_in_attribute): New.
1854 (dwarf2out_locviews_in_loclist): New.
1855 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
1856 (enum dw_line_info_opcode): Add LI_adv_address.
1857 (struct dw_line_info_table): Add view.
1858 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
1859 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
1860 (zero_view_p): New variable.
1861 (ZERO_VIEW_P): New macro.
1862 (output_asm_line_debug_info): New.
1863 (struct var_loc_node): Add view.
1864 (add_AT_view_list, AT_loc_list): New.
1865 (add_var_loc_to_decl): Add view param. Test it against last.
1866 (new_loc_list): Add view params. Record them.
1867 (AT_loc_list_ptr): Handle loc and view lists.
1868 (view_list_to_loc_list_val_node): New.
1869 (print_dw_val): Handle dw_val_class_view_list.
1870 (size_of_die): Likewise.
1871 (value_format): Likewise.
1872 (loc_list_has_views): New.
1873 (gen_llsym): Set vl_symbol too.
1874 (maybe_gen_llsym, skip_loc_list_entry): New.
1875 (dwarf2out_maybe_output_loclist_view_pair): New.
1876 (output_loc_list): Output view list or entries too.
1877 (output_view_list_offset): New.
1878 (output_die): Handle dw_val_class_view_list.
1879 (output_dwarf_version): New.
1880 (output_compilation_unit_header): Use it.
1881 (output_skeleton_debug_sections): Likewise.
1882 (output_rnglists, output_line_info): Likewise.
1883 (output_pubnames, output_aranges): Update version comments.
1884 (output_one_line_info_table): Output view numbers in asm comments.
1885 (dw_loc_list): Determine current endview, pass it to new_loc_list.
1886 Call maybe_gen_llsym.
1887 (loc_list_from_tree_1): Adjust.
1888 (add_AT_location_description): Create view list attribute if
1889 needed, check it's absent otherwise.
1890 (convert_cfa_to_fb_loc_list): Adjust.
1891 (maybe_emit_file): Call output_asm_line_debug_info for test.
1892 (dwarf2out_var_location): Reset views as needed. Precompute
1893 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
1894 attribute. Set view.
1895 (new_line_info_table): Reset next view.
1896 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
1897 (dwarf2out_source_line): Likewise. Output view resets and labels to
1898 the assembler, or select appropriate line info opcodes.
1899 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
1900 (optimize_string_length): Catch it. Adjust.
1901 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
1902 dw_val_class_view_list, and remove it if no longer needed.
1903 (hash_loc_list): Hash view numbers.
1904 (loc_list_hasher::equal): Compare them.
1905 (optimize_location_lists): Check whether a view list symbol is
1906 needed, and whether the locview attribute is present, and
1907 whether they match. Remove the locview attribute if no longer
1909 (index_location_lists): Call skip_loc_list_entry for test.
1910 (dwarf2out_finish): Call output_asm_line_debug_info for test.
1911 Use output_dwarf_version.
1912 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
1913 (struct dw_val_node): Add val_view_list.
1914 * final.c (SEEN_NEXT_VIEW): New.
1915 (set_next_view_needed): New.
1916 (clear_next_view_needed): New.
1917 (maybe_output_next_view): New.
1918 (final_start_function): Rename to...
1919 (final_start_function_1): ... this. Take pointer to FIRST,
1920 add SEEN parameter. Emit param bindings in the initial view.
1921 (final_start_function): Reintroduce SEEN-less interface.
1922 (final): Rename to...
1923 (final_1): ... this. Take SEEN parameter. Output final pending
1924 next view at the end.
1925 (final): Reintroduce seen-less interface.
1926 (final_scan_insn): Output pending next view before switching
1927 sections or ending a block. Mark the next view as needed when
1928 outputting variable locations. Notify debug backend of section
1929 changes, and of location view changes.
1930 (rest_of_handle_final): Adjust.
1931 * toplev.c (process_options): Autodetect value for debug variable
1932 location views option. Warn on incompat5 without -gdwarf-5.
1933 * doc/invoke.texi (gvariable-location-views): New.
1934 (gvariable-location-views=incompat5): New.
1935 (gno-variable-location-views): New.
1937 2018-02-08 David Malcolm <dmalcolm@redhat.com>
1939 PR tree-optimization/84136
1940 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
1941 that the result of find_edge is non-NULL.
1943 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1946 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
1947 storing integer register in SImode. Fix cost of 256 and 512
1948 byte aligned SSE register store.
1950 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1952 * config/i386/i386.c (ix86_multiplication_cost): Fix
1953 multiplication cost for TARGET_AVX512DQ.
1955 2018-02-08 Marek Polacek <polacek@redhat.com>
1957 PR tree-optimization/84238
1958 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
1961 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1963 PR tree-optimization/84265
1964 * tree-vect-stmts.c (vectorizable_store): Don't treat
1965 VMAT_CONTIGUOUS accesses as grouped.
1966 (vectorizable_load): Likewise.
1968 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1970 PR tree-optimization/81635
1971 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
1972 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
1973 (test_round_for_mask): New functions.
1974 (wide_int_cc_tests): Call test_round_for_mask.
1975 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
1976 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
1977 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
1978 range returned by get_range_info.
1980 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1983 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
1984 * symtab.c: Include builtins.h
1985 (symtab_node::output_to_lto_symbol_table_p): Move here
1986 from lto-streamer-out.c:output_symbol_p.
1987 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
1988 (output_symbol_p): Move all logic to symtab.c
1989 (produce_symtab): Update.
1991 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1993 * config/s390/s390-opts.h (enum indirect_branch): Define.
1994 * config/s390/s390-protos.h (s390_return_addr_from_memory)
1995 (s390_indirect_branch_via_thunk)
1996 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
1997 (enum s390_indirect_branch_type): Define.
1998 * config/s390/s390.c (struct s390_frame_layout, struct
1999 machine_function): Remove.
2000 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
2001 (indirect_branch_table_label_no, indirect_branch_table_name):
2003 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
2004 (enum s390_indirect_branch_option): Define.
2005 (s390_return_addr_from_memory): New function.
2006 (s390_handle_string_attribute): New function.
2007 (s390_attribute_table): Add new attribute handler.
2008 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
2009 (s390_indirect_branch_via_thunk): New function.
2010 (s390_indirect_branch_via_inline_thunk): New function.
2011 (s390_function_ok_for_sibcall): When jumping via thunk disallow
2012 sibling call optimization for non z10 compiles.
2013 (s390_emit_call): Force indirect branch target to be a single
2014 register. Add r1 clobber for non-z10 compiles.
2015 (s390_emit_epilogue): Emit return jump via return_use expander.
2016 (s390_reorg): Handle JUMP_INSNs as execute targets.
2017 (s390_option_override_internal): Perform validity checks for the
2018 new command line options.
2019 (s390_indirect_branch_attrvalue): New function.
2020 (s390_indirect_branch_settings): New function.
2021 (s390_set_current_function): Invoke s390_indirect_branch_settings.
2022 (s390_output_indirect_thunk_function): New function.
2023 (s390_code_end): Implement target hook.
2024 (s390_case_values_threshold): Implement target hook.
2025 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
2027 * config/s390/s390.h (struct s390_frame_layout)
2028 (struct machine_function): Move here from s390.c.
2029 (TARGET_INDIRECT_BRANCH_NOBP_RET)
2030 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
2031 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
2032 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
2033 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
2034 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
2035 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
2036 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
2037 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
2038 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
2039 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
2040 (mnemonic attribute): Add values which aren't recognized
2042 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
2043 pattern for branch conversion. Fix mnemonic attribute.
2044 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
2045 indirect branch via thunk if requested.
2046 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
2047 ("*indirect_jump"): Disable for branch conversion using out of
2049 ("indirect_jump_via_thunk<mode>_z10")
2050 ("indirect_jump_via_thunk<mode>")
2051 ("indirect_jump_via_inlinethunk<mode>_z10")
2052 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
2053 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
2054 ("casesi_jump_via_inlinethunk<mode>_z10")
2055 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
2056 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
2057 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
2058 ("*indirect2_jump"): Disable for branch conversion.
2059 ("casesi_jump"): Turn into expander and expand patterns for branch
2061 ("return_use"): New expander.
2062 ("*return"): Emit return via thunk and rename it to ...
2063 ("*return<mode>"): ... this one.
2064 * config/s390/s390.opt: Add new options and and enum for the
2067 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2069 * lra-constraints.c (match_reload): Unconditionally use
2070 gen_lowpart_SUBREG, rather than selecting between that
2071 and equivalent gen_rtx_SUBREG code.
2073 2018-02-08 Richard Biener <rguenther@suse.de>
2075 PR tree-optimization/84233
2076 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
2077 changed flag instead of boguously re-using phi_inserted.
2079 2018-02-08 Martin Jambor <mjambor@suse.cz>
2081 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
2082 static local variables.
2084 2018-02-08 Richard Biener <rguenther@suse.de>
2086 PR tree-optimization/84278
2087 * tree-vect-stmts.c (vectorizable_store): When looking for
2088 smaller vector types to perform grouped strided loads/stores
2089 make sure the mode is supported by the target.
2090 (vectorizable_load): Likewise.
2092 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2094 * config/aarch64/aarch64.c (aarch64_components_for_bb):
2095 Increase LDP/STP opportunities by adding adjacent callee-saves.
2097 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2099 PR rtl-optimization/84068
2100 PR rtl-optimization/83459
2101 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
2103 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
2105 PR tree-optimization/84224
2106 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
2107 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
2110 2018-02-07 Iain Sandoe <iain@codesourcery.com>
2113 * config/rs6000/altivec.md (*restore_world): Remove LR use.
2114 * config/rs6000/predicates.md (restore_world_operation): Adjust op
2115 count, remove one USE.
2117 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
2119 * doc/install.texi (Configuration): Document the
2120 --with-long-double-format={ibm,ieee} PowerPC configuration
2124 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
2125 Convert from define_expand to be define_insn_and_split. Rework
2126 float/double/_Float128 conversions to QI/HI/SImode to work with
2127 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
2128 conversions to QI/HImode types did a store and then a load to
2129 truncate the value. For conversions to VSX registers, don't split
2130 the insn, instead emit the code directly. Use the code iterator
2131 any_fix to combine signed and unsigned conversions.
2132 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
2133 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
2134 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2135 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
2136 (fix_<mode>di2_hw): Likewise.
2137 (fixuns_<mode>di2_hw): Likewise.
2138 (fix_<mode>si2_hw): Likewise.
2139 (fixuns_<mode>si2_hw): Likewise.
2140 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
2141 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2142 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
2143 fix<uns>_trunc<SFDF:mode>si2_p8.
2144 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
2146 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
2147 (fix<uns>_<mode>_mem): Likewise.
2148 (fctiw<u>z_<mode>_mem): Likewise.
2149 (fix<uns>_<mode>_mem): Likewise.
2150 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
2151 the register allocator from doing a direct move to the GPRs to do
2152 a store, and instead use the ISA 3.0 store byte/half-word from
2153 vector register instruction. For IEEE 128-bit floating point,
2154 also optimize stores of 32-bit ints.
2155 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
2157 2018-02-07 Alan Hayward <alan.hayward@arm.com>
2159 * genextract.c (push_pathstr_operand): New function to support
2161 (walk_rtx): Call push_pathstr_operand.
2162 (print_path): Support [a-zA-Z].
2164 2018-02-07 Richard Biener <rguenther@suse.de>
2166 PR tree-optimization/84037
2167 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
2168 (cse_and_gimplify_to_preheader): Declare.
2169 (vect_get_place_in_interleaving_chain): Likewise.
2170 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
2172 (_loop_vec_info::~_loop_vec_info): Delete it.
2173 (cse_and_gimplify_to_preheader): New function.
2174 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
2175 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
2176 (vectorizable_load): Likewise. For grouped stores always base
2177 the IV on the first element.
2178 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
2179 condition before gimplifying.
2181 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2183 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
2184 *DIV_EXPR and *MOD_EXPR.
2186 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
2189 * config/i386/i386.c (ix86_option_override_internal): Mask out
2190 the CF_SET bit when checking -fcf-protection.
2192 2018-02-07 Tom de Vries <tom@codesourcery.com>
2195 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
2198 2018-02-07 Richard Biener <rguenther@suse.de>
2200 PR tree-optimization/84204
2201 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
2204 PR tree-optimization/84205
2205 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
2206 special-case isl_ast_op_zdiv_r.
2208 PR tree-optimization/84223
2209 * graphite-scop-detection.c (gather_bbs::before_dom_children):
2210 Only add conditions from within the region.
2211 (gather_bbs::after_dom_children): Adjust.
2213 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
2216 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
2217 * config/avr/avr.md: Only post-reload split REG-REG moves if
2218 either register is GENERAL_REG_P.
2220 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2222 PR tree-optimization/84235
2223 * tree-ssa-scopedtables.c
2224 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
2225 if the subtraction is performed in floating point type where NaNs are
2226 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
2227 build 1. Formatting fix.
2229 2018-02-06 Jakub Jelinek <jakub@redhat.com>
2232 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
2233 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
2234 and skip it regardless of bb boundaries. Use CALL_P macro,
2235 don't test INSN_P (insn) together with CALL_P or JUMP_P check
2236 unnecessarily, formatting fix.
2238 2018-02-06 Michael Collison <michael.collison@arm.com>
2240 * config/arm/thumb2.md:
2241 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
2242 (*thumb_mov_notscc): Ditto.
2244 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
2247 * config/rs6000/rs6000.md (su code attribute): Use "u" for
2248 unsigned_fix, not "s".
2250 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2252 * configure.ac (gcc_fn_eh_frame_ro): New function.
2253 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
2254 correct .eh_frame permissions.
2255 * configure: Regenerate.
2257 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
2259 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
2262 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2264 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2265 Display warning message for -mno-speculate-indirect-jumps.
2267 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
2269 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
2271 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
2273 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
2275 PR tree-optimization/84225
2276 * tree-eh.c (find_trapping_overflow): Only call
2277 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
2279 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2282 * config/i386/i386.c: Reimplement the check of possible options
2283 -mibt/-mshstk conbination. Change error messages.
2284 * doc/invoke.texi: Fix a typo: remove extra '='.
2286 2018-02-06 Marek Polacek <polacek@redhat.com>
2288 PR tree-optimization/84228
2289 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
2291 2018-02-06 Tamar Christina <tamar.christina@arm.com>
2294 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
2295 emitted arch directives.
2296 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
2297 __ARM_FEATURE_COPROC before changing architectures.
2299 2018-02-06 Richard Biener <rguenther@suse.de>
2301 * config/i386/i386.c (print_reg): Fix typo.
2302 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
2304 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
2306 * configure: Regenerate.
2308 2018-02-05 Martin Sebor <msebor@redhat.com>
2310 PR tree-optimization/83369
2311 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
2314 2018-02-05 Martin Liska <mliska@suse.cz>
2316 * doc/invoke.texi: Cherry-pick upstream r323995.
2318 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
2320 * ira.c (ira_init_register_move_cost): Adjust comment.
2322 2018-02-05 Martin Liska <mliska@suse.cz>
2324 PR gcov-profile/84137
2325 * doc/gcov.texi: Fix typo in documentation.
2327 2018-02-05 Martin Liska <mliska@suse.cz>
2329 PR gcov-profile/83879
2330 * doc/gcov.texi: Document necessity of --dynamic-list-data when
2331 using dlopen functionality.
2333 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
2335 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
2336 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
2337 _mm_maskz_range_ss, _mm_mask_range_round_ss,
2338 _mm_maskz_range_round_ss): New intrinsics.
2339 (__builtin_ia32_rangesd128_round)
2340 (__builtin_ia32_rangess128_round): Remove.
2341 (__builtin_ia32_rangesd128_mask_round,
2342 __builtin_ia32_rangess128_mask_round): New builtins.
2343 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
2344 __builtin_ia32_rangess128_round): Remove.
2345 (__builtin_ia32_rangesd128_mask_round,
2346 __builtin_ia32_rangess128_mask_round): New builtins.
2347 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
2348 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
2349 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
2350 "<round_saeonly_constraint>")): Changed to ...
2351 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
2352 "<round_saeonly_scalar_constraint>")): ... this.
2353 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
2354 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
2355 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
2356 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
2357 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
2359 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
2361 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
2363 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
2364 Remove all values except native, 8540 and 8548.
2366 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
2368 * config/i386/i386.c (ix86_output_function_return): Pass
2369 INVALID_REGNUM, instead of -1, as invalid register number to
2370 indirect_thunk_name and output_indirect_thunk.
2372 2018-02-02 Julia Koval <julia.koval@intel.com>
2374 * config.gcc: Add -march=icelake.
2375 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
2376 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
2377 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
2378 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
2379 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
2380 (processor_target_table): Add icelake.
2381 (ix86_option_override_internal): Handle new PTAs.
2382 (get_builtin_code_for_version): Handle icelake.
2383 (M_INTEL_COREI7_ICELAKE): New.
2384 (fold_builtin_cpu): Handle icelake.
2385 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
2386 * doc/invoke.texi: Add -march=icelake.
2388 2018-02-02 Julia Koval <julia.koval@intel.com>
2390 * config/i386/i386.c (ix86_option_override_internal): Change flags type
2391 to wide_int_bitmask.
2392 * wide-int-bitmask.h: New.
2394 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2397 * config/i386/i386.md: Replace Pmode with word_mode in
2398 builtin_setjmp_setup and builtin_longjmp to support x32.
2400 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
2404 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
2406 (rs6000_supported_cpu_names): New static variable.
2407 (linux_cpu_translation_table): Likewise.
2408 (elf_platform) <cpu>: Define new static variable and use it.
2409 Translate kernel AT_PLATFORM name to canonical name if needed.
2410 Error if platform name is unknown.
2412 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
2415 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
2417 2018-02-01 Jeff Law <law@redhat.com>
2420 * config/i386/i386.c (release_scratch_register_on_entry): Add new
2421 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
2422 the scratch if RELEASE_VIA_POP is false.
2423 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
2424 If we have to save a temporary register, decrement SIZE appropriately.
2425 Pass new arguments to release_scratch_register_on_entry.
2426 (ix86_adjust_stack_and_probe): Likewise.
2427 (ix86_emit_probe_stack_range): Pass new arguments to
2428 release_scratch_register_on_entry.
2430 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
2432 PR rtl-optimization/84157
2433 * combine.c (change_zero_ext): Use REG_P predicate in
2434 front of HARD_REGISTER_P predicate.
2436 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
2438 * config/avr/avr.c (avr_option_override): Move disabling of
2439 -fdelete-null-pointer-checks to...
2440 * common/config/avr/avr-common.c (avr_option_optimization_table):
2443 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2445 PR tree-optimization/81635
2446 * tree-data-ref.c (split_constant_offset_1): For types that
2447 wrap on overflow, try to use range info to prove that wrapping
2450 2018-02-01 Renlin Li <renlin.li@arm.com>
2453 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
2455 (aarch64_register_move_cost): Likewise.
2456 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
2458 (REG_CLASS_NAMES): Likewise.
2459 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
2460 TAILCALL_ADDR_REGS. Remove IP registers.
2461 * config/aarch64/aarch64.md (Ucs): Update register constraint.
2463 2018-02-01 Richard Biener <rguenther@suse.de>
2465 * domwalk.h (dom_walker::dom_walker): Add additional constructor
2466 for specifying RPO order and allow NULL for that.
2467 * domwalk.c (dom_walker::dom_walker): Likewise.
2468 (dom_walker::walk): Handle NULL RPO order.
2469 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
2471 (rewrite_update_dom_walker): Likewise.
2472 (mark_def_dom_walker): Likewise.
2474 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2476 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
2477 (aarch64_maybe_expand_sve_subreg_move): Declare.
2478 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
2479 * config/aarch64/predicates.md (aarch64_any_register_operand): New
2481 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
2482 that are semantically a reverse operation.
2483 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
2484 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
2485 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
2487 (aarch64_can_change_mode_class): For big-endian, forbid changes
2488 between two SVE modes if they have different element sizes.
2490 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2492 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
2493 the TImode handling for big-endian targets.
2495 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2497 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
2498 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
2500 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
2501 Remove BSWAP handing for big-endian targets and use the form of
2502 LD1RQ appropariate for the mode.
2504 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2506 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
2507 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
2510 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2513 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
2514 check for operands that need to go through aarch64_sve_reload_be.
2516 2018-02-01 Jakub Jelinek <jakub@redhat.com>
2518 PR tree-optimization/81661
2519 PR tree-optimization/84117
2520 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
2521 * tree-eh.c: Include gimplify.h.
2522 (find_trapping_overflow, replace_trapping_overflow,
2523 rewrite_to_non_trapping_overflow): New functions.
2524 * tree-vect-loop.c: Include tree-eh.h.
2525 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
2526 * tree-data-ref.c: Include tree-eh.h.
2527 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
2529 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
2531 PR rtl-optimization/84123
2532 * combine.c (change_zero_ext): Check if hard register satisfies
2533 can_change_dest_mode before calling gen_lowpart_SUBREG.
2535 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
2538 * ira.c (ira_init_register_move_cost): Remove assert.
2540 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2542 PR rtl-optimization/84071
2543 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
2544 * doc/tm.texi: Regenerate.
2546 2018-01-31 Richard Biener <rguenther@suse.de>
2548 PR tree-optimization/84132
2549 * tree-data-ref.c (analyze_miv_subscript): Properly
2550 check whether evolution_function_is_affine_multivariate_p
2551 before calling gcd_of_steps_may_divide_p.
2553 2018-01-31 Julia Koval <julia.koval@intel.com>
2556 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
2557 * config/i386/i386.md (rdpid_rex64) New.
2558 (rdpid): Make 32bit only.
2560 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
2563 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
2564 an IDENTIFIER_NODE for FUNCTION_TYPE's.
2566 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2569 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
2571 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
2573 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2575 PR rtl-optimization/84071
2576 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
2577 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
2579 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2581 * config/arc/arc.c (arc_handle_aux_attribute): New function.
2582 (arc_attribute_table): Add 'aux' attribute.
2583 (arc_in_small_data_p): Consider aux like variables.
2584 (arc_is_aux_reg_p): New function.
2585 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
2586 (arc_get_aux_arg): New function.
2587 (prepare_move_operands): Handle aux-register access.
2588 (arc_handle_aux_attribute): New function.
2589 * doc/extend.texi (ARC Variable attributes): Add subsection.
2591 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2593 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
2594 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
2595 (arc_attribute_table): Add 'uncached' attribute.
2596 (arc_print_operand): Print '.di' flag for uncached memory
2598 (arc_in_small_data_p): Do not consider for small data the uncached
2600 (arc_is_uncached_mem_p): New function.
2601 * config/arc/predicates.md (compact_store_memory_operand): Check
2602 for uncached memory accesses.
2603 (nonvol_nonimm_operand): Likewise.
2604 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
2606 2018-01-31 Jakub Jelinek <jakub@redhat.com>
2609 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
2610 falign-loops=): Add Optimization flag.
2612 2018-01-30 Jeff Law <law@redhat.com>
2615 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
2616 INT_REGISTERS_SAVED. Check it prior to calling
2617 get_scratch_register_on_entry.
2618 (ix86_adjust_stack_and_probe): Similarly.
2619 (ix86_emit_probe_stack_range): Similarly.
2620 (ix86_expand_prologue): Corresponding changes.
2622 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2625 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
2626 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
2628 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
2631 * lra-constraints.c (curr_insn_transform): Process AND in the
2634 2018-01-30 Jakub Jelinek <jakub@redhat.com>
2636 PR rtl-optimization/83986
2637 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
2638 dependence against last_pending_memory_flush in addition to
2641 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
2643 PR tree-optimization/81611
2644 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
2647 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2650 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
2653 2018-01-30 Richard Biener <rguenther@suse.de>
2654 Jakub Jelinek <jakub@redhat.com>
2656 PR tree-optimization/84111
2657 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
2658 inner loops added during recursion, as they don't have up-to-date
2661 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2664 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
2665 (can_inline_edge_by_limits_p): ... here.
2666 (can_early_inline_edge_p, check_callers,
2667 update_caller_keys, update_callee_keys, recursive_inlining,
2668 add_new_edges_to_heap, speculation_useful_p,
2669 inline_small_functions,
2670 inline_small_functions, flatten_function,
2671 inline_to_all_callers_1): Update.
2673 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2675 * profile-count.c (profile_count::combine_with_ipa_count): Handle
2678 2018-01-30 Richard Biener <rguenther@suse.de>
2680 PR tree-optimization/83008
2681 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
2682 invariant and constant vector uses in stmts when they need
2685 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2688 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
2689 * configure: Regenerate.
2691 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2693 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
2695 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
2696 Use gen_rtx_REG rather than gen_lowpart.
2698 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2700 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
2701 rather than 0 when creating partial subregs.
2703 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2705 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
2708 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
2711 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
2712 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
2713 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
2714 flags. This restores the settings used before the 2017-07-24.
2715 Turning off pre increment/decrement/modify allows IVOPTS to
2716 optimize DF/SF loops where the index is an int.
2718 2018-01-29 Richard Biener <rguenther@suse.de>
2719 Kelvin Nilsen <kelvin@gcc.gnu.org>
2722 * tree-vect-stmts.c (vectorizable_call): Don't call
2723 targetm.vectorize_builtin_md_vectorized_function if callee is
2726 2018-01-22 Carl Love <cel@us.ibm.com>
2728 * doc/extend.tex: Fix typo in second arg in
2729 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
2731 2018-01-29 Richard Biener <rguenther@suse.de>
2733 PR tree-optimization/84086
2734 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
2735 (flush_ssaname_freelist): When SSA names were released reset
2736 the SCEV hash table.
2738 2018-01-29 Richard Biener <rguenther@suse.de>
2740 PR tree-optimization/84057
2741 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
2742 removed paths when removing edges.
2744 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
2746 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
2747 -mfunction-return=@var{choice}.
2749 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
2752 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
2754 (layout::print_source_line): Likewise.
2755 (test_get_line_width_without_trailing_whitespace): Add test cases.
2757 2018-01-27 Jakub Jelinek <jakub@redhat.com>
2760 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
2763 2018-01-26 Jim Wilson <jimw@sifive.com>
2765 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
2767 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
2770 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2772 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
2773 and CMP + SUB-immediate -> SUBS.
2775 2018-01-26 Martin Sebor <msebor@redhat.com>
2777 PR tree-optimization/83896
2778 * tree-ssa-strlen.c (get_string_len): Rename...
2779 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
2780 Avoid assuming length is constant.
2781 (handle_char_store): Use HOST_WIDE_INT for string length.
2783 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
2786 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
2787 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
2789 2018-01-26 Richard Biener <rguenther@suse.de>
2791 PR rtl-optimization/84003
2792 * dse.c (record_store): Only record redundant stores when
2793 the earlier store aliases at least all accesses the later one does.
2795 2018-01-26 Jakub Jelinek <jakub@redhat.com>
2797 PR rtl-optimization/83985
2798 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
2799 REG_CFA_RESTORE insns.
2800 (delete_unmarked_insns): Don't ignore separate shrink wrapping
2801 REG_CFA_RESTORE insns here.
2804 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
2805 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
2807 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2809 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
2810 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
2811 (arc_init): Likewise.
2812 (arc_override_options): Likewise.
2813 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
2815 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
2817 * config/arc/arc.h (TARGET_DBNZ): Define.
2818 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
2819 properly set the tune attribute.
2820 (dbnz): Use TARGET_DBNZ guard.
2821 * config/arc/arc.opt (mtune): Add core3 option.
2823 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2825 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
2826 recognize new pic like addresses.
2827 (arc_delegitimize_address): Clean up.
2829 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2831 * config/arc/arc-arches.def: Option mrf16 valid for all
2833 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
2834 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
2835 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
2836 * config/arc/arc-tables.opt: Regenerate.
2837 * config/arc/arc.c (arc_conditional_register_usage): Handle
2838 reduced register file case.
2839 (arc_file_start): Set must have build attributes.
2840 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
2842 * config/arc/arc.opt (mrf16): Add new option.
2843 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
2844 * config/arc/genmultilib.awk: Handle new mrf16 option.
2845 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
2846 * config/arc/t-multilib: Regenerate.
2847 * doc/invoke.texi (ARC Options): Document mrf16 option.
2849 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2851 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
2852 * config/arc/arc.c (arc_handle_secure_attribute): New function.
2853 (arc_attribute_table): Add 'secure_call' attribute.
2854 (arc_print_operand): Print secure call operand.
2855 (arc_function_ok_for_sibcall): Don't optimize tail calls when
2857 (arc_is_secure_call_p): New function. * config/arc/arc.md
2858 (call_i): Add support for sjli instruction.
2859 (call_value_i): Likewise.
2860 * config/arc/constraints.md (Csc): New constraint.
2862 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2863 John Eric Martin <John.Martin@emmicro-us.com>
2865 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
2866 * config/arc/arc.c (_arc_jli_section): New struct.
2867 (arc_jli_section): New type.
2868 (rc_jli_sections): New static variable.
2869 (arc_handle_jli_attribute): New function.
2870 (arc_attribute_table): Add jli_always and jli_fixed attribute.
2871 (arc_file_end): New function.
2872 (TARGET_ASM_FILE_END): Define.
2873 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
2874 (arc_add_jli_section): New function.
2875 (jli_call_scan): Likewise.
2876 (arc_reorg): Call jli_call_scan.
2877 (arc_output_addsi): Remove 'S' from printing asm operand.
2878 (arc_is_jli_call_p): New function.
2879 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
2881 (movhi_insn): Likewise.
2882 (movsi_insn): Likewise.
2883 (movsi_set_cc_insn): Likewise.
2884 (loadqi_update): Likewise.
2885 (load_zeroextendqisi_update): Likewise.
2886 (load_signextendqisi_update): Likewise.
2887 (loadhi_update): Likewise.
2888 (load_zeroextendhisi_update): Likewise.
2889 (load_signextendhisi_update): Likewise.
2890 (loadsi_update): Likewise.
2891 (loadsf_update): Likewise.
2892 (movsicc_insn): Likewise.
2893 (bset_insn): Likewise.
2894 (bxor_insn): Likewise.
2895 (bclr_insn): Likewise.
2896 (bmsk_insn): Likewise.
2897 (bicsi3_insn): Likewise.
2898 (cmpsi_cc_c_insn): Likewise.
2899 (movsi_ne): Likewise.
2900 (movsi_cond_exec): Likewise.
2901 (clrsbsi2): Likewise.
2912 (call_i): Remove 'S' asm letter, add jli instruction.
2913 (call_value_i): Likewise.
2914 * config/arc/arc.op (mjli-always): New option.
2915 * config/arc/constraints.md (Cji): New constraint.
2916 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
2918 (subsf3_fpx): Likewise.
2919 (mulsf3_fpx): Likewise.
2920 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
2922 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
2924 * doc/invoke.texi (ARC): Document mjli-always option.
2926 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
2928 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
2929 avoid addition with 0 and use incw and decw where possible.
2931 2018-01-26 Richard Biener <rguenther@suse.de>
2933 PR tree-optimization/81082
2934 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
2935 association if it requires casting to unsigned.
2936 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
2937 from fold_plusminus_mult_expr to catch important cases late when
2938 range info is available.
2940 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2942 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
2943 * configure.ac (hidden_linkonce): New test.
2944 * configure: Regenerate.
2945 * config.in: Regenerate.
2947 2018-01-26 Julia Koval <julia.koval@intel.com>
2949 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
2950 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
2951 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
2952 _mm_mask_bitshuffle_epi64_mask): Fix type.
2953 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
2954 USI_FTYPE_V4DI_V4DI_USI): Remove.
2955 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
2956 __builtin_ia32_vpshufbitqmb256_mask,
2957 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
2958 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
2959 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
2961 2018-01-26 Alan Modra <amodra@gmail.com>
2964 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
2965 UNSPEC_VBPERMQ. Sort other unspecs.
2967 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
2969 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
2971 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
2974 * predict.c (drop_profile): Do not push/pop cfun; update also
2976 (handle_missing_profiles): Fix logic looking for zero profiles.
2978 2018-01-25 Jakub Jelinek <jakub@redhat.com>
2981 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
2982 on functions with #pragma omp declare simd or functions with simd
2984 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
2985 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
2986 Remove trailing \n from warning_at calls.
2988 2018-01-25 Tom de Vries <tom@codesourcery.com>
2991 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
2992 for neutered workers.
2994 2018-01-24 Joseph Myers <joseph@codesourcery.com>
2997 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
2998 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
3000 2018-01-24 Jeff Law <law@redhat.com>
3003 * i386.c (get_probe_interval): Move to earlier point.
3004 (ix86_compute_frame_layout): If -fstack-clash-protection and
3005 the frame is larger than the probe interval, then use pushes
3006 to save registers rather than reg->mem moves.
3007 (ix86_expand_prologue): Remove conditional for int_registers_saved
3010 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
3013 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
3014 min/max for never referenced object.
3016 2018-01-24 Jakub Jelinek <jakub@redhat.com>
3019 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
3021 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
3022 attributes from DECL_ATTRIBUTES (decl) without affecting
3023 DECL_ATTRIBUTES (current_function_decl).
3024 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
3025 functions with non-NULL DECL_ABSTRACT_ORIGIN.
3027 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
3029 PR tree-optimization/83979
3030 * fold-const.c (fold_comparison): Use constant_boolean_node
3031 instead of boolean_{true,false}_node.
3033 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
3035 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
3038 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3040 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3041 Simplify the clause that sets the length attribute.
3042 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3043 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
3044 clause that sets the length attribute.
3045 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3047 2018-01-24 Tom de Vries <tom@codesourcery.com>
3050 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
3051 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
3052 Add strict parameter.
3053 (prevent_branch_around_nothing): Insert dummy insn between branch to
3054 label and label with no ptx insn inbetween.
3055 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
3057 2018-01-24 Tom de Vries <tom@codesourcery.com>
3060 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3061 for neutered threads in warp.
3062 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
3064 2018-01-24 Richard Biener <rguenther@suse.de>
3066 PR tree-optimization/83176
3067 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
3070 2018-01-24 Richard Biener <rguenther@suse.de>
3072 PR tree-optimization/82819
3073 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
3074 code generating pluses that are no-ops in the target precision.
3076 2018-01-24 Richard Biener <rguenther@suse.de>
3079 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
3081 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3083 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
3084 to merge probabilities.
3085 * predict.c (probably_never_executed): Also mark as cold functions
3086 with global 0 profile and guessed local profile.
3087 * profile-count.c (profile_probability::combine_with_count): New
3089 * profile-count.h (profile_probability::operator*,
3090 profile_probability::operator*=, profile_probability::operator/,
3091 profile_probability::operator/=): Reduce precision to adjusted
3092 and set value to guessed on contradictory divisions.
3093 (profile_probability::combine_with_freq): Remove.
3094 (profile_probability::combine_wiht_count): Declare.
3095 (profile_count::force_nonzero):: Set to adjusted.
3096 (profile_count::probability_in):: Set quality to adjusted.
3097 * tree-ssa-tail-merge.c (replace_block_by): Use
3100 2018-01-23 Andrew Waterman <andrew@sifive.com>
3101 Jim Wilson <jimw@sifive.com>
3103 * config/riscv/riscv.c (riscv_stack_boundary): New.
3104 (riscv_option_override): Set riscv_stack_boundary. Handle
3105 riscv_preferred_stack_boundary_arg.
3106 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
3107 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
3108 (STACK_BOUNDARY): Set to riscv_stack_boundary.
3109 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
3110 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
3111 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
3113 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
3116 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
3117 of struct ix86_frame.
3118 (ix86_expand_epilogue): Likewise. Add a local variable for
3119 the reg_save_offset field in struct ix86_frame.
3121 2018-01-23 Bin Cheng <bin.cheng@arm.com>
3123 PR tree-optimization/82604
3124 * tree-loop-distribution.c (enum partition_kind): New enum item
3125 PKIND_PARTIAL_MEMSET.
3126 (partition_builtin_p): Support above new enum item.
3127 (generate_code_for_partition): Ditto.
3128 (compute_access_range): Differentiate cases that equality can be
3129 proven at all loops, the innermost loops or no loops.
3130 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
3131 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
3132 (finalize_partitions, distribute_loop): Don't fuse partition of
3133 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
3134 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
3137 2018-01-23 Martin Liska <mliska@suse.cz>
3139 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
3140 order to ignore the predictor.
3141 (PRED_POLYMORPHIC_CALL): Likewise.
3142 (PRED_RECURSIVE_CALL): Likewise.
3144 2018-01-23 Martin Liska <mliska@suse.cz>
3146 * tree-profile.c (tree_profiling): Print function header to
3147 aware reader which function we are working on.
3148 * value-prof.c (gimple_find_values_to_profile): Do not print
3149 not interesting value histograms.
3151 2018-01-23 Martin Liska <mliska@suse.cz>
3153 * profile-count.h (enum profile_quality): Add
3154 profile_uninitialized as the first value. Do not number values
3155 as they are zero based.
3156 (profile_count::verify): Update sanity check.
3157 (profile_probability::verify): Likewise.
3159 2018-01-23 Nathan Sidwell <nathan@acm.org>
3161 * doc/invoke.texi (ffor-scope): Deprecate.
3163 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3165 PR tree-optimization/83510
3166 * domwalk.c (set_all_edges_as_executable): New function.
3167 (dom_walker::dom_walker): Convert bool param
3168 "skip_unreachable_blocks" to enum reachability. Move setup of
3169 edge flags to set_all_edges_as_executable and only do it when
3170 reachability is REACHABLE_BLOCKS.
3171 * domwalk.h (enum dom_walker::reachability): New enum.
3172 (dom_walker::dom_walker): Convert bool param
3173 "skip_unreachable_blocks" to enum reachability.
3174 (set_all_edges_as_executable): New decl.
3175 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
3176 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
3178 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
3179 but converting true to REACHABLE_BLOCKS.
3180 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
3182 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
3183 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
3184 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
3186 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
3187 if check_all_array_refs will be called.
3189 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3191 * tree.c (selftest::test_location_wrappers): Add more test
3194 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3196 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
3197 (selftest::test_bit_in_range): Likewise.
3199 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3202 * doc/sourcebuild.texi (vect_float): Say that the selector
3203 only describes the situation when -funsafe-math-optimizations is on.
3204 (vect_float_strict): Document.
3206 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3208 PR tree-optimization/83965
3209 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
3210 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
3211 instead of checking only for a reduction.
3212 (vect_recog_widen_sum_pattern): Likewise.
3214 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3216 * predict.c (probably_never_executed): Only use precise profile info.
3217 (compute_function_frequency): Skip after inlining hack since we now
3218 have quality checking.
3220 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3222 * profile-count.h (profile_probability::very_unlikely,
3223 profile_probability::unlikely, profile_probability::even): Set
3224 precision to guessed.
3226 2018-01-23 Richard Biener <rguenther@suse.de>
3228 PR tree-optimization/83963
3229 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
3230 Properly terminate dominator walk when crossing the exit edge not
3231 when visiting its source block.
3233 2018-01-23 Jakub Jelinek <jakub@redhat.com>
3236 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
3237 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
3239 2018-01-22 Jakub Jelinek <jakub@redhat.com>
3241 PR tree-optimization/83957
3242 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
3243 semicolon after for body surrounded by braces.
3245 PR tree-optimization/83081
3246 * profile-count.h (profile_probability::split): New method.
3247 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
3248 Use profile_probability::split.
3249 (do_compare_rtx_and_jump): Fix adjustment of probabilities
3250 when splitting a single conditional jump into 2.
3252 2018-01-22 David Malcolm <dmalcolm@redhat.com>
3254 PR tree-optimization/69452
3255 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
3258 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3260 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
3261 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
3262 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
3264 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3266 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
3267 * config/rl78/rl78.md (movdi): New define_expand.
3268 * config/rl78/rl78.c (rl78_split_movdi): New function.
3270 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
3273 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
3275 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
3276 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
3277 128-bit to produce an UNSPEC move to get the double word with the
3278 signbit and then a shift directly to do signbit.
3279 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
3280 implementation with a new version that just does either a direct
3281 move or a regular move. Move memory interface to separate insns.
3282 Move insns so they are next to the expander.
3283 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
3284 with signbit move. Split big and little endian case.
3285 (signbit<mode>2_dm_mem_le): Likewise.
3286 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
3287 (signbit<mode>2_dm2): Likewise.
3289 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3291 * config/rl78/rl78.md (anddi3): New define_expand.
3293 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3295 * config/rl78/rl78.md (umindi3): New define_expand.
3297 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3299 * config/rl78/rl78.md (smindi3): New define_expand.
3301 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3303 * config/rl78/rl78.md (smaxdi3): New define_expand.
3305 2018-01-22 Carl Love <cel@us.ibm.com>
3307 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
3308 LVX_V1TI): Add macro expansion.
3309 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
3310 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
3311 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
3312 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
3313 Change check to determine if the instruction is a byte reversing
3314 entry. Fix typo in comment.
3315 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
3316 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
3317 Add def_builtin calls for new builtins.
3318 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
3319 Add define_insn expansion.
3321 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3323 * config/rl78/rl78.md (umaxdi3): New define_expand.
3325 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3327 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
3328 for non-QImode registers.
3330 2018-01-22 Richard Biener <rguenther@suse.de>
3332 PR tree-optimization/83963
3333 * graphite-scop-detection.c (scop_detection::get_sese): Delay
3334 including the loop exit block.
3335 (scop_detection::merge_sese): Likewise.
3336 (scop_detection::add_scop): Do it here instead.
3338 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3340 * doc/sourcebuild.texi (arm_softfloat): Document.
3342 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
3345 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
3346 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
3347 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
3349 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3350 David Edelsohn <dje.gcc@gmail.com>
3353 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3354 Change "crset eq" to "crset 2".
3355 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3356 (*call_indirect_aix<mode>_nospec): Likewise.
3357 (*call_value_indirect_aix<mode>_nospec): Likewise.
3358 (*call_indirect_elfv2<mode>_nospec): Likewise.
3359 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
3360 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
3361 change assembly output from . to $.
3362 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3363 (indirect_jump<mode>_nospec): Change assembly output from . to $.
3364 (*tablejump<mode>_internal1_nospec): Likewise.
3366 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
3369 * config/sh/sh_optimize_sett_clrt.cc:
3370 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
3372 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
3374 PR tree-optimization/83940
3375 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
3376 offset_dt to vect_constant_def rather than vect_unknown_def_type.
3377 (vect_check_load_store_mask): Add a mask_dt_out parameter and
3378 use it to pass back the definition type.
3379 (vect_check_store_rhs): Likewise rhs_dt_out.
3380 (vect_build_gather_load_calls): Add a mask_dt argument and use
3381 it instead of a call to vect_is_simple_use.
3382 (vectorizable_store): Update calls to vect_check_load_store_mask
3383 and vect_check_store_rhs. Use the dt returned by the latter instead
3384 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
3385 instead of calls to vect_is_simple_use. Pass the scalar rather
3386 than the vector operand to vect_is_simple_use when handling
3387 second and subsequent copies of an rhs value.
3388 (vectorizable_load): Update calls to vect_check_load_store_mask
3389 and vect_build_gather_load_calls. Use the cached mask_dt and
3390 gs_info.offset_dt instead of calls to vect_is_simple_use.
3392 2018-01-20 Jakub Jelinek <jakub@redhat.com>
3395 * tree-emutls.c: Include gimplify.h.
3396 (lower_emutls_2): New function.
3397 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
3398 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
3399 it before further processing.
3402 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
3403 UINTVAL (trueop1) instead of INTVAL (op1).
3405 2018-01-19 Jakub Jelinek <jakub@redhat.com>
3409 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
3410 INCOMING_FRAME_SP_OFFSET if not defined.
3411 (scan_trace): Add ENTRY argument. If true and
3412 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
3413 emit a note to adjust the CFA offset.
3414 (create_cfi_notes): Adjust scan_trace callers.
3415 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
3416 INCOMING_FRAME_SP_OFFSET in the CIE.
3417 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
3418 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
3420 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
3421 * doc/tm.texi: Regenerated.
3423 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3425 PR rtl-optimization/83147
3426 * lra-constraints.c (remove_inheritance_pseudos): Use
3427 lra_substitute_pseudo_within_insn.
3429 2018-01-19 Tom de Vries <tom@codesourcery.com>
3430 Cesar Philippidis <cesar@codesourcery.com>
3433 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
3435 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
3438 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
3439 spaces for function labels.
3441 2018-01-19 Martin Liska <mliska@suse.cz>
3443 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
3444 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
3445 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
3446 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
3447 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
3448 (PRED_CONST_RETURN): Change from 69 to 65.
3449 (PRED_NULL_RETURN): Change from 91 to 71.
3450 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
3451 (PRED_LOOP_GUARD): Change from 66 to 73.
3453 2018-01-19 Martin Liska <mliska@suse.cz>
3455 * predict.c (predict_insn_def): Add new assert.
3456 (struct branch_predictor): Change type to signed integer.
3457 (test_prediction_value_range): Amend test to cover
3459 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
3460 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
3461 (PRED_LOOP_ITERATIONS_MAX): Likewise.
3462 (PRED_LOOP_IV_COMPARE): Likewise.
3463 * predict.h (PROB_UNINITIALIZED): Define new constant.
3465 2018-01-19 Martin Liska <mliska@suse.cz>
3467 * predict.c (dump_prediction): Add new format for
3468 analyze_brprob.py script which is enabled with -details
3470 * profile-count.h (precise_p): New function.
3472 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3474 PR tree-optimization/83922
3475 * tree-vect-loop.c (vect_verify_full_masking): Return false if
3476 there are no statements that need masking.
3477 (vect_active_double_reduction_p): New function.
3478 (vect_analyze_loop_operations): Use it when handling phis that
3479 are not in the loop header.
3481 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3483 PR tree-optimization/83914
3484 * tree-vect-loop.c (vectorizable_induction): Don't convert
3485 init_expr or apply the peeling adjustment for inductions
3486 that are nested within the vectorized loop.
3488 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3490 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
3493 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3497 * function.h (gimplify_parameters): Add gimple_seq * argument.
3498 * function.c: Include gimple.h and options.h.
3499 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
3500 for the added local temporaries if needed.
3501 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
3502 if there are any parameter cleanups, wrap whole body into a
3503 try/finally with the cleanups.
3505 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
3508 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
3509 Use GET_MODE_CLASS for scalar floating point.
3511 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3515 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
3516 Fix call of call_cgraph_insertion_hooks.
3518 2018-01-18 Martin Sebor <msebor@redhat.com>
3520 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
3522 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3525 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
3528 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
3531 * common.opt: (-ffile-prefix-map): New option.
3532 * opts.c (common_handle_option): Defer it.
3533 * opts-global.c (handle_common_deferred_options): Handle it.
3534 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
3535 * file-prefix-map.h: New file.
3536 (remap_debug_filename, add_debug_prefix_map): ...here.
3537 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
3538 * final.c (debug_prefix_map, add_debug_prefix_map
3539 remap_debug_filename): Move to...
3540 * file-prefix-map.c: New file.
3541 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
3542 generalize, get rid of alloca(), use strrchr() instead of strchr().
3543 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
3544 Implement in terms of add_prefix_map().
3545 (remap_macro_filename, remap_debug_filename): Implement in term of
3547 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
3548 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
3549 * dbxout.c: Include file-prefix-map.h.
3550 * varasm.c: Likewise.
3551 * vmsdbgout.c: Likewise.
3552 * xcoffout.c: Likewise.
3553 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
3554 * doc/cppopts.texi (-fmacro-prefix-map): Document.
3555 * doc/invoke.texi (-ffile-prefix-map): Document.
3556 (-fdebug-prefix-map): Update description.
3558 2018-01-18 Martin Liska <mliska@suse.cz>
3560 * config/i386/i386.c (indirect_thunk_name): Document that also
3562 (output_indirect_thunk): Document why both instructions
3563 (pause and lfence) are generated.
3565 2018-01-18 Richard Biener <rguenther@suse.de>
3567 PR tree-optimization/83887
3568 * graphite-scop-detection.c
3569 (scop_detection::get_nearest_dom_with_single_entry): Remove.
3570 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
3571 (scop_detection::merge_sese): Re-implement with a flood-fill
3572 algorithm that properly finds a SESE region if it exists.
3574 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3577 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
3578 pointer_diff optimizations use view_convert instead of convert.
3580 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3582 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3583 Generate different code for -mno-speculate-indirect-jumps.
3584 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3585 (*call_indirect_aix<mode>): Disable for
3586 -mno-speculate-indirect-jumps.
3587 (*call_indirect_aix<mode>_nospec): New define_insn.
3588 (*call_value_indirect_aix<mode>): Disable for
3589 -mno-speculate-indirect-jumps.
3590 (*call_value_indirect_aix<mode>_nospec): New define_insn.
3591 (*sibcall_nonlocal_sysv<mode>): Generate different code for
3592 -mno-speculate-indirect-jumps.
3593 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3595 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3597 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
3598 long double type, set the flags for noting the default long double
3599 type, even if we don't pass or return a long double type.
3601 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
3604 * ipa-inline.c (flatten_function): Do not overwrite final inlining
3607 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
3609 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
3610 support for merge[hl].
3611 (fold_mergehl_helper): New helper function.
3612 (tree-vector-builder.h): New #include for tree_vector_builder usage.
3613 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
3614 (altivec_vmrglw_direct): Add xxmrglw insn.
3616 2018-01-17 Andrew Waterman <andrew@sifive.com>
3618 * config/riscv/riscv.c (riscv_conditional_register_usage): If
3619 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
3621 2018-01-17 David Malcolm <dmalcolm@redhat.com>
3624 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
3625 call the lto_location_cache before reading the
3626 DECL_SOURCE_LOCATION of the types.
3628 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3629 Richard Sandiford <richard.sandiford@linaro.org>
3631 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
3632 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
3633 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
3634 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
3635 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
3637 * config/aarch64/constraints.md (aarch64_movti_operand):
3639 * config/aarch64/predicates.md (Uti): Add new constraint.
3641 2018-01-17 Carl Love <cel@us.ibm.com>
3643 * config/rs6000/vsx.md (define_expand xl_len_r,
3644 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
3645 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
3647 (define_expand, define_insn): Move the shift left from the
3648 define_insn to the define_expand for lxvl and stxvl instructions.
3649 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
3650 and XL_LEN_R definitions to PURE.
3652 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
3654 * config/i386/i386.c (indirect_thunk_name): Declare regno
3655 as unsigned int. Compare regno with INVALID_REGNUM.
3656 (output_indirect_thunk): Ditto.
3657 (output_indirect_thunk_function): Ditto.
3658 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
3659 in the call to output_indirect_thunk_function.
3661 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
3664 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
3665 rather than the size of inner_type to determine the stack slot size
3666 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
3668 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
3671 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
3674 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
3676 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
3677 endian Linux systems to optionally enable multilibs for selecting
3678 the long double type if the user configured an explicit type.
3679 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
3680 have no long double multilibs if not defined.
3681 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
3682 warn if the user used -mabi={ieee,ibm}longdouble and we built
3683 multilibs for long double.
3684 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
3685 appropriate multilib option.
3686 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
3688 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
3689 for building long double multilibs.
3690 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
3692 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
3694 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
3697 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
3699 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
3702 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
3705 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
3708 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3710 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
3711 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
3713 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
3715 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
3716 different rtl trees depending on TARGET_64BIT.
3717 (rs6000_gen_lvx): Likewise.
3719 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3721 * config/visium/visium.md (nop): Tweak comment.
3722 (hazard_nop): Likewise.
3724 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3726 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
3727 -mspeculate-indirect-jumps.
3728 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
3729 for -mno-speculate-indirect-jumps.
3730 (*call_indirect_elfv2<mode>_nospec): New define_insn.
3731 (*call_value_indirect_elfv2<mode>): Disable for
3732 -mno-speculate-indirect-jumps.
3733 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
3734 (indirect_jump): Emit different RTL for
3735 -mno-speculate-indirect-jumps.
3736 (*indirect_jump<mode>): Disable for
3737 -mno-speculate-indirect-jumps.
3738 (*indirect_jump<mode>_nospec): New define_insn.
3739 (tablejump): Emit different RTL for
3740 -mno-speculate-indirect-jumps.
3741 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
3742 (tablejumpsi_nospec): New define_expand.
3743 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
3744 (tablejumpdi_nospec): New define_expand.
3745 (*tablejump<mode>_internal1): Disable for
3746 -mno-speculate-indirect-jumps.
3747 (*tablejump<mode>_internal1_nospec): New define_insn.
3748 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
3751 2018-01-16 Artyom Skrobov tyomitch@gmail.com
3753 * caller-save.c (insert_save): Drop unnecessary parameter. All
3756 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3757 Richard Biener <rguenth@suse.de>
3760 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
3761 return early, inline manually is_gimple_sizepos. Make sure if we
3762 call gimplify_expr we don't end up with a gimple constant.
3763 * tree.c (variably_modified_type_p): Don't return true for
3764 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
3765 * gimplify.h (is_gimple_sizepos): Remove.
3767 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3769 PR tree-optimization/83857
3770 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
3771 vectorizable_live_operation for pure SLP statements.
3772 (vectorizable_live_operation): Handle PHIs.
3774 2018-01-16 Richard Biener <rguenther@suse.de>
3776 PR tree-optimization/83867
3777 * tree-vect-stmts.c (vect_transform_stmt): Precompute
3778 nested_in_vect_loop_p since the scalar stmt may get invalidated.
3780 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3783 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
3784 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
3785 If off is not INTEGER_CST, issue a may not be aligned warning
3786 rather than isn't aligned. Use isn%'t rather than isn't.
3787 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
3789 <case MULT_EXPR>: Improve the case when bottom and one of the
3790 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
3791 operand, in that case check if the other operand is multiple of
3792 bottom divided by the INTEGER_CST operand.
3794 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3797 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
3798 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
3799 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
3800 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
3801 * config/pa/pa.c (pa_function_arg_advance): Likewise.
3802 (pa_function_arg, pa_arg_partial_bytes): Likewise.
3803 (pa_function_arg_size): New function.
3805 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3807 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
3808 in a separate statement.
3810 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3812 PR tree-optimization/83847
3813 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
3814 group gathers and scatters.
3816 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3818 PR rtl-optimization/86620
3819 * params.def (max-sched-ready-insns): Bump minimum value to 1.
3821 PR rtl-optimization/83213
3822 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
3823 to last if both are JUMP_INSNs.
3825 PR tree-optimization/83843
3826 * gimple-ssa-store-merging.c
3827 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
3828 store_immediate_info for bswap/nop orig_stores.
3830 2018-01-15 Andrew Waterman <andrew@sifive.com>
3832 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
3834 <UDIV>: Increase cost if !TARGET_DIV.
3836 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
3838 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
3839 (define_attr "cr_logical_3op"): New.
3840 (cceq_ior_compare): Adjust.
3841 (cceq_ior_compare_complement): Adjust.
3842 (*cceq_rev_compare): Adjust.
3843 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
3844 (is_cracked_insn): Adjust.
3845 (insn_must_be_first_in_group): Adjust.
3846 * config/rs6000/40x.md: Adjust.
3847 * config/rs6000/440.md: Adjust.
3848 * config/rs6000/476.md: Adjust.
3849 * config/rs6000/601.md: Adjust.
3850 * config/rs6000/603.md: Adjust.
3851 * config/rs6000/6xx.md: Adjust.
3852 * config/rs6000/7450.md: Adjust.
3853 * config/rs6000/7xx.md: Adjust.
3854 * config/rs6000/8540.md: Adjust.
3855 * config/rs6000/cell.md: Adjust.
3856 * config/rs6000/e300c2c3.md: Adjust.
3857 * config/rs6000/e500mc.md: Adjust.
3858 * config/rs6000/e500mc64.md: Adjust.
3859 * config/rs6000/e5500.md: Adjust.
3860 * config/rs6000/e6500.md: Adjust.
3861 * config/rs6000/mpc.md: Adjust.
3862 * config/rs6000/power4.md: Adjust.
3863 * config/rs6000/power5.md: Adjust.
3864 * config/rs6000/power6.md: Adjust.
3865 * config/rs6000/power7.md: Adjust.
3866 * config/rs6000/power8.md: Adjust.
3867 * config/rs6000/power9.md: Adjust.
3868 * config/rs6000/rs64.md: Adjust.
3869 * config/rs6000/titan.md: Adjust.
3871 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3873 * config/i386/predicates.md (indirect_branch_operand): Rewrite
3874 ix86_indirect_branch_register logic.
3876 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3878 * config/i386/constraints.md (Bs): Update
3879 ix86_indirect_branch_register check. Don't check
3880 ix86_indirect_branch_register with GOT_memory_operand.
3882 * config/i386/predicates.md (GOT_memory_operand): Don't check
3883 ix86_indirect_branch_register here.
3884 (GOT32_symbol_operand): Likewise.
3886 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3888 * config/i386/predicates.md (constant_call_address_operand):
3889 Rewrite ix86_indirect_branch_register logic.
3890 (sibcall_insn_operand): Likewise.
3892 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3894 * config/i386/constraints.md (Bs): Replace
3895 ix86_indirect_branch_thunk_register with
3896 ix86_indirect_branch_register.
3898 * config/i386/i386.md (indirect_jump): Likewise.
3899 (tablejump): Likewise.
3900 (*sibcall_memory): Likewise.
3901 (*sibcall_value_memory): Likewise.
3902 Peepholes of indirect call and jump via memory: Likewise.
3903 * config/i386/i386.opt: Likewise.
3904 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3905 (GOT_memory_operand): Likewise.
3906 (call_insn_operand): Likewise.
3907 (sibcall_insn_operand): Likewise.
3908 (GOT32_symbol_operand): Likewise.
3910 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3913 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
3914 type rather than type addr's type points to.
3915 (expand_omp_atomic_mutex): Likewise.
3916 (expand_omp_atomic): Likewise.
3918 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3921 * config/i386/i386.c (output_indirect_thunk_function): Use
3922 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
3923 for __x86_return_thunk.
3925 2018-01-15 Richard Biener <rguenther@suse.de>
3928 * expmed.c (extract_bit_field_1): Fix typo.
3930 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3933 * config/arm/iterators.md (VF): New mode iterator.
3934 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
3935 Remove integer-related logic from pattern.
3936 (neon_vabd<mode>_3): Likewise.
3938 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3941 * common.opt (fstrict-overflow): No longer an alias.
3942 (fwrapv-pointer): New option.
3943 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
3944 also for pointer types based on flag_wrapv_pointer.
3945 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
3946 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
3947 opts->x_flag_wrapv got set.
3948 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
3949 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
3950 POINTER_TYPE_OVERFLOW_UNDEFINED.
3951 * match.pd: Likewise in address comparison pattern.
3952 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
3954 2018-01-15 Richard Biener <rguenther@suse.de>
3957 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
3958 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
3959 Reset type names to their identifier if their TYPE_DECL doesn't
3960 have linkage (and thus is used for ODR and devirt).
3961 (save_debug_info_for_decl): Remove.
3962 (save_debug_info_for_type): Likewise.
3963 (add_tree_to_fld_list): Adjust.
3964 * tree-pretty-print.c (dump_generic_node): Make dumping of
3965 type names more robust.
3967 2018-01-15 Richard Biener <rguenther@suse.de>
3969 * BASE-VER: Bump to 8.0.1.
3971 2018-01-14 Martin Sebor <msebor@redhat.com>
3974 * builtins.c (check_access): Avoid warning when the no-warning bit
3977 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
3979 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
3980 * ira-color (allocno_hard_regs_compare): Likewise.
3982 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
3985 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
3986 Use .pushsection/.popsection.
3988 2018-01-14 Martin Sebor <msebor@redhat.com>
3991 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
3993 2018-01-14 Jakub Jelinek <jakub@redhat.com>
3995 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
3996 entry from extra_headers.
3997 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
3998 extra_headers, make the list bitwise identical to the i?86-*-* one.
4000 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4002 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
4003 -mcmodel=large with -mindirect-branch=thunk,
4004 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
4005 -mfunction-return=thunk-extern.
4006 * doc/invoke.texi: Document -mcmodel=large is incompatible with
4007 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
4008 -mfunction-return=thunk and -mfunction-return=thunk-extern.
4010 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4012 * config/i386/i386.c (print_reg): Print the name of the full
4013 integer register without '%'.
4014 (ix86_print_operand): Handle 'V'.
4015 * doc/extend.texi: Document 'V' modifier.
4017 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4019 * config/i386/constraints.md (Bs): Disallow memory operand for
4020 -mindirect-branch-register.
4022 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4023 (GOT_memory_operand): Likewise.
4024 (call_insn_operand): Likewise.
4025 (sibcall_insn_operand): Likewise.
4026 (GOT32_symbol_operand): Likewise.
4027 * config/i386/i386.md (indirect_jump): Call convert_memory_address
4028 for -mindirect-branch-register.
4029 (tablejump): Likewise.
4030 (*sibcall_memory): Likewise.
4031 (*sibcall_value_memory): Likewise.
4032 Disallow peepholes of indirect call and jump via memory for
4033 -mindirect-branch-register.
4034 (*call_pop): Replace m with Bw.
4035 (*call_value_pop): Likewise.
4036 (*sibcall_pop_memory): Replace m with Bs.
4037 * config/i386/i386.opt (mindirect-branch-register): New option.
4038 * doc/invoke.texi: Document -mindirect-branch-register option.
4040 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4042 * config/i386/i386-protos.h (ix86_output_function_return): New.
4043 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
4044 set function_return_type.
4045 (indirect_thunk_name): Add ret_p to indicate thunk for function
4047 (output_indirect_thunk_function): Pass false to
4048 indirect_thunk_name.
4049 (ix86_output_indirect_branch_via_reg): Likewise.
4050 (ix86_output_indirect_branch_via_push): Likewise.
4051 (output_indirect_thunk_function): Create alias for function
4052 return thunk if regno < 0.
4053 (ix86_output_function_return): New function.
4054 (ix86_handle_fndecl_attribute): Handle function_return.
4055 (ix86_attribute_table): Add function_return.
4056 * config/i386/i386.h (machine_function): Add
4057 function_return_type.
4058 * config/i386/i386.md (simple_return_internal): Use
4059 ix86_output_function_return.
4060 (simple_return_internal_long): Likewise.
4061 * config/i386/i386.opt (mfunction-return=): New option.
4062 (indirect_branch): Mention -mfunction-return=.
4063 * doc/extend.texi: Document function_return function attribute.
4064 * doc/invoke.texi: Document -mfunction-return= option.
4066 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4068 * config/i386/i386-opts.h (indirect_branch): New.
4069 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
4070 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
4071 with local indirect jump when converting indirect call and jump.
4072 (ix86_set_indirect_branch_type): New.
4073 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
4074 (indirectlabelno): New.
4075 (indirect_thunk_needed): Likewise.
4076 (indirect_thunk_bnd_needed): Likewise.
4077 (indirect_thunks_used): Likewise.
4078 (indirect_thunks_bnd_used): Likewise.
4079 (INDIRECT_LABEL): Likewise.
4080 (indirect_thunk_name): Likewise.
4081 (output_indirect_thunk): Likewise.
4082 (output_indirect_thunk_function): Likewise.
4083 (ix86_output_indirect_branch_via_reg): Likewise.
4084 (ix86_output_indirect_branch_via_push): Likewise.
4085 (ix86_output_indirect_branch): Likewise.
4086 (ix86_output_indirect_jmp): Likewise.
4087 (ix86_code_end): Call output_indirect_thunk_function if needed.
4088 (ix86_output_call_insn): Call ix86_output_indirect_branch if
4090 (ix86_handle_fndecl_attribute): Handle indirect_branch.
4091 (ix86_attribute_table): Add indirect_branch.
4092 * config/i386/i386.h (machine_function): Add indirect_branch_type
4093 and has_local_indirect_jump.
4094 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
4096 (tablejump): Likewise.
4097 (*indirect_jump): Use ix86_output_indirect_jmp.
4098 (*tablejump_1): Likewise.
4099 (simple_return_indirect_internal): Likewise.
4100 * config/i386/i386.opt (mindirect-branch=): New option.
4101 (indirect_branch): New.
4104 (thunk-inline): Likewise.
4105 (thunk-extern): Likewise.
4106 * doc/extend.texi: Document indirect_branch function attribute.
4107 * doc/invoke.texi: Document -mindirect-branch= option.
4109 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
4112 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
4114 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
4116 * ipa-inline.c (want_inline_small_function_p): Return false if
4117 inlining has already failed with CIF_FINAL_ERROR.
4118 (update_caller_keys): Call want_inline_small_function_p before
4120 (update_callee_keys): Likewise.
4122 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4124 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
4126 (rs6000_quadword_masked_address_p): Likewise.
4127 (quad_aligned_load_p): Likewise.
4128 (quad_aligned_store_p): Likewise.
4129 (const_load_sequence_p): Add comment to describe the outer-most loop.
4130 (mimic_memory_attributes_and_flags): New function.
4131 (rs6000_gen_stvx): Likewise.
4132 (replace_swapped_aligned_store): Likewise.
4133 (rs6000_gen_lvx): Likewise.
4134 (replace_swapped_aligned_load): Likewise.
4135 (replace_swapped_load_constant): Capitalize argument name in
4136 comment describing this function.
4137 (rs6000_analyze_swaps): Add a third pass to search for vector loads
4138 and stores that access quad-word aligned addresses and replace
4139 with stvx or lvx instructions when appropriate.
4140 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
4141 New function prototype.
4142 (rs6000_quadword_masked_address_p): Likewise.
4143 (rs6000_gen_lvx): Likewise.
4144 (rs6000_gen_stvx): Likewise.
4145 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
4146 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
4147 when memory address is aligned.
4148 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
4149 this split to select lvx instruction when memory address is aligned.
4150 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
4151 instruction when memory address is aligned.
4152 (*vsx_le_perm_load_v16qi): Likewise.
4153 (four unnamed splitters): Modify to select the stvx instruction
4154 when memory is aligned.
4156 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
4158 * predict.c (determine_unlikely_bbs): Handle correctly BBs
4159 which appears in the queue multiple times.
4161 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4162 Alan Hayward <alan.hayward@arm.com>
4163 David Sherwood <david.sherwood@arm.com>
4165 * tree-vectorizer.h (vec_lower_bound): New structure.
4166 (_loop_vec_info): Add check_nonzero and lower_bounds.
4167 (LOOP_VINFO_CHECK_NONZERO): New macro.
4168 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
4169 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
4170 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
4171 fields. Make seg_len the distance travelled, not including the
4173 (dr_direction_indicator): Declare.
4174 (dr_zero_step_indicator): Likewise.
4175 (dr_known_forward_stride_p): Likewise.
4176 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
4178 (runtime_alias_check_p): Allow runtime alias checks with
4180 (operator ==): Compare access_size and align.
4181 (prune_runtime_alias_test_list): Rework for new distinction between
4182 the access_size and seg_len.
4183 (create_intersect_range_checks_index): Likewise. Cope with polynomial
4185 (get_segment_min_max): New function.
4186 (create_intersect_range_checks): Use it.
4187 (dr_step_indicator): New function.
4188 (dr_direction_indicator): Likewise.
4189 (dr_zero_step_indicator): Likewise.
4190 (dr_known_forward_stride_p): Likewise.
4191 * tree-loop-distribution.c (data_ref_segment_size): Return
4192 DR_STEP * (niters - 1).
4193 (compute_alias_check_pairs): Update call to the dr_with_seg_len
4195 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
4196 (vect_preserves_scalar_order_p): New function, split out from...
4197 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
4198 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
4199 (vect_vfa_access_size): New function.
4200 (vect_vfa_align): Likewise.
4201 (vect_compile_time_alias): Take access_size_a and access_b arguments.
4202 (dump_lower_bound): New function.
4203 (vect_check_lower_bound): Likewise.
4204 (vect_small_gap_p): Likewise.
4205 (vectorizable_with_step_bound_p): Likewise.
4206 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
4207 depencies if the vectorization factor is 1. Convert the checks
4208 for nonzero steps into checks on the bounds of DR_STEP. Try using
4209 a bunds check for variable steps if the minimum required step is
4210 relatively small. Update calls to the dr_with_seg_len
4211 constructor and to vect_compile_time_alias.
4212 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
4214 (vect_loop_versioning): Call it.
4215 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
4217 (vect_estimate_min_profitable_iters): Account for any bounds checks.
4219 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4220 Alan Hayward <alan.hayward@arm.com>
4221 David Sherwood <david.sherwood@arm.com>
4223 * doc/sourcebuild.texi (vect_scatter_store): Document.
4224 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
4226 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
4228 * genopinit.c (main): Add supports_vec_scatter_store and
4229 supports_vec_scatter_store_cached to target_optabs.
4230 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
4231 IFN_MASK_SCATTER_STORE.
4232 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
4234 * internal-fn.h (internal_store_fn_p): Declare.
4235 (internal_fn_stored_value_index): Likewise.
4236 * internal-fn.c (scatter_store_direct): New macro.
4237 (expand_scatter_store_optab_fn): New function.
4238 (direct_scatter_store_optab_supported_p): New macro.
4239 (internal_store_fn_p): New function.
4240 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
4241 IFN_MASK_SCATTER_STORE.
4242 (internal_fn_mask_index): Likewise.
4243 (internal_fn_stored_value_index): New function.
4244 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
4246 * optabs-query.h (supports_vec_scatter_store_p): Declare.
4247 * optabs-query.c (supports_vec_scatter_store_p): New function.
4248 * tree-vectorizer.h (vect_get_store_rhs): Declare.
4249 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
4250 true for scatter stores.
4251 (vect_gather_scatter_fn_p): Handle scatter stores too.
4252 (vect_check_gather_scatter): Consider using scatter stores if
4253 supports_vec_scatter_store_p.
4254 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
4256 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4257 internal_fn_stored_value_index.
4258 (check_load_store_masking): Handle scatter stores too.
4259 (vect_get_store_rhs): Make public.
4260 (vectorizable_call): Use internal_store_fn_p.
4261 (vectorizable_store): Handle scatter store internal functions.
4262 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
4263 when deciding whether the end of the group has been reached.
4264 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
4265 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
4266 (mask_scatter_store<mode>): New insns.
4268 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4269 Alan Hayward <alan.hayward@arm.com>
4270 David Sherwood <david.sherwood@arm.com>
4272 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
4273 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
4274 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
4276 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
4277 Use vect_truncate_gather_scatter_offset if we can't treat the
4278 operation as a normal gather load or scatter store.
4279 (get_group_load_store_type): Take the gather_scatter_info
4280 as argument. Try using a gather load or scatter store for
4281 single-element groups.
4282 (get_load_store_type): Update calls to get_group_load_store_type
4283 and vect_use_strided_gather_scatters_p.
4285 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4286 Alan Hayward <alan.hayward@arm.com>
4287 David Sherwood <david.sherwood@arm.com>
4289 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
4290 optional tree argument.
4291 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
4293 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
4294 but continue to use the current value as a fallback.
4295 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
4296 to compare the updates.
4297 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
4298 (get_load_store_type): Use it when handling a strided access.
4299 (vect_get_strided_load_store_ops): New function.
4300 (vect_get_data_ptr_increment): Likewise.
4301 (vectorizable_load): Handle strided gather loads. Always pass
4302 a step to vect_create_data_ref_ptr and bump_vector_ptr.
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 * doc/md.texi (gather_load@var{m}): Document.
4309 (mask_gather_load@var{m}): Likewise.
4310 * genopinit.c (main): Add supports_vec_gather_load and
4311 supports_vec_gather_load_cached to target_optabs.
4312 * optabs-tree.c (init_tree_optimization_optabs): Use
4313 ggc_cleared_alloc to allocate target_optabs.
4314 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
4315 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
4317 * internal-fn.h (internal_load_fn_p): Declare.
4318 (internal_gather_scatter_fn_p): Likewise.
4319 (internal_fn_mask_index): Likewise.
4320 (internal_gather_scatter_fn_supported_p): Likewise.
4321 * internal-fn.c (gather_load_direct): New macro.
4322 (expand_gather_load_optab_fn): New function.
4323 (direct_gather_load_optab_supported_p): New macro.
4324 (direct_internal_fn_optab): New function.
4325 (internal_load_fn_p): Likewise.
4326 (internal_gather_scatter_fn_p): Likewise.
4327 (internal_fn_mask_index): Likewise.
4328 (internal_gather_scatter_fn_supported_p): Likewise.
4329 * optabs-query.c (supports_at_least_one_mode_p): New function.
4330 (supports_vec_gather_load_p): Likewise.
4331 * optabs-query.h (supports_vec_gather_load_p): Declare.
4332 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
4333 and memory_type field.
4334 (NUM_PATTERNS): Bump to 15.
4335 * tree-vect-data-refs.c: Include internal-fn.h.
4336 (vect_gather_scatter_fn_p): New function.
4337 (vect_describe_gather_scatter_call): Likewise.
4338 (vect_check_gather_scatter): Try using internal functions for
4339 gather loads. Recognize existing calls to a gather load function.
4340 (vect_analyze_data_refs): Consider using gather loads if
4341 supports_vec_gather_load_p.
4342 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
4343 (vect_get_gather_scatter_offset_type): Likewise.
4344 (vect_convert_mask_for_vectype): Likewise.
4345 (vect_add_conversion_to_patterm): Likewise.
4346 (vect_try_gather_scatter_pattern): Likewise.
4347 (vect_recog_gather_scatter_pattern): New pattern recognizer.
4348 (vect_vect_recog_func_ptrs): Add it.
4349 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4350 internal_fn_mask_index and internal_gather_scatter_fn_p.
4351 (check_load_store_masking): Take the gather_scatter_info as an
4352 argument and handle gather loads.
4353 (vect_get_gather_scatter_ops): New function.
4354 (vectorizable_call): Check internal_load_fn_p.
4355 (vectorizable_load): Likewise. Handle gather load internal
4357 (vectorizable_store): Update call to check_load_store_masking.
4358 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
4359 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
4360 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
4361 (aarch64_gather_scale_operand_d): New predicates.
4362 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
4363 (mask_gather_load<mode>): New insns.
4365 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4366 Alan Hayward <alan.hayward@arm.com>
4367 David Sherwood <david.sherwood@arm.com>
4369 * optabs.def (fold_left_plus_optab): New optab.
4370 * doc/md.texi (fold_left_plus_@var{m}): Document.
4371 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
4372 * internal-fn.c (fold_left_direct): Define.
4373 (expand_fold_left_optab_fn): Likewise.
4374 (direct_fold_left_optab_supported_p): Likewise.
4375 * fold-const-call.c (fold_const_fold_left): New function.
4376 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
4377 * tree-parloops.c (valid_reduction_p): New function.
4378 (gather_scalar_reductions): Use it.
4379 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
4380 (vect_finish_replace_stmt): Declare.
4381 * tree-vect-loop.c (fold_left_reduction_fn): New function.
4382 (needs_fold_left_reduction_p): New function, split out from...
4383 (vect_is_simple_reduction): ...here. Accept reductions that
4384 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
4385 (vect_force_simple_reduction): Also store the reduction type in
4386 the assignment's STMT_VINFO_REDUC_TYPE.
4387 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
4388 (merge_with_identity): New function.
4389 (vect_expand_fold_left): Likewise.
4390 (vectorize_fold_left_reduction): Likewise.
4391 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
4392 scalar phi in place for it. Check for target support and reject
4393 cases that would reassociate the operation. Defer the transform
4394 phase to vectorize_fold_left_reduction.
4395 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
4396 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
4397 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
4399 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4401 * tree-if-conv.c (predicate_mem_writes): Remove redundant
4402 call to ifc_temp_var.
4404 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4405 Alan Hayward <alan.hayward@arm.com>
4406 David Sherwood <david.sherwood@arm.com>
4408 * target.def (legitimize_address_displacement): Take the original
4409 offset as a poly_int.
4410 * targhooks.h (default_legitimize_address_displacement): Update
4412 * targhooks.c (default_legitimize_address_displacement): Likewise.
4413 * doc/tm.texi: Regenerate.
4414 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
4415 as an argument, moving assert of ad->disp == ad->disp_term to...
4416 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
4417 Try calling targetm.legitimize_address_displacement before expanding
4418 the address rather than afterwards, and adjust for the new interface.
4419 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
4420 Match the new hook interface. Handle SVE addresses.
4421 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
4424 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4426 * Makefile.in (OBJS): Add early-remat.o.
4427 * target.def (select_early_remat_modes): New hook.
4428 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
4429 * doc/tm.texi: Regenerate.
4430 * targhooks.h (default_select_early_remat_modes): Declare.
4431 * targhooks.c (default_select_early_remat_modes): New function.
4432 * timevar.def (TV_EARLY_REMAT): New timevar.
4433 * passes.def (pass_early_remat): New pass.
4434 * tree-pass.h (make_pass_early_remat): Declare.
4435 * early-remat.c: New file.
4436 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
4438 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
4440 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4441 Alan Hayward <alan.hayward@arm.com>
4442 David Sherwood <david.sherwood@arm.com>
4444 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
4445 vfm1 with a bound_epilog parameter.
4446 (vect_do_peeling): Update calls accordingly, and move the prologue
4447 call earlier in the function. Treat the base bound_epilog as 0 for
4448 fully-masked loops and retain vf - 1 for other loops. Add 1 to
4449 this base when peeling for gaps.
4450 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
4451 with fully-masked loops.
4452 (vect_estimate_min_profitable_iters): Handle the single peeled
4453 iteration in that case.
4455 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4456 Alan Hayward <alan.hayward@arm.com>
4457 David Sherwood <david.sherwood@arm.com>
4459 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
4460 single-element interleaving even if the size is not a power of 2.
4461 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
4462 accesses for single-element interleaving if the group size is
4465 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4466 Alan Hayward <alan.hayward@arm.com>
4467 David Sherwood <david.sherwood@arm.com>
4469 * doc/md.texi (fold_extract_last_@var{m}): Document.
4470 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
4471 * optabs.def (fold_extract_last_optab): New optab.
4472 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
4473 * internal-fn.c (fold_extract_direct): New macro.
4474 (expand_fold_extract_optab_fn): Likewise.
4475 (direct_fold_extract_optab_supported_p): Likewise.
4476 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
4477 * tree-vect-loop.c (vect_model_reduction_cost): Handle
4478 EXTRACT_LAST_REDUCTION.
4479 (get_initial_def_for_reduction): Do not create an initial vector
4480 for EXTRACT_LAST_REDUCTION reductions.
4481 (vectorizable_reduction): Leave the scalar phi in place for
4482 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
4483 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
4484 epilogue code for EXTRACT_LAST_REDUCTION and defer the
4485 transform phase to vectorizable_condition.
4486 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
4488 (vect_finish_stmt_generation): ...here.
4489 (vect_finish_replace_stmt): New function.
4490 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
4491 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
4493 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
4495 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4496 Alan Hayward <alan.hayward@arm.com>
4497 David Sherwood <david.sherwood@arm.com>
4499 * doc/md.texi (extract_last_@var{m}): Document.
4500 * optabs.def (extract_last_optab): New optab.
4501 * internal-fn.def (EXTRACT_LAST): New internal function.
4502 * internal-fn.c (cond_unary_direct): New macro.
4503 (expand_cond_unary_optab_fn): Likewise.
4504 (direct_cond_unary_optab_supported_p): Likewise.
4505 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
4506 loops using EXTRACT_LAST.
4507 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
4508 (extract_last_<mode>): ...this optab.
4509 (vec_extract<mode><Vel>): Update accordingly.
4511 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4512 Alan Hayward <alan.hayward@arm.com>
4513 David Sherwood <david.sherwood@arm.com>
4515 * target.def (empty_mask_is_expensive): New hook.
4516 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
4517 * doc/tm.texi: Regenerate.
4518 * targhooks.h (default_empty_mask_is_expensive): Declare.
4519 * targhooks.c (default_empty_mask_is_expensive): New function.
4520 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
4521 if the target says that empty masks are expensive.
4522 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
4524 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
4526 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4527 Alan Hayward <alan.hayward@arm.com>
4528 David Sherwood <david.sherwood@arm.com>
4530 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
4531 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
4532 (vect_use_loop_mask_for_alignment_p): New function.
4533 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
4534 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
4535 niters_skip argument. Make sure that the first niters_skip elements
4536 of the first iteration are inactive.
4537 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
4538 Update call to vect_set_loop_masks_directly.
4539 (get_misalign_in_elems): New function, split out from...
4540 (vect_gen_prolog_loop_niters): ...here.
4541 (vect_update_init_of_dr): Take a code argument that specifies whether
4542 the adjustment should be added or subtracted.
4543 (vect_update_init_of_drs): Likewise.
4544 (vect_prepare_for_masked_peels): New function.
4545 (vect_do_peeling): Skip prologue peeling if we're using a mask
4546 instead. Update call to vect_update_inits_of_drs.
4547 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4549 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
4550 alignment. Do not include the number of peeled iterations in
4551 the minimum threshold in that case.
4552 (vectorizable_induction): Adjust the start value down by
4553 LOOP_VINFO_MASK_SKIP_NITERS iterations.
4554 (vect_transform_loop): Call vect_prepare_for_masked_peels.
4555 Take the number of skipped iterations into account when calculating
4557 * tree-vect-stmts.c (vect_gen_while_not): New function.
4559 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4560 Alan Hayward <alan.hayward@arm.com>
4561 David Sherwood <david.sherwood@arm.com>
4563 * doc/sourcebuild.texi (vect_fully_masked): Document.
4564 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
4566 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
4568 (vect_analyze_loop_2): ...here. Don't check the vectorization
4569 factor against the number of loop iterations if the loop is
4572 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4573 Alan Hayward <alan.hayward@arm.com>
4574 David Sherwood <david.sherwood@arm.com>
4576 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
4577 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
4578 (dump_groups): Update accordingly.
4579 (iv_use::mem_type): New member variable.
4580 (address_p): New function.
4581 (record_use): Add a mem_type argument and initialize the new
4583 (record_group_use): Add a mem_type argument. Use address_p.
4584 Remove obsolete null checks of base_object. Update call to record_use.
4585 (find_interesting_uses_op): Update call to record_group_use.
4586 (find_interesting_uses_cond): Likewise.
4587 (find_interesting_uses_address): Likewise.
4588 (get_mem_type_for_internal_fn): New function.
4589 (find_address_like_use): Likewise.
4590 (find_interesting_uses_stmt): Try find_address_like_use before
4591 calling find_interesting_uses_op.
4592 (addr_offset_valid_p): Use the iv mem_type field as the type
4593 of the addressed memory.
4594 (add_autoinc_candidates): Likewise.
4595 (get_address_cost): Likewise.
4596 (split_small_address_groups_p): Use address_p.
4597 (split_address_groups): Likewise.
4598 (add_iv_candidate_for_use): Likewise.
4599 (autoinc_possible_for_pair): Likewise.
4600 (rewrite_groups): Likewise.
4601 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
4602 (determine_group_iv_cost): Update after split of USE_ADDRESS.
4603 (get_alias_ptr_type_for_ptr_address): New function.
4604 (rewrite_use_address): Rewrite address uses in calls that were
4605 identified by find_address_like_use.
4607 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4608 Alan Hayward <alan.hayward@arm.com>
4609 David Sherwood <david.sherwood@arm.com>
4611 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
4613 * gimple-expr.h (is_gimple_addressable: Likewise.
4614 * gimple-expr.c (is_gimple_address): Likewise.
4615 * internal-fn.c (expand_call_mem_ref): New function.
4616 (expand_mask_load_optab_fn): Use it.
4617 (expand_mask_store_optab_fn): Likewise.
4619 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4620 Alan Hayward <alan.hayward@arm.com>
4621 David Sherwood <david.sherwood@arm.com>
4623 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
4624 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
4625 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
4626 (cond_umax@var{mode}): Document.
4627 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
4628 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
4629 (cond_umin_optab, cond_umax_optab): New optabs.
4630 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
4631 (COND_IOR, COND_XOR): New internal functions.
4632 * internal-fn.h (get_conditional_internal_fn): Declare.
4633 * internal-fn.c (cond_binary_direct): New macro.
4634 (expand_cond_binary_optab_fn): Likewise.
4635 (direct_cond_binary_optab_supported_p): Likewise.
4636 (get_conditional_internal_fn): New function.
4637 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
4638 Cope with reduction statements that are vectorized as calls rather
4640 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
4641 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
4642 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
4643 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4644 (UNSPEC_COND_EOR): New unspecs.
4645 (optab): Add mappings for them.
4646 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
4647 (sve_int_op, sve_fp_op): New int attributes.
4649 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4650 Alan Hayward <alan.hayward@arm.com>
4651 David Sherwood <david.sherwood@arm.com>
4653 * optabs.def (while_ult_optab): New optab.
4654 * doc/md.texi (while_ult@var{m}@var{n}): Document.
4655 * internal-fn.def (WHILE_ULT): New internal function.
4656 * internal-fn.h (direct_internal_fn_supported_p): New override
4657 that takes two types as argument.
4658 * internal-fn.c (while_direct): New macro.
4659 (expand_while_optab_fn): New function.
4660 (convert_optab_supported_p): Likewise.
4661 (direct_while_optab_supported_p): New macro.
4662 * wide-int.h (wi::udiv_ceil): New function.
4663 * tree-vectorizer.h (rgroup_masks): New structure.
4664 (vec_loop_masks): New typedef.
4665 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
4667 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
4668 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
4669 (vect_max_vf): New function.
4670 (slpeel_make_loop_iterate_ntimes): Delete.
4671 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
4672 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
4673 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
4674 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
4675 internal-fn.h, stor-layout.h and optabs-query.h.
4676 (vect_set_loop_mask): New function.
4677 (add_preheader_seq): Likewise.
4678 (add_header_seq): Likewise.
4679 (interleave_supported_p): Likewise.
4680 (vect_maybe_permute_loop_masks): Likewise.
4681 (vect_set_loop_masks_directly): Likewise.
4682 (vect_set_loop_condition_masked): Likewise.
4683 (vect_set_loop_condition_unmasked): New function, split out from
4684 slpeel_make_loop_iterate_ntimes.
4685 (slpeel_make_loop_iterate_ntimes): Rename to..
4686 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
4687 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
4688 (vect_do_peeling): Update call accordingly.
4689 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
4691 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4692 mask_compare_type, can_fully_mask_p and fully_masked_p.
4693 (release_vec_loop_masks): New function.
4694 (_loop_vec_info): Use it to free the loop masks.
4695 (can_produce_all_loop_masks_p): New function.
4696 (vect_get_max_nscalars_per_iter): Likewise.
4697 (vect_verify_full_masking): Likewise.
4698 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
4699 retries, and free the mask rgroups before retrying. Check loop-wide
4700 reasons for disallowing fully-masked loops. Make the final decision
4701 about whether use a fully-masked loop or not.
4702 (vect_estimate_min_profitable_iters): Do not assume that peeling
4703 for the number of iterations will be needed for fully-masked loops.
4704 (vectorizable_reduction): Disable fully-masked loops.
4705 (vectorizable_live_operation): Likewise.
4706 (vect_halve_mask_nunits): New function.
4707 (vect_double_mask_nunits): Likewise.
4708 (vect_record_loop_mask): Likewise.
4709 (vect_get_loop_mask): Likewise.
4710 (vect_transform_loop): Handle the case in which the final loop
4711 iteration might handle a partial vector. Call vect_set_loop_condition
4712 instead of slpeel_make_loop_iterate_ntimes.
4713 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
4714 (check_load_store_masking): New function.
4715 (prepare_load_store_mask): Likewise.
4716 (vectorizable_store): Handle fully-masked loops.
4717 (vectorizable_load): Likewise.
4718 (supportable_widening_operation): Use vect_halve_mask_nunits for
4720 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
4721 (vect_gen_while): New function.
4722 * config/aarch64/aarch64.md (umax<mode>3): New expander.
4723 (aarch64_uqdec<mode>): New insn.
4725 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4726 Alan Hayward <alan.hayward@arm.com>
4727 David Sherwood <david.sherwood@arm.com>
4729 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
4730 (reduc_xor_scal_optab): New optabs.
4731 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
4732 (reduc_xor_scal_@var{m}): Document.
4733 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
4734 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
4736 * fold-const-call.c (fold_const_call): Handle them.
4737 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
4738 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
4739 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
4740 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
4741 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
4742 (UNSPEC_XORV): New unspecs.
4743 (optab): Add entries for them.
4744 (BITWISEV): New int iterator.
4745 (bit_reduc_op): New int attributes.
4747 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4748 Alan Hayward <alan.hayward@arm.com>
4749 David Sherwood <david.sherwood@arm.com>
4751 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
4752 * internal-fn.def (VEC_SHL_INSERT): New internal function.
4753 * optabs.def (vec_shl_insert_optab): New optab.
4754 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
4755 (duplicate_and_interleave): Likewise.
4756 * tree-vect-loop.c: Include internal-fn.h.
4757 (neutral_op_for_slp_reduction): New function, split out from
4758 get_initial_defs_for_reduction.
4759 (get_initial_def_for_reduction): Handle option 2 for variable-length
4760 vectors by loading the neutral value into a vector and then shifting
4761 the initial value into element 0.
4762 (get_initial_defs_for_reduction): Replace the code argument with
4763 the neutral value calculated by neutral_op_for_slp_reduction.
4764 Use gimple_build_vector for constant-length vectors.
4765 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
4766 but the first group_size elements have a neutral value.
4767 Use duplicate_and_interleave otherwise.
4768 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
4769 Update call to get_initial_defs_for_reduction. Handle SLP
4770 reductions for variable-length vectors by creating one vector
4771 result for each scalar result, with the elements associated
4772 with other scalar results stubbed out with the neutral value.
4773 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
4774 Require IFN_VEC_SHL_INSERT for double reductions on
4775 variable-length vectors, or SLP reductions that have
4776 a neutral value. Require can_duplicate_and_interleave_p
4777 support for variable-length unchained SLP reductions if there
4778 is no neutral value, such as for MIN/MAX reductions. Also require
4779 the number of vector elements to be a multiple of the number of
4780 SLP statements when doing variable-length unchained SLP reductions.
4781 Update call to vect_create_epilog_for_reduction.
4782 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
4783 and remove initial values.
4784 (duplicate_and_interleave): Make public.
4785 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
4786 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
4788 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4789 Alan Hayward <alan.hayward@arm.com>
4790 David Sherwood <david.sherwood@arm.com>
4792 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
4793 (can_duplicate_and_interleave_p): New function.
4794 (vect_get_and_check_slp_defs): Take the vector of statements
4795 rather than just the current one. Remove excess parentheses.
4796 Restriction rejectinon of vect_constant_def and vect_external_def
4797 for variable-length vectors to boolean types, or types for which
4798 can_duplicate_and_interleave_p is false.
4799 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
4800 (duplicate_and_interleave): New function.
4801 (vect_get_constant_vectors): Use gimple_build_vector for
4802 constant-length vectors and suitable variable-length constant
4803 vectors. Use duplicate_and_interleave for other variable-length
4804 vectors. Don't defer the update when inserting new statements.
4806 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4807 Alan Hayward <alan.hayward@arm.com>
4808 David Sherwood <david.sherwood@arm.com>
4810 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
4811 min_profitable_iters doesn't go negative.
4813 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4814 Alan Hayward <alan.hayward@arm.com>
4815 David Sherwood <david.sherwood@arm.com>
4817 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
4818 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
4819 * optabs.def (vec_mask_load_lanes_optab): New optab.
4820 (vec_mask_store_lanes_optab): Likewise.
4821 * internal-fn.def (MASK_LOAD_LANES): New internal function.
4822 (MASK_STORE_LANES): Likewise.
4823 * internal-fn.c (mask_load_lanes_direct): New macro.
4824 (mask_store_lanes_direct): Likewise.
4825 (expand_mask_load_optab_fn): Handle masked operations.
4826 (expand_mask_load_lanes_optab_fn): New macro.
4827 (expand_mask_store_optab_fn): Handle masked operations.
4828 (expand_mask_store_lanes_optab_fn): New macro.
4829 (direct_mask_load_lanes_optab_supported_p): Likewise.
4830 (direct_mask_store_lanes_optab_supported_p): Likewise.
4831 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
4833 (vect_load_lanes_supported): Likewise.
4834 * tree-vect-data-refs.c (strip_conversion): New function.
4835 (can_group_stmts_p): Likewise.
4836 (vect_analyze_data_ref_accesses): Use it instead of checking
4837 for a pair of assignments.
4838 (vect_store_lanes_supported): Take a masked_p parameter.
4839 (vect_load_lanes_supported): Likewise.
4840 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
4841 vect_store_lanes_supported and vect_load_lanes_supported.
4842 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4843 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
4844 parameter. Don't allow gaps for masked accesses.
4845 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
4846 and vect_load_lanes_supported.
4847 (get_load_store_type): Take a masked_p parameter and update
4848 call to get_group_load_store_type.
4849 (vectorizable_store): Update call to get_load_store_type.
4850 Handle IFN_MASK_STORE_LANES.
4851 (vectorizable_load): Update call to get_load_store_type.
4852 Handle IFN_MASK_LOAD_LANES.
4854 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4855 Alan Hayward <alan.hayward@arm.com>
4856 David Sherwood <david.sherwood@arm.com>
4858 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
4860 * config/aarch64/aarch64-protos.h
4861 (aarch64_sve_struct_memory_operand_p): Declare.
4862 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
4863 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
4864 (VPRED, vpred): Handle SVE structure modes.
4865 * config/aarch64/constraints.md (Utx): New constraint.
4866 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
4867 (aarch64_sve_struct_nonimmediate_operand): New predicates.
4868 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
4869 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
4870 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
4871 structure modes. Split into pieces after RA.
4872 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
4873 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
4875 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
4876 SVE structure modes.
4877 (aarch64_classify_address): Likewise.
4878 (sizetochar): Move earlier in file.
4879 (aarch64_print_operand): Handle SVE register lists.
4880 (aarch64_array_mode): New function.
4881 (aarch64_sve_struct_memory_operand_p): Likewise.
4882 (TARGET_ARRAY_MODE): Redefine.
4884 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4885 Alan Hayward <alan.hayward@arm.com>
4886 David Sherwood <david.sherwood@arm.com>
4888 * target.def (array_mode): New target hook.
4889 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
4890 * doc/tm.texi: Regenerate.
4891 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
4892 * hooks.c (hook_optmode_mode_uhwi_none): New function.
4893 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
4895 * stor-layout.c (mode_for_array): Likewise. Support polynomial
4898 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4899 Alan Hayward <alan.hayward@arm.com>
4900 David Sherwood <david.sherwood@arm.com>
4902 * fold-const.c (fold_binary_loc): Check the argument types
4903 rather than the result type when testing for a vector operation.
4905 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4907 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
4908 * doc/tm.texi: Regenerate.
4910 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4911 Alan Hayward <alan.hayward@arm.com>
4912 David Sherwood <david.sherwood@arm.com>
4914 * doc/invoke.texi (-msve-vector-bits=): Document new option.
4915 (sve): Document new AArch64 extension.
4916 * doc/md.texi (w): Extend the description of the AArch64
4917 constraint to include SVE vectors.
4918 (Upl, Upa): Document new AArch64 predicate constraints.
4919 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
4921 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
4922 (msve-vector-bits=): New option.
4923 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
4924 SVE when these are disabled.
4925 (sve): New extension.
4926 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
4927 modes. Adjust their number of units based on aarch64_sve_vg.
4928 (MAX_BITSIZE_MODE_ANY_MODE): Define.
4929 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
4930 aarch64_addr_query_type.
4931 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
4932 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
4933 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
4934 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
4935 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
4936 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
4937 (aarch64_simd_imm_zero_p): Delete.
4938 (aarch64_check_zero_based_sve_index_immediate): Declare.
4939 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4940 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4941 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4942 (aarch64_sve_float_mul_immediate_p): Likewise.
4943 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4945 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
4946 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
4947 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
4948 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
4949 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
4950 (aarch64_regmode_natural_size): Likewise.
4951 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
4952 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
4954 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
4955 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
4956 for VG and the SVE predicate registers.
4957 (V_ALIASES): Add a "z"-prefixed alias.
4958 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
4959 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
4960 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
4961 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
4962 (REG_CLASS_NAMES): Add entries for them.
4963 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
4964 and the predicate registers.
4965 (aarch64_sve_vg): Declare.
4966 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
4967 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
4968 (REGMODE_NATURAL_SIZE): Define.
4969 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
4971 * config/aarch64/aarch64.c: Include cfgrtl.h.
4972 (simd_immediate_info): Add a constructor for series vectors,
4973 and an associated step field.
4974 (aarch64_sve_vg): New variable.
4975 (aarch64_dbx_register_number): Handle VG and the predicate registers.
4976 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
4977 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
4978 (VEC_ANY_DATA, VEC_STRUCT): New constants.
4979 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
4980 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
4981 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
4982 (aarch64_get_mask_mode): New functions.
4983 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
4984 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
4985 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
4986 predicate modes and predicate registers. Explicitly restrict
4987 GPRs to modes of 16 bytes or smaller. Only allow FP registers
4988 to store a vector mode if it is recognized by
4989 aarch64_classify_vector_mode.
4990 (aarch64_regmode_natural_size): New function.
4991 (aarch64_hard_regno_caller_save_mode): Return the original mode
4993 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
4994 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
4995 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
4996 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
4998 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
4999 does not overlap dest if the function is frame-related. Handle
5001 (aarch64_split_add_offset): New function.
5002 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
5003 them aarch64_add_offset.
5004 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
5005 and update call to aarch64_sub_sp.
5006 (aarch64_add_cfa_expression): New function.
5007 (aarch64_expand_prologue): Pass extra temporary registers to the
5008 functions above. Handle the case in which we need to emit new
5009 DW_CFA_expressions for registers that were originally saved
5010 relative to the stack pointer, but now have to be expressed
5011 relative to the frame pointer.
5012 (aarch64_output_mi_thunk): Pass extra temporary registers to the
5014 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
5015 IP0 and IP1 values for SVE frames.
5016 (aarch64_expand_vec_series): New function.
5017 (aarch64_expand_sve_widened_duplicate): Likewise.
5018 (aarch64_expand_sve_const_vector): Likewise.
5019 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
5020 Handle SVE constants. Use emit_move_insn to move a force_const_mem
5021 into the register, rather than emitting a SET directly.
5022 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
5023 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
5024 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
5025 (offset_9bit_signed_scaled_p): New functions.
5026 (aarch64_replicate_bitmask_imm): New function.
5027 (aarch64_bitmask_imm): Use it.
5028 (aarch64_cannot_force_const_mem): Reject expressions involving
5029 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
5030 (aarch64_classify_index): Handle SVE indices, by requiring
5031 a plain register index with a scale that matches the element size.
5032 (aarch64_classify_address): Handle SVE addresses. Assert that
5033 the mode of the address is VOIDmode or an integer mode.
5034 Update call to aarch64_classify_symbol.
5035 (aarch64_classify_symbolic_expression): Update call to
5036 aarch64_classify_symbol.
5037 (aarch64_const_vec_all_in_range_p): New function.
5038 (aarch64_print_vector_float_operand): Likewise.
5039 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
5040 "vN" for FP registers with SVE modes. Handle (const ...) vectors
5041 and the FP immediates 1.0 and 0.5.
5042 (aarch64_print_address_internal): Handle SVE addresses.
5043 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
5044 (aarch64_regno_regclass): Handle predicate registers.
5045 (aarch64_secondary_reload): Handle big-endian reloads of SVE
5047 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
5048 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
5049 (aarch64_convert_sve_vector_bits): New function.
5050 (aarch64_override_options): Use it to handle -msve-vector-bits=.
5051 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5053 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
5054 Handle SVE vector and predicate modes. Accept VL-based constants
5055 that need only one temporary register, and VL offsets that require
5056 no temporary registers.
5057 (aarch64_conditional_register_usage): Mark the predicate registers
5058 as fixed if SVE isn't available.
5059 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
5060 Return true for SVE vector and predicate modes.
5061 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
5062 rather than an unsigned int. Handle SVE modes.
5063 (aarch64_preferred_simd_mode): Update call accordingly. Handle
5065 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
5067 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5068 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5069 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5070 (aarch64_sve_float_mul_immediate_p): New functions.
5071 (aarch64_sve_valid_immediate): New function.
5072 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
5073 Explicitly reject structure modes. Check for INDEX constants.
5074 Handle PTRUE and PFALSE constants.
5075 (aarch64_check_zero_based_sve_index_immediate): New function.
5076 (aarch64_simd_imm_zero_p): Delete.
5077 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
5078 vector modes. Accept constants in the range of CNT[BHWD].
5079 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
5080 ask for an Advanced SIMD mode.
5081 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
5082 (aarch64_simd_vector_alignment): Handle SVE predicates.
5083 (aarch64_vectorize_preferred_vector_alignment): New function.
5084 (aarch64_simd_vector_alignment_reachable): Use it instead of
5086 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
5087 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
5089 (MAX_VECT_LEN): Delete.
5090 (expand_vec_perm_d): Add a vec_flags field.
5091 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
5092 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
5093 (aarch64_evpc_ext): Don't apply a big-endian lane correction
5095 (aarch64_evpc_rev): Rename to...
5096 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
5097 (aarch64_evpc_rev_global): New function.
5098 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
5099 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
5101 (aarch64_evpc_sve_tbl): New function.
5102 (aarch64_expand_vec_perm_const_1): Update after rename of
5103 aarch64_evpc_rev. Handle SVE permutes too, trying
5104 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
5105 than aarch64_evpc_tbl.
5106 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
5107 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
5108 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
5109 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
5110 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
5111 (aarch64_expand_sve_vcond): New functions.
5112 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
5113 of aarch64_vector_mode_p.
5114 (aarch64_dwarf_poly_indeterminate_value): New function.
5115 (aarch64_compute_pressure_classes): Likewise.
5116 (aarch64_can_change_mode_class): Likewise.
5117 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
5118 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
5119 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
5120 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
5121 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
5122 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
5123 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
5124 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
5126 (Dn, Dl, Dr): Accept const as well as const_vector.
5127 (Dz): Likewise. Compare against CONST0_RTX.
5128 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
5129 of "vector" where appropriate.
5130 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
5131 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
5132 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
5133 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
5134 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
5135 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
5136 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
5137 (v_int_equiv): Extend to SVE modes.
5138 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
5140 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
5141 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
5142 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
5143 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
5144 (SVE_COND_FP_CMP): New int iterators.
5145 (perm_hilo): Handle the new unpack unspecs.
5146 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
5148 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
5149 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
5150 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
5151 (aarch64_equality_operator, aarch64_constant_vector_operand)
5152 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
5153 (aarch64_sve_nonimmediate_operand): Likewise.
5154 (aarch64_sve_general_operand): Likewise.
5155 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
5156 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
5157 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
5158 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
5159 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
5160 (aarch64_sve_float_arith_immediate): Likewise.
5161 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
5162 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
5163 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
5164 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
5165 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
5166 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
5167 (aarch64_sve_float_arith_operand): Likewise.
5168 (aarch64_sve_float_arith_with_sub_operand): Likewise.
5169 (aarch64_sve_float_mul_operand): Likewise.
5170 (aarch64_sve_vec_perm_operand): Likewise.
5171 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
5172 (aarch64_mov_operand): Accept const_poly_int and const_vector.
5173 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
5174 as well as const_vector.
5175 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
5176 in file. Use CONST0_RTX and CONSTM1_RTX.
5177 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
5178 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
5179 Use aarch64_simd_imm_zero.
5180 * config/aarch64/aarch64-sve.md: New file.
5181 * config/aarch64/aarch64.md: Include it.
5182 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
5183 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
5184 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
5185 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
5186 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
5187 (sve): New attribute.
5188 (enabled): Disable instructions with the sve attribute unless
5190 (movqi, movhi): Pass CONST_POLY_INT operaneds through
5191 aarch64_expand_mov_immediate.
5192 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
5193 CNT[BHSD] immediates.
5194 (movti): Split CONST_POLY_INT moves into two halves.
5195 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
5196 Split additions that need a temporary here if the destination
5197 is the stack pointer.
5198 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
5199 (*add<mode>3_poly_1): New instruction.
5200 (set_clobber_cc): New expander.
5202 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5204 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
5205 parameter and use it instead of GET_MODE_SIZE (innermode). Use
5206 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
5207 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
5208 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
5209 Change innermode from fixed_mode_size to machine_mode.
5210 (simplify_subreg): Update call accordingly. Handle a constant-sized
5211 subreg of a variable-length CONST_VECTOR.
5213 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5214 Alan Hayward <alan.hayward@arm.com>
5215 David Sherwood <david.sherwood@arm.com>
5217 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
5218 (add_offset_to_base): New function, split out from...
5219 (create_mem_ref): ...here. When handling a scale other than 1,
5220 check first whether the address is valid without the offset.
5221 Add it into the base if so, leaving the index and scale as-is.
5223 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5226 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
5227 fold_for_warn before checking if arg2 is INTEGER_CST.
5229 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
5231 * config/rs6000/predicates.md (load_multiple_operation): Delete.
5232 (store_multiple_operation): Delete.
5233 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
5234 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
5235 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
5236 guarded by TARGET_STRING.
5237 (rs6000_output_load_multiple): Delete.
5238 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
5239 OPTION_MASK_STRING / TARGET_STRING handling.
5240 (print_operand) <'N', 'O'>: Add comment that these are unused now.
5241 (const rs6000_opt_masks) <"string">: Change mask to 0.
5242 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
5243 (MASK_STRING): Delete.
5244 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
5246 (load_multiple): Delete.
5253 (store_multiple): Delete.
5260 (movmemsi_8reg): Delete.
5261 (corresponding unnamed define_insn): Delete.
5262 (movmemsi_6reg): Delete.
5263 (corresponding unnamed define_insn): Delete.
5264 (movmemsi_4reg): Delete.
5265 (corresponding unnamed define_insn): Delete.
5266 (movmemsi_2reg): Delete.
5267 (corresponding unnamed define_insn): Delete.
5268 (movmemsi_1reg): Delete.
5269 (corresponding unnamed define_insn): Delete.
5270 * config/rs6000/rs6000.opt (mno-string): New.
5271 (mstring): Replace by deprecation warning stub.
5272 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
5274 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5276 * regrename.c (regrename_do_replace): If replacing the same
5277 reg multiple times, try to reuse last created gen_raw_REG.
5280 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
5281 main to workaround a bug in GDB.
5283 2018-01-12 Tom de Vries <tom@codesourcery.com>
5286 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
5288 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
5290 PR rtl-optimization/80481
5291 * ira-color.c (get_cap_member): New function.
5292 (allocnos_conflict_by_live_ranges_p): Use it.
5293 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
5294 (setup_slot_coalesced_allocno_live_ranges): Ditto.
5296 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
5299 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
5300 (*saddl_se_1): Ditto.
5302 (*ssubl_se_1): Ditto.
5304 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5306 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
5307 rather than wi::to_widest for DR_INITs.
5308 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
5309 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
5310 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
5312 (vect_analyze_group_access_1): Note that here.
5314 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5316 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
5317 polynomial type sizes.
5319 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5321 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
5322 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
5323 (gimple_add_tmp_var): Likewise.
5325 2018-01-12 Martin Liska <mliska@suse.cz>
5327 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
5328 (gimple_alloc_sizes): Likewise.
5329 (dump_gimple_statistics): Use PRIu64 in printf format.
5330 * gimple.h: Change uint64_t to int.
5332 2018-01-12 Martin Liska <mliska@suse.cz>
5334 * tree-core.h: Use uint64_t instead of int.
5335 * tree.c (tree_node_counts): Likewise.
5336 (tree_node_sizes): Likewise.
5337 (dump_tree_statistics): Use PRIu64 in printf format.
5339 2018-01-12 Martin Liska <mliska@suse.cz>
5341 * Makefile.in: As qsort_chk is implemented in vec.c, add
5342 vec.o to linkage of gencfn-macros.
5343 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
5344 passing the info to record_node_allocation_statistics.
5345 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
5347 * ggc-common.c (struct ggc_usage): Add operator== and use
5348 it in operator< and compare function.
5349 * mem-stats.h (struct mem_usage): Likewise.
5350 * vec.c (struct vec_usage): Remove operator< and compare
5351 function. Can be simply inherited.
5353 2018-01-12 Martin Jambor <mjambor@suse.cz>
5356 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
5357 * tree-ssa-math-opts.c: Include domwalk.h.
5358 (convert_mult_to_fma_1): New function.
5359 (fma_transformation_info): New type.
5360 (fma_deferring_state): Likewise.
5361 (cancel_fma_deferring): New function.
5362 (result_of_phi): Likewise.
5363 (last_fma_candidate_feeds_initial_phi): Likewise.
5364 (convert_mult_to_fma): Added deferring logic, split actual
5365 transformation to convert_mult_to_fma_1.
5366 (math_opts_dom_walker): New type.
5367 (math_opts_dom_walker::after_dom_children): New method, body moved
5368 here from pass_optimize_widening_mul::execute, added deferring logic
5370 (pass_optimize_widening_mul::execute): Moved most of code to
5371 math_opts_dom_walker::after_dom_children.
5372 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
5373 * config/i386/i386.c (ix86_option_override_internal): Added
5374 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
5376 2018-01-12 Richard Biener <rguenther@suse.de>
5379 * dwarf2out.c (gen_variable_die): Do not reset old_die for
5380 inline instance vars.
5382 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
5385 * config/rx/rx.c (rx_is_restricted_memory_address):
5388 2018-01-12 Richard Biener <rguenther@suse.de>
5390 PR tree-optimization/80846
5391 * target.def (split_reduction): New target hook.
5392 * targhooks.c (default_split_reduction): New function.
5393 * targhooks.h (default_split_reduction): Declare.
5394 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
5395 target requests first reduce vectors by combining low and high
5397 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
5398 (get_vectype_for_scalar_type_and_size): Export.
5399 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
5400 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
5401 * doc/tm.texi: Regenerate.
5402 * config/i386/i386.c (ix86_split_reduction): Implement
5403 TARGET_VECTORIZE_SPLIT_REDUCTION.
5405 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5408 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
5409 in PIC mode except for TARGET_VXWORKS_RTP.
5410 * config/sparc/sparc.c: Include cfgrtl.h.
5411 (TARGET_INIT_PIC_REG): Define.
5412 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
5413 (sparc_pic_register_p): New predicate.
5414 (sparc_legitimate_address_p): Use it.
5415 (sparc_legitimize_pic_address): Likewise.
5416 (sparc_delegitimize_address): Likewise.
5417 (sparc_mode_dependent_address_p): Likewise.
5418 (gen_load_pcrel_sym): Remove 4th parameter.
5419 (load_got_register): Adjust call to above. Remove obsolete stuff.
5420 (sparc_expand_prologue): Do not call load_got_register here.
5421 (sparc_flat_expand_prologue): Likewise.
5422 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
5423 (sparc_use_pseudo_pic_reg): New function.
5424 (sparc_init_pic_reg): Likewise.
5425 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5426 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
5428 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
5430 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
5431 Add item for branch_cost.
5433 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5435 PR rtl-optimization/83565
5436 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
5437 not extend the result to a larger mode for rotate operations.
5438 (num_sign_bit_copies1): Likewise.
5440 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5443 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
5445 Use values-Xc.o for -pedantic.
5446 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
5448 2018-01-12 Martin Liska <mliska@suse.cz>
5451 * ipa-devirt.c (final_warning_record::grow_type_warnings):
5453 (possible_polymorphic_call_targets): Use it.
5454 (ipa_devirt): Likewise.
5456 2018-01-12 Martin Liska <mliska@suse.cz>
5458 * profile-count.h (enum profile_quality): Use 0 as invalid
5459 enum value of profile_quality.
5461 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
5463 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
5464 -mext-string options.
5466 2018-01-12 Richard Biener <rguenther@suse.de>
5468 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
5469 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
5470 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
5472 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
5474 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5476 * configure.ac (--with-long-double-format): Add support for the
5477 configuration option to change the default long double format on
5479 * config.gcc (powerpc*-linux*-*): Likewise.
5480 * configure: Regenerate.
5481 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
5482 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
5483 used without modification.
5485 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5487 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
5488 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
5489 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
5490 MISC_BUILTIN_SPEC_BARRIER.
5491 (rs6000_init_builtins): Likewise.
5492 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
5494 (speculation_barrier): New define_insn.
5495 * doc/extend.texi: Document __builtin_speculation_barrier.
5497 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5500 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
5501 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
5502 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
5504 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
5505 integral modes instead of "ss" and "sd".
5506 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
5507 vectors with 32-bit and 64-bit elements.
5508 (vecdupssescalarmodesuffix): New mode attribute.
5509 (vec_dup<mode>): Use it.
5511 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
5514 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
5515 frame if argument is passed on stack.
5517 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5520 * ree.c (combine_reaching_defs): Optimize also
5521 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
5522 reg2=any_extend(exp); reg1=reg2;, formatting fix.
5524 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5527 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
5529 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5532 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
5533 after they are computed.
5535 2018-01-11 Bin Cheng <bin.cheng@arm.com>
5537 PR tree-optimization/83695
5538 * gimple-loop-linterchange.cc
5539 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
5540 reset cached scev information after interchange.
5541 (pass_linterchange::execute): Remove call to scev_reset_htab.
5543 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5545 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
5546 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
5547 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
5548 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
5549 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
5550 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
5551 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
5552 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
5553 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
5554 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
5555 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
5556 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
5557 (V_lane_reg): Likewise.
5558 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
5560 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
5561 (vfmal_lane_low<mode>_intrinsic,
5562 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
5563 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
5564 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
5565 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
5566 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
5567 vfmsl_lane_high<mode>_intrinsic): New define_insns.
5569 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5571 * config/arm/arm-cpus.in (fp16fml): New feature.
5572 (ALL_SIMD): Add fp16fml.
5573 (armv8.2-a): Add fp16fml as an option.
5574 (armv8.3-a): Likewise.
5575 (armv8.4-a): Add fp16fml as part of fp16.
5576 * config/arm/arm.h (TARGET_FP16FML): Define.
5577 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
5579 * config/arm/arm-modes.def (V2HF): Define.
5580 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
5581 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
5582 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
5583 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
5584 vfmsl_low, vfmsl_high): New set of builtins.
5585 * config/arm/iterators.md (PLUSMINUS): New code iterator.
5586 (vfml_op): New code attribute.
5587 (VFMLHALVES): New int iterator.
5588 (VFML, VFMLSEL): New mode attributes.
5589 (V_reg): Define mapping for V2HF.
5590 (V_hi, V_lo): New mode attributes.
5591 (VF_constraint): Likewise.
5592 (vfml_half, vfml_half_selector): New int attributes.
5593 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
5595 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
5596 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
5598 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
5599 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
5600 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
5601 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
5603 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
5604 Document new effective target and option set.
5606 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5608 * config/arm/arm-cpus.in (armv8_4): New feature.
5609 (ARMv8_4a): New fgroup.
5610 (armv8.4-a): New arch.
5611 * config/arm/arm-tables.opt: Regenerate.
5612 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
5613 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
5614 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
5615 Add matching rules for -march=armv8.4-a and extensions.
5616 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
5618 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
5621 * config/rx/rx.md (BW): New mode attribute.
5622 (sync_lock_test_and_setsi): Add mode suffix to insn output.
5624 2018-01-11 Richard Biener <rguenther@suse.de>
5626 PR tree-optimization/83435
5627 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
5628 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
5629 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
5631 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5632 Alan Hayward <alan.hayward@arm.com>
5633 David Sherwood <david.sherwood@arm.com>
5635 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
5637 (aarch64_classify_address): Initialize it. Track polynomial offsets.
5638 (aarch64_print_address_internal): Use it to check for a zero offset.
5640 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5641 Alan Hayward <alan.hayward@arm.com>
5642 David Sherwood <david.sherwood@arm.com>
5644 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
5645 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
5646 Return a poly_int64 rather than a HOST_WIDE_INT.
5647 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
5648 rather than a HOST_WIDE_INT.
5649 * config/aarch64/aarch64.h (aarch64_frame): Protect with
5650 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
5651 hard_fp_offset, frame_size, initial_adjust, callee_offset and
5652 final_offset from HOST_WIDE_INT to poly_int64.
5653 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
5654 to_constant when getting the number of units in an Advanced SIMD
5656 (aarch64_builtin_vectorized_function): Check for a constant number
5658 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
5660 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
5661 attribute instead of GET_MODE_NUNITS.
5662 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
5663 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
5664 GET_MODE_SIZE for fixed-size registers.
5665 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
5666 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
5667 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
5668 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
5669 (aarch64_print_operand, aarch64_print_address_internal)
5670 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
5671 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
5672 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
5673 Handle polynomial GET_MODE_SIZE.
5674 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
5675 wider than SImode without modification.
5676 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
5677 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
5678 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
5679 passing and returning SVE modes.
5680 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
5681 rather than GEN_INT.
5682 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
5683 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
5684 (aarch64_allocate_and_probe_stack_space): Likewise.
5685 (aarch64_layout_frame): Cope with polynomial offsets.
5686 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
5687 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
5689 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
5690 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
5691 poly_int64 rather than a HOST_WIDE_INT.
5692 (aarch64_get_separate_components, aarch64_process_components)
5693 (aarch64_expand_prologue, aarch64_expand_epilogue)
5694 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
5695 (aarch64_anchor_offset): New function, split out from...
5696 (aarch64_legitimize_address): ...here.
5697 (aarch64_builtin_vectorization_cost): Handle polynomial
5698 TYPE_VECTOR_SUBPARTS.
5699 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
5701 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
5702 number of elements from the PARALLEL rather than the mode.
5703 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
5704 rather than GET_MODE_BITSIZE.
5705 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
5706 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
5707 (aarch64_expand_vec_perm_const_1): Handle polynomial
5708 d->perm.length () and d->perm elements.
5709 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
5710 Apply to_constant to d->perm elements.
5711 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
5712 polynomial CONST_VECTOR_NUNITS.
5713 (aarch64_move_pointer): Take amount as a poly_int64 rather
5715 (aarch64_progress_pointer): Avoid temporary variable.
5716 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
5717 the mode attribute instead of GET_MODE.
5719 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5720 Alan Hayward <alan.hayward@arm.com>
5721 David Sherwood <david.sherwood@arm.com>
5723 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
5724 x exists before using it.
5725 (aarch64_add_constant_internal): Rename to...
5726 (aarch64_add_offset_1): ...this. Replace regnum with separate
5727 src and dest rtxes. Handle the case in which they're different,
5728 including when the offset is zero. Replace scratchreg with an rtx.
5729 Use 2 additions if there is no spare register into which we can
5730 move a 16-bit constant.
5731 (aarch64_add_constant): Delete.
5732 (aarch64_add_offset): Replace reg with separate src and dest
5733 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
5734 Use aarch64_add_offset_1.
5735 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
5736 an rtx rather than an int. Take the delta as a poly_int64
5737 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
5738 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
5739 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
5740 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
5741 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
5743 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
5744 aarch64_add_constant.
5746 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5748 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
5749 Use scalar_float_mode.
5751 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5753 * config/aarch64/aarch64-simd.md
5754 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
5755 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
5756 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
5757 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
5758 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
5759 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
5760 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
5761 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
5762 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
5763 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
5765 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5768 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
5769 targ_options->x_arm_arch_string is non NULL.
5771 2018-01-11 Tamar Christina <tamar.christina@arm.com>
5773 * config/aarch64/aarch64.h
5774 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
5776 2018-01-11 Sudakshina Das <sudi.das@arm.com>
5779 * expmed.c (emit_store_flag_force): Swap if const op0
5780 and change VOIDmode to mode of op0.
5782 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5784 PR rtl-optimization/83761
5785 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
5786 than bytes to mode_for_size.
5788 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5791 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
5792 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
5795 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5798 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
5799 when in layout mode.
5800 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
5801 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
5804 2018-01-10 Michael Collison <michael.collison@arm.com>
5806 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
5807 * config/aarch64/aarch64-option-extension.def: Add
5808 AARCH64_OPT_EXTENSION of 'fp16fml'.
5809 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5810 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
5811 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
5812 * config/aarch64/constraints.md (Ui7): New constraint.
5813 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
5814 (VFMLA_SEL_W): Ditto.
5817 (VFMLA16_LOW): New int iterator.
5818 (VFMLA16_HIGH): Ditto.
5819 (UNSPEC_FMLAL): New unspec.
5820 (UNSPEC_FMLSL): Ditto.
5821 (UNSPEC_FMLAL2): Ditto.
5822 (UNSPEC_FMLSL2): Ditto.
5823 (f16mac): New code attribute.
5824 * config/aarch64/aarch64-simd-builtins.def
5825 (aarch64_fmlal_lowv2sf): Ditto.
5826 (aarch64_fmlsl_lowv2sf): Ditto.
5827 (aarch64_fmlalq_lowv4sf): Ditto.
5828 (aarch64_fmlslq_lowv4sf): Ditto.
5829 (aarch64_fmlal_highv2sf): Ditto.
5830 (aarch64_fmlsl_highv2sf): Ditto.
5831 (aarch64_fmlalq_highv4sf): Ditto.
5832 (aarch64_fmlslq_highv4sf): Ditto.
5833 (aarch64_fmlal_lane_lowv2sf): Ditto.
5834 (aarch64_fmlsl_lane_lowv2sf): Ditto.
5835 (aarch64_fmlal_laneq_lowv2sf): Ditto.
5836 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
5837 (aarch64_fmlalq_lane_lowv4sf): Ditto.
5838 (aarch64_fmlsl_lane_lowv4sf): Ditto.
5839 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
5840 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
5841 (aarch64_fmlal_lane_highv2sf): Ditto.
5842 (aarch64_fmlsl_lane_highv2sf): Ditto.
5843 (aarch64_fmlal_laneq_highv2sf): Ditto.
5844 (aarch64_fmlsl_laneq_highv2sf): Ditto.
5845 (aarch64_fmlalq_lane_highv4sf): Ditto.
5846 (aarch64_fmlsl_lane_highv4sf): Ditto.
5847 (aarch64_fmlalq_laneq_highv4sf): Ditto.
5848 (aarch64_fmlsl_laneq_highv4sf): Ditto.
5849 * config/aarch64/aarch64-simd.md:
5850 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
5851 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5852 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
5853 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5854 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
5855 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
5856 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
5857 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
5858 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
5859 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
5860 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
5861 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
5862 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
5863 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
5864 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
5865 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
5866 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
5867 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
5868 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
5869 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
5870 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
5871 (vfmlsl_low_u32): Ditto.
5872 (vfmlalq_low_u32): Ditto.
5873 (vfmlslq_low_u32): Ditto.
5874 (vfmlal_high_u32): Ditto.
5875 (vfmlsl_high_u32): Ditto.
5876 (vfmlalq_high_u32): Ditto.
5877 (vfmlslq_high_u32): Ditto.
5878 (vfmlal_lane_low_u32): Ditto.
5879 (vfmlsl_lane_low_u32): Ditto.
5880 (vfmlal_laneq_low_u32): Ditto.
5881 (vfmlsl_laneq_low_u32): Ditto.
5882 (vfmlalq_lane_low_u32): Ditto.
5883 (vfmlslq_lane_low_u32): Ditto.
5884 (vfmlalq_laneq_low_u32): Ditto.
5885 (vfmlslq_laneq_low_u32): Ditto.
5886 (vfmlal_lane_high_u32): Ditto.
5887 (vfmlsl_lane_high_u32): Ditto.
5888 (vfmlal_laneq_high_u32): Ditto.
5889 (vfmlsl_laneq_high_u32): Ditto.
5890 (vfmlalq_lane_high_u32): Ditto.
5891 (vfmlslq_lane_high_u32): Ditto.
5892 (vfmlalq_laneq_high_u32): Ditto.
5893 (vfmlslq_laneq_high_u32): Ditto.
5894 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
5895 (AARCH64_FL_FOR_ARCH8_4): New.
5896 (AARCH64_ISA_F16FML): New ISA flag.
5897 (TARGET_F16FML): New feature flag for fp16fml.
5898 (doc/invoke.texi): Document new fp16fml option.
5900 2018-01-10 Michael Collison <michael.collison@arm.com>
5902 * config/aarch64/aarch64-builtins.c:
5903 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
5904 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5905 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
5906 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
5907 (AARCH64_ISA_SHA3): New ISA flag.
5908 (TARGET_SHA3): New feature flag for sha3.
5909 * config/aarch64/iterators.md (sha512_op): New int attribute.
5910 (CRYPTO_SHA512): New int iterator.
5911 (UNSPEC_SHA512H): New unspec.
5912 (UNSPEC_SHA512H2): Ditto.
5913 (UNSPEC_SHA512SU0): Ditto.
5914 (UNSPEC_SHA512SU1): Ditto.
5915 * config/aarch64/aarch64-simd-builtins.def
5916 (aarch64_crypto_sha512hqv2di): New builtin.
5917 (aarch64_crypto_sha512h2qv2di): Ditto.
5918 (aarch64_crypto_sha512su0qv2di): Ditto.
5919 (aarch64_crypto_sha512su1qv2di): Ditto.
5920 (aarch64_eor3qv8hi): Ditto.
5921 (aarch64_rax1qv2di): Ditto.
5922 (aarch64_xarqv2di): Ditto.
5923 (aarch64_bcaxqv8hi): Ditto.
5924 * config/aarch64/aarch64-simd.md:
5925 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
5926 (aarch64_crypto_sha512su0qv2di): Ditto.
5927 (aarch64_crypto_sha512su1qv2di): Ditto.
5928 (aarch64_eor3qv8hi): Ditto.
5929 (aarch64_rax1qv2di): Ditto.
5930 (aarch64_xarqv2di): Ditto.
5931 (aarch64_bcaxqv8hi): Ditto.
5932 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
5933 (vsha512h2q_u64): Ditto.
5934 (vsha512su0q_u64): Ditto.
5935 (vsha512su1q_u64): Ditto.
5936 (veor3q_u16): Ditto.
5937 (vrax1q_u64): Ditto.
5939 (vbcaxq_u16): Ditto.
5940 * config/arm/types.md (crypto_sha512): New type attribute.
5941 (crypto_sha3): Ditto.
5942 (doc/invoke.texi): Document new sha3 option.
5944 2018-01-10 Michael Collison <michael.collison@arm.com>
5946 * config/aarch64/aarch64-builtins.c:
5947 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
5948 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5949 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
5950 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
5951 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
5952 (AARCH64_ISA_SM4): New ISA flag.
5953 (TARGET_SM4): New feature flag for sm4.
5954 * config/aarch64/aarch64-simd-builtins.def
5955 (aarch64_sm3ss1qv4si): Ditto.
5956 (aarch64_sm3tt1aq4si): Ditto.
5957 (aarch64_sm3tt1bq4si): Ditto.
5958 (aarch64_sm3tt2aq4si): Ditto.
5959 (aarch64_sm3tt2bq4si): Ditto.
5960 (aarch64_sm3partw1qv4si): Ditto.
5961 (aarch64_sm3partw2qv4si): Ditto.
5962 (aarch64_sm4eqv4si): Ditto.
5963 (aarch64_sm4ekeyqv4si): Ditto.
5964 * config/aarch64/aarch64-simd.md:
5965 (aarch64_sm3ss1qv4si): Ditto.
5966 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
5967 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
5968 (aarch64_sm4eqv4si): Ditto.
5969 (aarch64_sm4ekeyqv4si): Ditto.
5970 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
5971 (sm3part_op): Ditto.
5972 (CRYPTO_SM3TT): Ditto.
5973 (CRYPTO_SM3PART): Ditto.
5974 (UNSPEC_SM3SS1): New unspec.
5975 (UNSPEC_SM3TT1A): Ditto.
5976 (UNSPEC_SM3TT1B): Ditto.
5977 (UNSPEC_SM3TT2A): Ditto.
5978 (UNSPEC_SM3TT2B): Ditto.
5979 (UNSPEC_SM3PARTW1): Ditto.
5980 (UNSPEC_SM3PARTW2): Ditto.
5981 (UNSPEC_SM4E): Ditto.
5982 (UNSPEC_SM4EKEY): Ditto.
5983 * config/aarch64/constraints.md (Ui2): New constraint.
5984 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
5985 * config/arm/types.md (crypto_sm3): New type attribute.
5986 (crypto_sm4): Ditto.
5987 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
5988 (vsm3tt1aq_u32): Ditto.
5989 (vsm3tt1bq_u32): Ditto.
5990 (vsm3tt2aq_u32): Ditto.
5991 (vsm3tt2bq_u32): Ditto.
5992 (vsm3partw1q_u32): Ditto.
5993 (vsm3partw2q_u32): Ditto.
5994 (vsm4eq_u32): Ditto.
5995 (vsm4ekeyq_u32): Ditto.
5996 (doc/invoke.texi): Document new sm4 option.
5998 2018-01-10 Michael Collison <michael.collison@arm.com>
6000 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
6001 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
6002 (AARCH64_FL_FOR_ARCH8_4): New.
6003 (AARCH64_FL_V8_4): New flag.
6004 (doc/invoke.texi): Document new armv8.4-a option.
6006 2018-01-10 Michael Collison <michael.collison@arm.com>
6008 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
6009 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
6010 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
6011 * config/aarch64/aarch64-option-extension.def: Add
6012 AARCH64_OPT_EXTENSION of 'sha2'.
6013 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
6014 (crypto): Disable sha2 and aes if crypto disabled.
6015 (crypto): Enable aes and sha2 if enabled.
6016 (simd): Disable sha2 and aes if simd disabled.
6017 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
6019 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
6020 (TARGET_SHA2): New feature flag for sha2.
6021 (TARGET_AES): New feature flag for aes.
6022 * config/aarch64/aarch64-simd.md:
6023 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
6024 conditional on TARGET_AES.
6025 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
6026 (aarch64_crypto_sha1hsi): Make pattern conditional
6028 (aarch64_crypto_sha1hv4si): Ditto.
6029 (aarch64_be_crypto_sha1hv4si): Ditto.
6030 (aarch64_crypto_sha1su1v4si): Ditto.
6031 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
6032 (aarch64_crypto_sha1su0v4si): Ditto.
6033 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
6034 (aarch64_crypto_sha256su0v4si): Ditto.
6035 (aarch64_crypto_sha256su1v4si): Ditto.
6036 (doc/invoke.texi): Document new aes and sha2 options.
6038 2018-01-10 Martin Sebor <msebor@redhat.com>
6040 PR tree-optimization/83781
6041 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
6044 2018-01-11 Martin Sebor <msebor@gmail.com>
6045 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6047 PR tree-optimization/83501
6048 PR tree-optimization/81703
6050 * tree-ssa-strlen.c (get_string_cst): Rename...
6051 (get_string_len): ...to this. Handle global constants.
6052 (handle_char_store): Adjust.
6054 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
6055 Jim Wilson <jimw@sifive.com>
6057 * config/riscv/riscv-protos.h (riscv_output_return): New.
6058 * config/riscv/riscv.c (struct machine_function): New naked_p field.
6059 (riscv_attribute_table, riscv_output_return),
6060 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
6061 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
6062 (riscv_compute_frame_info): Only compute frame->mask if not a naked
6064 (riscv_expand_prologue): Add early return for naked function.
6065 (riscv_expand_epilogue): Likewise.
6066 (riscv_function_ok_for_sibcall): Return false for naked function.
6067 (riscv_set_current_function): New.
6068 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
6069 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
6070 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
6071 * doc/extend.texi (RISC-V Function Attributes): New.
6073 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6075 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
6076 check for 128-bit long double before checking TCmode.
6077 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
6078 128-bit long doubles before checking TFmode or TCmode.
6079 (FLOAT128_IBM_P): Likewise.
6081 2018-01-10 Martin Sebor <msebor@redhat.com>
6083 PR tree-optimization/83671
6084 * builtins.c (c_strlen): Unconditionally return zero for the empty
6086 Use -Warray-bounds for warnings.
6087 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
6088 for non-constant array indices with COMPONENT_REF, arrays of
6089 arrays, and pointers to arrays.
6090 (gimple_fold_builtin_strlen): Determine and set length range for
6091 non-constant character arrays.
6093 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
6096 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
6099 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
6101 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
6103 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6106 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
6107 VECTOR_MEM_ALTIVEC_OR_VSX_P.
6108 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
6109 indexed_or_indirect_operand predicate.
6110 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
6111 (*vsx_le_perm_load_v8hi): Likewise.
6112 (*vsx_le_perm_load_v16qi): Likewise.
6113 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
6114 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
6115 (*vsx_le_perm_store_v8hi): Likewise.
6116 (*vsx_le_perm_store_v16qi): Likewise.
6117 (eight unnamed splitters): Likewise.
6119 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6121 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
6122 * config/rs6000/emmintrin.h: Likewise.
6123 * config/rs6000/mmintrin.h: Likewise.
6124 * config/rs6000/xmmintrin.h: Likewise.
6126 2018-01-10 David Malcolm <dmalcolm@redhat.com>
6129 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
6131 * tree.c (tree_nop_conversion): Return true for location wrapper
6133 (maybe_wrap_with_location): New function.
6134 (selftest::check_strip_nops): New function.
6135 (selftest::test_location_wrappers): New function.
6136 (selftest::tree_c_tests): Call it.
6137 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
6138 (maybe_wrap_with_location): New decl.
6139 (EXPR_LOCATION_WRAPPER_P): New macro.
6140 (location_wrapper_p): New inline function.
6141 (tree_strip_any_location_wrapper): New inline function.
6143 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
6146 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
6147 stack_realign_offset for the largest alignment of stack slot
6149 (ix86_find_max_used_stack_alignment): New function.
6150 (ix86_finalize_stack_frame_flags): Use it. Set
6151 max_used_stack_alignment if we don't realign stack.
6152 * config/i386/i386.h (machine_function): Add
6153 max_used_stack_alignment.
6155 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
6157 * config/arm/arm.opt (-mbranch-cost): New option.
6158 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
6161 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
6164 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
6165 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
6167 2018-01-10 Richard Biener <rguenther@suse.de>
6170 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
6171 early out so it also covers the case where we have a non-NULL
6174 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6176 PR tree-optimization/83753
6177 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
6178 for non-strided grouped accesses if the number of elements is 1.
6180 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6183 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
6184 * i386.h (TARGET_USE_GATHER): Define.
6185 * x86-tune.def (X86_TUNE_USE_GATHER): New.
6187 2018-01-10 Martin Liska <mliska@suse.cz>
6190 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
6191 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
6193 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
6194 CLEANUP_NO_PARTITIONING is not set.
6196 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6198 * doc/rtl.texi: Remove documentation of (const ...) wrappers
6199 for vectors, as a partial revert of r254296.
6200 * rtl.h (const_vec_p): Delete.
6201 (const_vec_duplicate_p): Don't test for vector CONSTs.
6202 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
6203 * expmed.c (make_tree): Likewise.
6206 * common.md (E, F): Use CONSTANT_P instead of checking for
6208 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
6209 checking for CONST_VECTOR.
6211 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6214 * predict.c (force_edge_cold): Handle in more sane way edges
6217 2018-01-09 Carl Love <cel@us.ibm.com>
6219 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
6221 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
6222 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
6223 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
6224 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
6225 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
6226 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
6227 * config/rs6000/rs6000-protos.h: Add extern defition for
6228 rs6000_generate_float2_double_code.
6229 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
6231 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
6232 (float2_v2df): Add define_expand.
6234 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
6237 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
6238 op_mode in the force_to_mode call.
6240 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6242 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
6243 instead of checking each element individually.
6244 (aarch64_evpc_uzp): Likewise.
6245 (aarch64_evpc_zip): Likewise.
6246 (aarch64_evpc_ext): Likewise.
6247 (aarch64_evpc_rev): Likewise.
6248 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
6249 instead of checking each element individually. Return true without
6251 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
6252 whether all selected elements come from the same input, instead of
6253 checking each element individually. Remove calls to gen_rtx_REG,
6254 start_sequence and end_sequence and instead assert that no rtl is
6257 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6259 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
6260 order of HIGH and CONST checks.
6262 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6264 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
6265 if the destination isn't an SSA_NAME.
6267 2018-01-09 Richard Biener <rguenther@suse.de>
6269 PR tree-optimization/83668
6270 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
6272 (canonicalize_loop_form): ... here, renamed from ...
6273 (canonicalize_loop_closed_ssa_form): ... this and amended to
6274 swap successor edges for loop exit blocks to make us use
6275 the RPO order we need for initial schedule generation.
6277 2018-01-09 Joseph Myers <joseph@codesourcery.com>
6279 PR tree-optimization/64811
6280 * match.pd: When optimizing comparisons with Inf, avoid
6281 introducing or losing exceptions from comparisons with NaN.
6283 2018-01-09 Martin Liska <mliska@suse.cz>
6286 * asan.c (shadow_mem_size): Add gcc_assert.
6288 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
6290 Don't save registers in main().
6293 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
6294 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
6295 * config/avr/avr.c (avr_set_current_function): Don't error if
6296 naked, OS_task or OS_main are specified at the same time.
6297 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
6299 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
6301 * common/config/avr/avr-common.c (avr_option_optimization_table):
6302 Switch on -mmain-is-OS_task for optimizing compilations.
6304 2018-01-09 Richard Biener <rguenther@suse.de>
6306 PR tree-optimization/83572
6307 * graphite.c: Include cfganal.h.
6308 (graphite_transform_loops): Connect infinite loops to exit
6309 and remove fake edges at the end.
6311 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6313 * ipa-inline.c (edge_badness): Revert accidental checkin.
6315 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6318 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
6319 symbols; not inline clones.
6321 2018-01-09 Jakub Jelinek <jakub@redhat.com>
6324 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
6325 hard registers. Formatting fixes.
6327 PR preprocessor/83722
6328 * gcc.c (try_generate_repro): Pass
6329 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
6330 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
6333 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
6334 Kito Cheng <kito.cheng@gmail.com>
6336 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
6337 (riscv_leaf_function_p): Delete.
6338 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
6340 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6342 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
6344 (do_ifelse): New function.
6345 (do_isel): New function.
6346 (do_sub3): New function.
6347 (do_add3): New function.
6348 (do_load_mask_compare): New function.
6349 (do_overlap_load_compare): New function.
6350 (expand_compare_loop): New function.
6351 (expand_block_compare): Call expand_compare_loop() when appropriate.
6352 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
6354 (-mblock-compare-inline-loop-limit): New option.
6356 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6359 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
6360 Reverse order of second and third operands in first alternative.
6361 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
6362 of first and second elements in UNSPEC_VPERMR vector.
6363 (altivec_expand_vec_perm_le): Likewise.
6365 2018-01-08 Jeff Law <law@redhat.com>
6367 PR rtl-optimizatin/81308
6368 * tree-switch-conversion.c (cfg_altered): New file scoped static.
6369 (process_switch): If group_case_labels makes a change, then set
6371 (pass_convert_switch::execute): If a switch is converted, then
6372 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
6374 PR rtl-optimization/81308
6375 * recog.c (split_all_insns): Conditionally cleanup the CFG after
6378 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
6380 PR target/83663 - Revert r255946
6381 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
6382 generation for cases where splatting a value is not useful.
6383 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
6384 across a vec_duplicate and a paradoxical subreg forming a vector
6385 mode to a vec_concat.
6387 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6389 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
6390 -march=armv8.3-a variants.
6391 * config/arm/t-multilib: Likewise.
6392 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
6394 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6396 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
6398 (cceq_ior_compare_complement): Give it a name so I can use it, and
6399 change boolean_or_operator predicate to boolean_operator so it can
6400 be used to generate a crand.
6401 (eqne): New code iterator.
6402 (bd/bd_neg): New code_attrs.
6403 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
6404 a single define_insn.
6405 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
6406 decrement (bdnzt/bdnzf/bdzt/bdzf).
6407 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
6408 with the new names of the branch decrement patterns, and added the
6409 names of the branch decrement conditional patterns.
6411 2018-01-08 Richard Biener <rguenther@suse.de>
6413 PR tree-optimization/83563
6414 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
6417 2018-01-08 Richard Biener <rguenther@suse.de>
6420 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
6422 2018-01-08 Richard Biener <rguenther@suse.de>
6424 PR tree-optimization/83685
6425 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
6426 references to abnormals.
6428 2018-01-08 Richard Biener <rguenther@suse.de>
6431 * dwarf2out.c (output_indirect_strings): Handle empty
6432 skeleton_debug_str_hash.
6433 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
6435 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6437 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
6438 (emit_store_direct): Likewise.
6439 (arc_trampoline_adjust_address): Likewise.
6440 (arc_asm_trampoline_template): New function.
6441 (arc_initialize_trampoline): Use asm_trampoline_template.
6442 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
6443 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
6444 * config/arc/arc.md (flush_icache): Delete pattern.
6446 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6448 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
6449 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
6452 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6455 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
6456 by not USED_FOR_TARGET.
6457 (make_pass_resolve_sw_modes): Likewise.
6459 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6461 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
6464 2018-01-08 Richard Biener <rguenther@suse.de>
6467 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
6469 2018-01-08 Richard Biener <rguenther@suse.de>
6472 * match.pd ((t * 2) / 2) -> t): Add missing :c.
6474 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
6477 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
6478 basic blocks with a small number of successors.
6479 (convert_control_dep_chain_into_preds): Improve handling of
6481 (dump_predicates): Split apart into...
6482 (dump_pred_chain): ...here...
6483 (dump_pred_info): ...and here.
6484 (can_one_predicate_be_invalidated_p): Add debugging printfs.
6485 (can_chain_union_be_invalidated_p): Improve check for invalidation
6487 (uninit_uses_cannot_happen): Avoid unnecessary if
6488 convert_control_dep_chain_into_preds yielded nothing.
6490 2018-01-06 Martin Sebor <msebor@redhat.com>
6492 PR tree-optimization/83640
6493 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
6494 subtracting negative offset from size.
6495 (builtin_access::overlap): Adjust offset bounds of the access to fall
6496 within the size of the object if possible.
6498 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
6500 PR rtl-optimization/83699
6501 * expmed.c (extract_bit_field_1): Restrict the vector usage of
6502 extract_bit_field_as_subreg to cases in which the extracted
6503 value is also a vector.
6505 * lra-constraints.c (process_alt_operands): Test for the equivalence
6506 substitutions when detecting a possible reload cycle.
6508 2018-01-06 Jakub Jelinek <jakub@redhat.com>
6511 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
6512 by default if flag_selective_schedling{,2}. Formatting fixes.
6514 PR rtl-optimization/83682
6515 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
6516 if it has non-VECTOR_MODE element mode.
6517 (vec_duplicate_p): Likewise.
6520 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
6521 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
6523 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6526 * config/i386/i386-builtin.def
6527 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
6528 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
6529 Require also OPTION_MASK_ISA_AVX512F in addition to
6530 OPTION_MASK_ISA_GFNI.
6531 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
6532 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
6533 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
6534 to OPTION_MASK_ISA_GFNI.
6535 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
6536 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
6537 OPTION_MASK_ISA_AVX512BW.
6538 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
6539 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
6540 addition to OPTION_MASK_ISA_GFNI.
6541 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
6542 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
6543 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
6544 to OPTION_MASK_ISA_GFNI.
6545 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
6546 a requirement for all ISAs rather than any of them with a few
6548 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
6550 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
6551 bitmasks to be enabled with 3 exceptions, instead of requiring any
6552 enabled ISA with lots of exceptions.
6553 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
6554 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
6555 Change avx512bw in isa attribute to avx512f.
6556 * config/i386/sgxintrin.h: Add license boilerplate.
6557 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
6558 to __AVX512F__ and __AVX512VL to __AVX512VL__.
6559 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
6560 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
6562 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
6563 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
6564 temporarily sse2 rather than sse if not enabled already.
6567 * config/i386/sse.md (VI248_VLBW): Rename to ...
6568 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
6569 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
6570 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
6571 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
6572 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
6573 mode iterator instead of VI248_VLBW.
6575 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
6577 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
6578 (record_modified): Skip clobbers; add debug output.
6579 (param_change_prob): Use sreal frequencies.
6581 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6583 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
6584 punt for user-aligned variables.
6586 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6588 * tree-chrec.c (chrec_contains_symbols): Return true for
6591 2018-01-05 Sudakshina Das <sudi.das@arm.com>
6594 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
6595 of (x|y) == x for BICS pattern.
6597 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6599 PR tree-optimization/83605
6600 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
6601 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
6604 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
6606 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
6607 * config/epiphany/rtems.h: New file.
6609 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6610 Uros Bizjak <ubizjak@gmail.com>
6613 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
6614 QIreg_operand instead of register_operand predicate.
6615 * config/i386/i386.c (ix86_rop_should_change_byte_p,
6616 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
6617 comments instead of -fmitigate[-_]rop.
6619 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6622 * cgraphunit.c (symbol_table::compile): Switch to text_section
6623 before calling assembly_start debug hook.
6624 * run-rtl-passes.c (run_rtl_passes): Likewise.
6627 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6629 * tree-vrp.c (extract_range_from_binary_expr_1): Check
6630 range_int_cst_p rather than !symbolic_range_p before calling
6631 extract_range_from_multiplicative_op_1.
6633 2018-01-04 Jeff Law <law@redhat.com>
6635 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
6636 redundant test in assertion.
6638 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6640 * doc/rtl.texi: Document machine_mode wrapper classes.
6642 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6644 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
6647 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6649 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
6650 the VEC_PERM_EXPR fold to fail.
6652 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6655 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
6656 to switched_sections.
6658 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6661 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
6664 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
6667 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
6668 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
6670 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6673 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
6674 is BLKmode and bitpos not zero or mode change is needed.
6676 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6679 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
6682 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
6685 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
6686 instead of MULT rtx. Update all corresponding splitters.
6688 (*ssub<modesuffix>): Ditto.
6690 (*cmp_sadd_di): Update split patterns.
6691 (*cmp_sadd_si): Ditto.
6692 (*cmp_sadd_sidi): Ditto.
6693 (*cmp_ssub_di): Ditto.
6694 (*cmp_ssub_si): Ditto.
6695 (*cmp_ssub_sidi): Ditto.
6696 * config/alpha/predicates.md (const23_operand): New predicate.
6697 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
6698 Look for ASHIFT, not MULT inner operand.
6699 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
6701 2018-01-04 Martin Liska <mliska@suse.cz>
6703 PR gcov-profile/83669
6704 * gcov.c (output_intermediate_file): Add version to intermediate
6706 * doc/gcov.texi: Document new field 'version' in intermediate
6707 file format. Fix location of '-k' option of gcov command.
6709 2018-01-04 Martin Liska <mliska@suse.cz>
6712 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
6714 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6716 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
6718 2018-01-03 Martin Sebor <msebor@redhat.com>
6720 PR tree-optimization/83655
6721 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
6722 checking calls with invalid arguments.
6724 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6726 * tree-vect-stmts.c (vect_get_store_rhs): New function.
6727 (vectorizable_mask_load_store): Delete.
6728 (vectorizable_call): Return false for masked loads and stores.
6729 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
6730 instead of gimple_assign_rhs1.
6731 (vectorizable_load): Handle IFN_MASK_LOAD.
6732 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
6734 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6736 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
6738 (vectorizable_mask_load_store): ...here.
6739 (vectorizable_load): ...and here.
6741 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6743 * tree-vect-stmts.c (vect_build_all_ones_mask)
6744 (vect_build_zero_merge_argument): New functions, split out from...
6745 (vectorizable_load): ...here.
6747 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6749 * tree-vect-stmts.c (vect_check_store_rhs): New function,
6751 (vectorizable_mask_load_store): ...here.
6752 (vectorizable_store): ...and here.
6754 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6756 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
6758 (vectorizable_mask_load_store): ...here.
6760 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6762 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
6763 (vect_model_store_cost): Take a vec_load_store_type instead of a
6765 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
6766 (vect_model_store_cost): Take a vec_load_store_type instead of a
6768 (vectorizable_mask_load_store): Update accordingly.
6769 (vectorizable_store): Likewise.
6770 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
6772 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6774 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
6775 IFN_MASK_LOAD calls here rather than...
6776 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
6778 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6779 Alan Hayward <alan.hayward@arm.com>
6780 David Sherwood <david.sherwood@arm.com>
6782 * expmed.c (extract_bit_field_1): For vector extracts,
6783 fall back to extract_bit_field_as_subreg if vec_extract
6786 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6787 Alan Hayward <alan.hayward@arm.com>
6788 David Sherwood <david.sherwood@arm.com>
6790 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
6791 they are variable or constant sized.
6792 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
6793 slots for constant-sized data.
6795 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6796 Alan Hayward <alan.hayward@arm.com>
6797 David Sherwood <david.sherwood@arm.com>
6799 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
6800 handling COND_EXPRs with boolean comparisons, try to find a better
6801 basis for the mask type than the boolean itself.
6803 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6805 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
6806 is calculated and how it can be overridden.
6807 * genmodes.c (max_bitsize_mode_any_mode): New variable.
6808 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
6810 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
6813 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6814 Alan Hayward <alan.hayward@arm.com>
6815 David Sherwood <david.sherwood@arm.com>
6817 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
6818 Remove the mode argument.
6819 (aarch64_simd_valid_immediate): Remove the mode and inverse
6821 * config/aarch64/iterators.md (bitsize): New iterator.
6822 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
6823 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
6824 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
6825 aarch64_simd_valid_immediate.
6826 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
6827 (aarch64_reg_or_bic_imm): Likewise.
6828 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
6829 with an insn_type enum and msl with a modifier_type enum.
6830 Replace element_width with a scalar_mode. Change the shift
6831 to unsigned int. Add constructors for scalar_float_mode and
6832 scalar_int_mode elements.
6833 (aarch64_vect_float_const_representable_p): Delete.
6834 (aarch64_can_const_movi_rtx_p)
6835 (aarch64_simd_scalar_immediate_valid_for_move)
6836 (aarch64_simd_make_constant): Update call to
6837 aarch64_simd_valid_immediate.
6838 (aarch64_advsimd_valid_immediate_hs): New function.
6839 (aarch64_advsimd_valid_immediate): Likewise.
6840 (aarch64_simd_valid_immediate): Remove mode and inverse
6841 arguments. Rewrite to use the above. Use const_vec_duplicate_p
6842 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
6843 and aarch64_float_const_representable_p on the result.
6844 (aarch64_output_simd_mov_immediate): Remove mode argument.
6845 Update call to aarch64_simd_valid_immediate and use of
6846 simd_immediate_info.
6847 (aarch64_output_scalar_simd_mov_immediate): Update call
6850 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6851 Alan Hayward <alan.hayward@arm.com>
6852 David Sherwood <david.sherwood@arm.com>
6854 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
6855 (mode_nunits): Likewise CONST_MODE_NUNITS.
6856 * machmode.def (ADJUST_NUNITS): Document.
6857 * genmodes.c (mode_data::need_nunits_adj): New field.
6858 (blank_mode): Update accordingly.
6859 (adj_nunits): New variable.
6860 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
6862 (emit_mode_size_inline): Set need_bytesize_adj for all modes
6863 listed in adj_nunits.
6864 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
6865 listed in adj_nunits. Don't emit case statements for such modes.
6866 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
6867 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
6868 nothing if adj_nunits is nonnull.
6869 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
6870 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
6871 (emit_mode_fbit): Update use of print_maybe_const_decl.
6872 (emit_move_size): Likewise. Treat the array as non-const
6874 (emit_mode_adjustments): Handle adj_nunits.
6876 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6878 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
6879 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
6880 (VECTOR_MODES): Use it.
6881 (make_vector_modes): Take the prefix as an argument.
6883 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6884 Alan Hayward <alan.hayward@arm.com>
6885 David Sherwood <david.sherwood@arm.com>
6887 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
6888 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
6889 for MODE_VECTOR_BOOL.
6890 * machmode.def (VECTOR_BOOL_MODE): Document.
6891 * genmodes.c (VECTOR_BOOL_MODE): New macro.
6892 (make_vector_bool_mode): New function.
6893 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
6895 * lto-streamer-in.c (lto_input_mode_table): Likewise.
6896 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
6898 * stor-layout.c (int_mode_for_mode): Likewise.
6899 * tree.c (build_vector_type_for_mode): Likewise.
6900 * varasm.c (output_constant_pool_2): Likewise.
6901 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
6902 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
6903 for MODE_VECTOR_BOOL.
6904 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
6905 of mode class checks.
6906 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
6907 instead of a list of mode class checks.
6908 (expand_vector_scalar_condition): Likewise.
6909 (type_for_widest_vector_mode): Handle BImode as an inner mode.
6911 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6912 Alan Hayward <alan.hayward@arm.com>
6913 David Sherwood <david.sherwood@arm.com>
6915 * machmode.h (mode_size): Change from unsigned short to
6917 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
6918 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6919 or if measurement_type is not polynomial.
6920 (fixed_size_mode::includes_p): Check for constant-sized modes.
6921 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
6922 return a poly_uint16 rather than an unsigned short.
6923 (emit_mode_size): Change the type of mode_size from unsigned short
6924 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
6925 (emit_mode_adjustments): Cope with polynomial vector sizes.
6926 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6928 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6930 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
6931 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
6932 * caller-save.c (setup_save_areas): Likewise.
6933 (replace_reg_with_saved_mem): Likewise.
6934 * calls.c (emit_library_call_value_1): Likewise.
6935 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
6936 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
6937 (gen_lowpart_for_combine): Likewise.
6938 * convert.c (convert_to_integer_1): Likewise.
6939 * cse.c (equiv_constant, cse_insn): Likewise.
6940 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
6941 (cselib_subst_to_values): Likewise.
6942 * dce.c (word_dce_process_block): Likewise.
6943 * df-problems.c (df_word_lr_mark_ref): Likewise.
6944 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
6945 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
6946 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
6947 (rtl_for_decl_location): Likewise.
6948 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
6949 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
6950 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
6951 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
6952 (expand_expr_real_1): Likewise.
6953 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
6954 (pad_below): Likewise.
6955 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6956 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
6957 * ira.c (get_subreg_tracking_sizes): Likewise.
6958 * ira-build.c (ira_create_allocno_objects): Likewise.
6959 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
6960 (ira_sort_regnos_for_alter_reg): Likewise.
6961 * ira-costs.c (record_operand_costs): Likewise.
6962 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
6963 (resolve_simple_move): Likewise.
6964 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
6965 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
6966 (lra_constraints): Likewise.
6967 (CONST_POOL_OK_P): Reject variable-sized modes.
6968 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
6969 (add_pseudo_to_slot, lra_spill): Likewise.
6970 * omp-low.c (omp_clause_aligned_alignment): Likewise.
6971 * optabs-query.c (get_best_extraction_insn): Likewise.
6972 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
6973 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
6974 (expand_mult_highpart, valid_multiword_target_p): Likewise.
6975 * recog.c (offsettable_address_addr_space_p): Likewise.
6976 * regcprop.c (maybe_mode_change): Likewise.
6977 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
6978 * regrename.c (build_def_use): Likewise.
6979 * regstat.c (dump_reg_info): Likewise.
6980 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
6981 (find_reloads, find_reloads_subreg_address): Likewise.
6982 * reload1.c (eliminate_regs_1): Likewise.
6983 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
6984 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
6985 (simplify_binary_operation_1, simplify_subreg): Likewise.
6986 * targhooks.c (default_function_arg_padding): Likewise.
6987 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
6988 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
6989 (verify_gimple_assign_ternary): Likewise.
6990 * tree-inline.c (estimate_move_cost): Likewise.
6991 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
6992 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
6993 (get_address_cost_ainc): Likewise.
6994 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
6995 (vect_supportable_dr_alignment): Likewise.
6996 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6997 (vectorizable_reduction): Likewise.
6998 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
6999 (vectorizable_operation, vectorizable_load): Likewise.
7000 * tree.c (build_same_sized_truth_vector_type): Likewise.
7001 * valtrack.c (cleanup_auto_inc_dec): Likewise.
7002 * var-tracking.c (emit_note_insn_var_location): Likewise.
7003 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
7004 (ADDR_VEC_ALIGN): Likewise.
7006 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7007 Alan Hayward <alan.hayward@arm.com>
7008 David Sherwood <david.sherwood@arm.com>
7010 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
7012 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
7013 or if measurement_type is polynomial.
7014 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
7015 * combine.c (make_extraction): Likewise.
7016 * dse.c (find_shift_sequence): Likewise.
7017 * dwarf2out.c (mem_loc_descriptor): Likewise.
7018 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
7019 (extract_bit_field, extract_low_bits): Likewise.
7020 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
7021 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
7022 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
7023 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
7024 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7025 * reload.c (find_reloads): Likewise.
7026 * reload1.c (alter_reg): Likewise.
7027 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
7028 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
7029 * tree-if-conv.c (predicate_mem_writes): Likewise.
7030 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
7031 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
7032 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
7033 * valtrack.c (dead_debug_insert_temp): Likewise.
7034 * varasm.c (mergeable_constant_section): Likewise.
7035 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
7037 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7038 Alan Hayward <alan.hayward@arm.com>
7039 David Sherwood <david.sherwood@arm.com>
7041 * expr.c (expand_assignment): Cope with polynomial mode sizes
7042 when assigning to a CONCAT.
7044 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7045 Alan Hayward <alan.hayward@arm.com>
7046 David Sherwood <david.sherwood@arm.com>
7048 * machmode.h (mode_precision): Change from unsigned short to
7050 (mode_to_precision): Return a poly_uint16 rather than an unsigned
7052 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
7053 or if measurement_type is not polynomial.
7054 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
7055 in which the mode is already known to be a scalar_int_mode.
7056 * genmodes.c (emit_mode_precision): Change the type of mode_precision
7057 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
7059 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7060 for GET_MODE_PRECISION.
7061 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7062 for GET_MODE_PRECISION.
7063 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
7065 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
7066 (expand_field_assignment, make_extraction): Likewise.
7067 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
7068 (get_last_value): Likewise.
7069 * convert.c (convert_to_integer_1): Likewise.
7070 * cse.c (cse_insn): Likewise.
7071 * expr.c (expand_expr_real_1): Likewise.
7072 * lra-constraints.c (simplify_operand_subreg): Likewise.
7073 * optabs-query.c (can_atomic_load_p): Likewise.
7074 * optabs.c (expand_atomic_load): Likewise.
7075 (expand_atomic_store): Likewise.
7076 * ree.c (combine_reaching_defs): Likewise.
7077 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
7078 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
7079 * tree.h (type_has_mode_precision_p): Likewise.
7080 * ubsan.c (instrument_si_overflow): Likewise.
7082 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7083 Alan Hayward <alan.hayward@arm.com>
7084 David Sherwood <david.sherwood@arm.com>
7086 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
7087 polynomial numbers of units.
7088 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
7089 (valid_vector_subparts_p): New function.
7090 (build_vector_type): Remove temporary shim and take the number
7091 of units as a poly_uint64 rather than an int.
7092 (build_opaque_vector_type): Take the number of units as a
7093 poly_uint64 rather than an int.
7094 * tree.c (build_vector_from_ctor): Handle polynomial
7095 TYPE_VECTOR_SUBPARTS.
7096 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
7097 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
7098 (build_vector_from_val): If the number of units is variable,
7099 use build_vec_duplicate_cst for constant operands and
7100 VEC_DUPLICATE_EXPR otherwise.
7101 (make_vector_type): Remove temporary is_constant ().
7102 (build_vector_type, build_opaque_vector_type): Take the number of
7103 units as a poly_uint64 rather than an int.
7104 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
7106 * cfgexpand.c (expand_debug_expr): Likewise.
7107 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
7108 (store_constructor, expand_expr_real_1): Likewise.
7109 (const_scalar_mask_from_tree): Likewise.
7110 * fold-const-call.c (fold_const_reduction): Likewise.
7111 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
7112 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
7113 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
7114 (fold_relational_const): Likewise.
7115 (native_interpret_vector): Likewise. Change the size from an
7116 int to an unsigned int.
7117 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
7118 TYPE_VECTOR_SUBPARTS.
7119 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
7120 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
7121 duplicating a non-constant operand into a variable-length vector.
7122 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
7123 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
7124 * ipa-icf.c (sem_variable::equals): Likewise.
7125 * match.pd: Likewise.
7126 * omp-simd-clone.c (simd_clone_subparts): Likewise.
7127 * print-tree.c (print_node): Likewise.
7128 * stor-layout.c (layout_type): Likewise.
7129 * targhooks.c (default_builtin_vectorization_cost): Likewise.
7130 * tree-cfg.c (verify_gimple_comparison): Likewise.
7131 (verify_gimple_assign_binary): Likewise.
7132 (verify_gimple_assign_ternary): Likewise.
7133 (verify_gimple_assign_single): Likewise.
7134 * tree-pretty-print.c (dump_generic_node): Likewise.
7135 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7136 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
7137 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
7138 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
7139 (vect_shift_permute_load_chain): Likewise.
7140 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
7141 (expand_vector_condition, optimize_vector_constructor): Likewise.
7142 (lower_vec_perm, get_compute_type): Likewise.
7143 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7144 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
7145 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
7146 (vect_recog_mask_conversion_pattern): Likewise.
7147 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
7148 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
7149 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7150 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
7151 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
7152 (vectorizable_shift, vectorizable_operation, vectorizable_store)
7153 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
7154 (supportable_widening_operation): Likewise.
7155 (supportable_narrowing_operation): Likewise.
7156 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
7158 * varasm.c (output_constant): Likewise.
7160 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7161 Alan Hayward <alan.hayward@arm.com>
7162 David Sherwood <david.sherwood@arm.com>
7164 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
7165 so that both the length == 3 and length != 3 cases set up their
7166 own permute vectors. Add comments explaining why we know the
7167 number of elements is constant.
7168 (vect_permute_load_chain): Likewise.
7170 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7171 Alan Hayward <alan.hayward@arm.com>
7172 David Sherwood <david.sherwood@arm.com>
7174 * machmode.h (mode_nunits): Change from unsigned char to
7176 (ONLY_FIXED_SIZE_MODES): New macro.
7177 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
7178 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
7179 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
7181 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
7182 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
7183 or if measurement_type is not polynomial.
7184 * genmodes.c (ZERO_COEFFS): New macro.
7185 (emit_mode_nunits_inline): Make mode_nunits_inline return a
7187 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
7188 Use ZERO_COEFFS when emitting initializers.
7189 * data-streamer.h (bp_pack_poly_value): New function.
7190 (bp_unpack_poly_value): Likewise.
7191 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7192 for GET_MODE_NUNITS.
7193 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7194 for GET_MODE_NUNITS.
7195 * tree.c (make_vector_type): Remove temporary shim and make
7196 the real function take the number of units as a poly_uint64
7198 (build_vector_type_for_mode): Handle polynomial nunits.
7199 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
7200 * emit-rtl.c (const_vec_series_p_1): Likewise.
7201 (gen_rtx_CONST_VECTOR): Likewise.
7202 * fold-const.c (test_vec_duplicate_folding): Likewise.
7203 * genrecog.c (validate_pattern): Likewise.
7204 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
7205 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7206 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
7207 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
7208 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
7209 * rtlanal.c (subreg_get_info): Likewise.
7210 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7211 (vect_grouped_load_supported): Likewise.
7212 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
7213 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
7214 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
7215 (simplify_const_unary_operation, simplify_binary_operation_1)
7216 (simplify_const_binary_operation, simplify_ternary_operation)
7217 (test_vector_ops_duplicate, test_vector_ops): Likewise.
7218 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
7219 instead of CONST_VECTOR_NUNITS.
7220 * varasm.c (output_constant_pool_2): Likewise.
7221 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
7222 explicit-encoded elements in the XVEC for variable-length vectors.
7224 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7226 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
7228 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7229 Alan Hayward <alan.hayward@arm.com>
7230 David Sherwood <david.sherwood@arm.com>
7232 * coretypes.h (fixed_size_mode): Declare.
7233 (fixed_size_mode_pod): New typedef.
7234 * builtins.h (target_builtins::x_apply_args_mode)
7235 (target_builtins::x_apply_result_mode): Change type to
7236 fixed_size_mode_pod.
7237 * builtins.c (apply_args_size, apply_result_size, result_vector)
7238 (expand_builtin_apply_args_1, expand_builtin_apply)
7239 (expand_builtin_return): Update accordingly.
7241 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7243 * cse.c (hash_rtx_cb): Hash only the encoded elements.
7244 * cselib.c (cselib_hash_rtx): Likewise.
7245 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
7246 CONST_VECTOR encoding.
7248 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7249 Jeff Law <law@redhat.com>
7252 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
7253 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
7254 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
7255 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
7258 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
7259 explicitly probe *sp in a noreturn function if there were any callee
7260 register saves or frame pointer is needed.
7262 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7265 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
7266 BLKmode for ternary, binary or unary expressions.
7269 * var-tracking.c (delete_vta_debug_insn): New inline function.
7270 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
7271 insns from get_insns () to NULL instead of each bb separately.
7272 Use delete_vta_debug_insn. No longer static.
7273 (vt_debug_insns_local, variable_tracking_main_1): Adjust
7274 delete_vta_debug_insns callers.
7275 * rtl.h (delete_vta_debug_insns): Declare.
7276 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
7277 instead of variable_tracking_main.
7279 2018-01-03 Martin Sebor <msebor@redhat.com>
7281 PR tree-optimization/83603
7282 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
7283 arguments past the endof the argument list in functions declared
7284 without a prototype.
7285 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
7286 Avoid checking when arguments are null.
7288 2018-01-03 Martin Sebor <msebor@redhat.com>
7291 * doc/extend.texi (attribute const): Fix a typo.
7292 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
7293 issuing -Wsuggest-attribute for void functions.
7295 2018-01-03 Martin Sebor <msebor@redhat.com>
7297 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
7298 offset_int::from instead of wide_int::to_shwi.
7299 (maybe_diag_overlap): Remove assertion.
7300 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
7301 * gimple-ssa-sprintf.c (format_directive): Same.
7302 (parse_directive): Same.
7303 (sprintf_dom_walker::compute_format_length): Same.
7304 (try_substitute_return_value): Same.
7306 2018-01-03 Jeff Law <law@redhat.com>
7309 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
7310 non-constant residual for zero at runtime and avoid probing in
7311 that case. Reorganize code for trailing problem to mirror handling
7314 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7316 PR tree-optimization/83501
7317 * tree-ssa-strlen.c (get_string_cst): New.
7318 (handle_char_store): Call get_string_cst.
7320 2018-01-03 Martin Liska <mliska@suse.cz>
7322 PR tree-optimization/83593
7323 * tree-ssa-strlen.c: Include tree-cfg.h.
7324 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
7325 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
7326 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
7328 (strlen_dom_walker::before_dom_children): Call
7329 gimple_purge_dead_eh_edges. Dump tranformation with details
7331 (strlen_dom_walker::before_dom_children): Update call by adding
7332 new argument cleanup_eh.
7333 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
7335 2018-01-03 Martin Liska <mliska@suse.cz>
7338 * cif-code.def (VARIADIC_THUNK): New enum value.
7339 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
7342 2018-01-03 Jan Beulich <jbeulich@suse.com>
7344 * sse.md (mov<mode>_internal): Tighten condition for when to use
7345 vmovdqu<ssescalarsize> for TI and OI modes.
7347 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7349 Update copyright years.
7351 2018-01-03 Martin Liska <mliska@suse.cz>
7354 * ipa-visibility.c (function_and_variable_visibility): Skip
7355 functions with noipa attribure.
7357 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7359 * gcc.c (process_command): Update copyright notice dates.
7360 * gcov-dump.c (print_version): Ditto.
7361 * gcov.c (print_version): Ditto.
7362 * gcov-tool.c (print_version): Ditto.
7363 * gengtype.c (create_file): Ditto.
7364 * doc/cpp.texi: Bump @copying's copyright year.
7365 * doc/cppinternals.texi: Ditto.
7366 * doc/gcc.texi: Ditto.
7367 * doc/gccint.texi: Ditto.
7368 * doc/gcov.texi: Ditto.
7369 * doc/install.texi: Ditto.
7370 * doc/invoke.texi: Ditto.
7372 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7374 * vector-builder.h (vector_builder::m_full_nelts): Change from
7375 unsigned int to poly_uint64.
7376 (vector_builder::full_nelts): Update prototype accordingly.
7377 (vector_builder::new_vector): Likewise.
7378 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
7379 (vector_builder::operator ==): Likewise.
7380 (vector_builder::finalize): Likewise.
7381 * int-vector-builder.h (int_vector_builder::int_vector_builder):
7382 Take the number of elements as a poly_uint64 rather than an
7384 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
7385 from unsigned int to poly_uint64.
7386 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
7387 (vec_perm_indices::new_vector): Likewise.
7388 (vec_perm_indices::length): Likewise.
7389 (vec_perm_indices::nelts_per_input): Likewise.
7390 (vec_perm_indices::input_nelts): Likewise.
7391 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
7392 number of elements per input as a poly_uint64 rather than an
7393 unsigned int. Use the original encoding for variable-length
7394 vectors, rather than clamping each individual element.
7395 For the second and subsequent elements in each pattern,
7396 clamp the step and base before clamping their sum.
7397 (vec_perm_indices::series_p): Handle polynomial element counts.
7398 (vec_perm_indices::all_in_range_p): Likewise.
7399 (vec_perm_indices_to_tree): Likewise.
7400 (vec_perm_indices_to_rtx): Likewise.
7401 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
7402 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
7403 (tree_vector_builder::new_binary_operation): Handle polynomial
7404 element counts. Return false if we need to know the number
7405 of elements at compile time.
7406 * fold-const.c (fold_vec_perm): Punt if the number of elements
7407 isn't known at compile time.
7409 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7411 * vec-perm-indices.h (vec_perm_builder): Change element type
7412 from HOST_WIDE_INT to poly_int64.
7413 (vec_perm_indices::element_type): Update accordingly.
7414 (vec_perm_indices::clamp): Handle polynomial element_types.
7415 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7416 (vec_perm_indices::all_in_range_p): Likewise.
7417 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
7419 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
7420 polynomial vec_perm_indices element types.
7421 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
7422 * fold-const.c (fold_vec_perm): Likewise.
7423 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
7424 * tree-vect-generic.c (lower_vec_perm): Likewise.
7425 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7426 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
7427 element type to HOST_WIDE_INT.
7429 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7430 Alan Hayward <alan.hayward@arm.com>
7431 David Sherwood <david.sherwood@arm.com>
7433 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
7434 rather than an int. Use plus_constant.
7435 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
7436 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
7438 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7439 Alan Hayward <alan.hayward@arm.com>
7440 David Sherwood <david.sherwood@arm.com>
7442 * calls.c (emit_call_1, expand_call): Change struct_value_size from
7443 a HOST_WIDE_INT to a poly_int64.
7445 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7446 Alan Hayward <alan.hayward@arm.com>
7447 David Sherwood <david.sherwood@arm.com>
7449 * calls.c (load_register_parameters): Cope with polynomial
7450 mode sizes. Require a constant size for BLKmode parameters
7451 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
7452 forces a parameter to be padded at the lsb end in order to
7453 fill a complete number of words, require the parameter size
7454 to be ordered wrt UNITS_PER_WORD.
7456 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7457 Alan Hayward <alan.hayward@arm.com>
7458 David Sherwood <david.sherwood@arm.com>
7460 * reload1.c (spill_stack_slot_width): Change element type
7461 from unsigned int to poly_uint64_pod.
7462 (alter_reg): Treat mode sizes as polynomial.
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 * reload.c (complex_word_subreg_p): New function.
7469 (reload_inner_reg_of_subreg, push_reload): Use it.
7471 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7472 Alan Hayward <alan.hayward@arm.com>
7473 David Sherwood <david.sherwood@arm.com>
7475 * lra-constraints.c (process_alt_operands): Reject matched
7476 operands whose sizes aren't ordered.
7477 (match_reload): Refer to this check here.
7479 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7480 Alan Hayward <alan.hayward@arm.com>
7481 David Sherwood <david.sherwood@arm.com>
7483 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
7484 that the mode size is in the set {1, 2, 4, 8, 16}.
7486 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7487 Alan Hayward <alan.hayward@arm.com>
7488 David Sherwood <david.sherwood@arm.com>
7490 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
7491 Use plus_constant instead of gen_rtx_PLUS.
7493 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7494 Alan Hayward <alan.hayward@arm.com>
7495 David Sherwood <david.sherwood@arm.com>
7497 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
7498 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
7499 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
7500 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
7501 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
7502 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
7503 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
7504 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
7505 * config/i386/i386.c (ix86_push_rounding): ...this new function.
7506 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
7508 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
7509 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
7510 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
7511 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
7512 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
7513 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
7514 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
7515 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
7516 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
7517 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
7519 * expr.c (emit_move_resolve_push): Treat the input and result
7520 of PUSH_ROUNDING as a poly_int64.
7521 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
7522 (emit_push_insn): Likewise.
7523 * lra-eliminations.c (mark_not_eliminable): Likewise.
7524 * recog.c (push_operand): Likewise.
7525 * reload1.c (elimination_effects): Likewise.
7526 * rtlanal.c (nonzero_bits1): Likewise.
7527 * calls.c (store_one_arg): Likewise. Require the padding to be
7528 known at compile time.
7530 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7531 Alan Hayward <alan.hayward@arm.com>
7532 David Sherwood <david.sherwood@arm.com>
7534 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
7535 Use plus_constant instead of gen_rtx_PLUS.
7537 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7538 Alan Hayward <alan.hayward@arm.com>
7539 David Sherwood <david.sherwood@arm.com>
7541 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
7544 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7545 Alan Hayward <alan.hayward@arm.com>
7546 David Sherwood <david.sherwood@arm.com>
7548 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
7549 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
7550 via stack temporaries. Treat the mode size as polynomial too.
7552 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7553 Alan Hayward <alan.hayward@arm.com>
7554 David Sherwood <david.sherwood@arm.com>
7556 * expr.c (expand_expr_real_2): When handling conversions involving
7557 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
7558 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
7559 as a poly_uint64 too.
7561 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7562 Alan Hayward <alan.hayward@arm.com>
7563 David Sherwood <david.sherwood@arm.com>
7565 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
7567 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7568 Alan Hayward <alan.hayward@arm.com>
7569 David Sherwood <david.sherwood@arm.com>
7571 * combine.c (can_change_dest_mode): Handle polynomial
7572 REGMODE_NATURAL_SIZE.
7573 * expmed.c (store_bit_field_1): Likewise.
7574 * expr.c (store_constructor): Likewise.
7575 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
7576 and polynomial REGMODE_NATURAL_SIZE.
7577 (gen_lowpart_common): Likewise.
7578 * reginfo.c (record_subregs_of_mode): Likewise.
7579 * rtlanal.c (read_modify_subreg_p): Likewise.
7581 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7582 Alan Hayward <alan.hayward@arm.com>
7583 David Sherwood <david.sherwood@arm.com>
7585 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
7586 numbers of elements.
7588 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7589 Alan Hayward <alan.hayward@arm.com>
7590 David Sherwood <david.sherwood@arm.com>
7592 * match.pd: Cope with polynomial numbers of vector elements.
7594 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7595 Alan Hayward <alan.hayward@arm.com>
7596 David Sherwood <david.sherwood@arm.com>
7598 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
7599 in a POINTER_PLUS_EXPR.
7601 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7602 Alan Hayward <alan.hayward@arm.com>
7603 David Sherwood <david.sherwood@arm.com>
7605 * omp-simd-clone.c (simd_clone_subparts): New function.
7606 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
7607 (ipa_simd_modify_function_body): Likewise.
7609 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7610 Alan Hayward <alan.hayward@arm.com>
7611 David Sherwood <david.sherwood@arm.com>
7613 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
7614 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
7615 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
7616 (expand_vector_condition, vector_element): Likewise.
7617 (subparts_gt): New function.
7618 (get_compute_type): Use subparts_gt.
7619 (count_type_subparts): Delete.
7620 (expand_vector_operations_1): Use subparts_gt instead of
7621 count_type_subparts.
7623 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7624 Alan Hayward <alan.hayward@arm.com>
7625 David Sherwood <david.sherwood@arm.com>
7627 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
7628 (vect_compile_time_alias): ...this new function. Do the calculation
7629 on poly_ints rather than trees.
7630 (vect_prune_runtime_alias_test_list): Update call accordingly.
7632 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7633 Alan Hayward <alan.hayward@arm.com>
7634 David Sherwood <david.sherwood@arm.com>
7636 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
7638 (vect_schedule_slp_instance): Likewise.
7640 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7641 Alan Hayward <alan.hayward@arm.com>
7642 David Sherwood <david.sherwood@arm.com>
7644 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
7645 constant and extern definitions for variable-length vectors.
7646 (vect_get_constant_vectors): Note that the number of units
7647 is known to be constant.
7649 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7650 Alan Hayward <alan.hayward@arm.com>
7651 David Sherwood <david.sherwood@arm.com>
7653 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
7654 of units as polynomial. Choose between WIDE and NARROW based
7657 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7658 Alan Hayward <alan.hayward@arm.com>
7659 David Sherwood <david.sherwood@arm.com>
7661 * tree-vect-stmts.c (simd_clone_subparts): New function.
7662 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
7664 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7665 Alan Hayward <alan.hayward@arm.com>
7666 David Sherwood <david.sherwood@arm.com>
7668 * tree-vect-stmts.c (vectorizable_call): Treat the number of
7669 vectors as polynomial. Use build_index_vector for
7672 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7673 Alan Hayward <alan.hayward@arm.com>
7674 David Sherwood <david.sherwood@arm.com>
7676 * tree-vect-stmts.c (get_load_store_type): Treat the number of
7677 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
7678 for variable-length vectors.
7679 (vectorizable_mask_load_store): Treat the number of units as
7680 polynomial, asserting that it is constant if the condition has
7681 already been enforced.
7682 (vectorizable_store, vectorizable_load): Likewise.
7684 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7685 Alan Hayward <alan.hayward@arm.com>
7686 David Sherwood <david.sherwood@arm.com>
7688 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
7689 of units as polynomial. Punt if we can't tell at compile time
7690 which vector contains the final result.
7692 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7693 Alan Hayward <alan.hayward@arm.com>
7694 David Sherwood <david.sherwood@arm.com>
7696 * tree-vect-loop.c (vectorizable_induction): Treat the number
7697 of units as polynomial. Punt on SLP inductions. Use an integer
7698 VEC_SERIES_EXPR for variable-length integer reductions. Use a
7699 cast of such a series for variable-length floating-point
7702 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7703 Alan Hayward <alan.hayward@arm.com>
7704 David Sherwood <david.sherwood@arm.com>
7706 * tree.h (build_index_vector): Declare.
7707 * tree.c (build_index_vector): New function.
7708 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
7709 of units as polynomial, forcibly converting it to a constant if
7710 vectorizable_reduction has already enforced the condition.
7711 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
7712 to create a {1,2,3,...} vector.
7713 (vectorizable_reduction): Treat the number of units as polynomial.
7714 Choose vectype_in based on the largest scalar element size rather
7715 than the smallest number of units. Enforce the restrictions
7718 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7719 Alan Hayward <alan.hayward@arm.com>
7720 David Sherwood <david.sherwood@arm.com>
7722 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
7723 number of units as polynomial.
7725 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7726 Alan Hayward <alan.hayward@arm.com>
7727 David Sherwood <david.sherwood@arm.com>
7729 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
7730 * target.def (autovectorize_vector_sizes): Return the vector sizes
7731 by pointer, using vector_sizes rather than a bitmask.
7732 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
7733 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
7734 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
7736 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
7737 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
7738 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
7739 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
7740 * omp-general.c (omp_max_vf): Likewise.
7741 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7742 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
7743 * tree-vect-loop.c (vect_analyze_loop): Likewise.
7744 * tree-vect-slp.c (vect_slp_bb): Likewise.
7745 * doc/tm.texi: Regenerate.
7746 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
7748 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
7749 the vector size as a poly_uint64 rather than an unsigned int.
7750 (current_vector_size): Change from an unsigned int to a poly_uint64.
7751 (get_vectype_for_scalar_type): Update accordingly.
7752 * tree.h (build_truth_vector_type): Take the size and number of
7753 units as a poly_uint64 rather than an unsigned int.
7754 (build_vector_type): Add a temporary overload that takes
7755 the number of units as a poly_uint64 rather than an unsigned int.
7756 * tree.c (make_vector_type): Likewise.
7757 (build_truth_vector_type): Take the number of units as a poly_uint64
7758 rather than an unsigned int.
7760 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7761 Alan Hayward <alan.hayward@arm.com>
7762 David Sherwood <david.sherwood@arm.com>
7764 * target.def (get_mask_mode): Take the number of units and length
7765 as poly_uint64s rather than unsigned ints.
7766 * targhooks.h (default_get_mask_mode): Update accordingly.
7767 * targhooks.c (default_get_mask_mode): Likewise.
7768 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
7769 * doc/tm.texi: Regenerate.
7771 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7772 Alan Hayward <alan.hayward@arm.com>
7773 David Sherwood <david.sherwood@arm.com>
7775 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
7776 * omp-general.c (omp_max_vf): Likewise.
7777 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
7778 (expand_omp_simd): Handle polynomial safelen.
7779 * omp-low.c (omplow_simd_context): Add a default constructor.
7780 (omplow_simd_context::max_vf): Change from int to poly_uint64.
7781 (lower_rec_simd_input_clauses): Update accordingly.
7782 (lower_rec_input_clauses): Likewise.
7784 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7785 Alan Hayward <alan.hayward@arm.com>
7786 David Sherwood <david.sherwood@arm.com>
7788 * tree-vectorizer.h (vect_nunits_for_cost): New function.
7789 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
7790 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
7791 (vect_analyze_slp_cost): Likewise.
7792 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
7793 (vect_model_load_cost): Likewise.
7795 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7796 Alan Hayward <alan.hayward@arm.com>
7797 David Sherwood <david.sherwood@arm.com>
7799 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
7800 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
7801 from an unsigned int * to a poly_uint64_pod *.
7802 (calculate_unrolling_factor): New function.
7803 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
7805 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7806 Alan Hayward <alan.hayward@arm.com>
7807 David Sherwood <david.sherwood@arm.com>
7809 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
7810 from an unsigned int to a poly_uint64.
7811 (_loop_vec_info::slp_unrolling_factor): Likewise.
7812 (_loop_vec_info::vectorization_factor): Change from an int
7814 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
7815 (vect_get_num_vectors): New function.
7816 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
7817 (vect_get_num_copies): Use vect_get_num_vectors.
7818 (vect_analyze_data_ref_dependences): Change max_vf from an int *
7819 to an unsigned int *.
7820 (vect_analyze_data_refs): Change min_vf from an int * to a
7822 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7823 than an unsigned HOST_WIDE_INT.
7824 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
7825 (vect_analyze_data_ref_dependence): Change max_vf from an int *
7826 to an unsigned int *.
7827 (vect_analyze_data_ref_dependences): Likewise.
7828 (vect_compute_data_ref_alignment): Handle polynomial vf.
7829 (vect_enhance_data_refs_alignment): Likewise.
7830 (vect_prune_runtime_alias_test_list): Likewise.
7831 (vect_shift_permute_load_chain): Likewise.
7832 (vect_supportable_dr_alignment): Likewise.
7833 (dependence_distance_ge_vf): Take the vectorization factor as a
7834 poly_uint64 rather than an unsigned HOST_WIDE_INT.
7835 (vect_analyze_data_refs): Change min_vf from an int * to a
7837 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
7838 vfm1 as a poly_uint64 rather than an int. Make the same change
7839 for the returned bound_scalar.
7840 (vect_gen_vector_loop_niters): Handle polynomial vf.
7841 (vect_do_peeling): Likewise. Update call to
7842 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
7843 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
7845 * tree-vect-loop.c (vect_determine_vectorization_factor)
7846 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
7847 (vect_get_known_peeling_cost): Likewise.
7848 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
7849 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
7850 (vect_transform_loop): Likewise. Use the lowest possible VF when
7851 updating the upper bounds of the loop.
7852 (vect_min_worthwhile_factor): Make static. Return an unsigned int
7854 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
7855 polynomial unroll factors.
7856 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
7857 (vect_make_slp_decision): Likewise.
7858 (vect_supported_load_permutation_p): Likewise, and polynomial
7860 (vect_analyze_slp_cost): Handle polynomial vf.
7861 (vect_slp_analyze_node_operations): Likewise.
7862 (vect_slp_analyze_bb_1): Likewise.
7863 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7864 than an unsigned HOST_WIDE_INT.
7865 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
7866 (vectorizable_load): Handle polynomial vf.
7867 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
7869 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
7871 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7872 Alan Hayward <alan.hayward@arm.com>
7873 David Sherwood <david.sherwood@arm.com>
7875 * match.pd: Handle bit operations involving three constants
7876 and try to fold one pair.
7878 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7880 * tree-vect-loop-manip.c: Include gimple-fold.h.
7881 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
7882 niters_maybe_zero parameters. Handle other cases besides a step of 1.
7883 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
7884 Add a path that uses a step of VF instead of 1, but disable it
7886 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
7887 and niters_no_overflow parameters. Update calls to
7888 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
7889 Create a new SSA name if the latter choses to use a ste other
7890 than zero, and return it via niters_vector_mult_vf_var.
7891 * tree-vect-loop.c (vect_transform_loop): Update calls to
7892 vect_do_peeling, vect_gen_vector_loop_niters and
7893 slpeel_make_loop_iterate_ntimes.
7894 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
7895 (vect_gen_vector_loop_niters): Update declarations after above changes.
7897 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
7899 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
7900 128-bit round to integer instructions.
7901 (ceil<mode>2): Likewise.
7902 (btrunc<mode>2): Likewise.
7903 (round<mode>2): Likewise.
7905 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7907 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
7908 unaligned VSX load/store on P8/P9.
7909 (expand_block_clear): Allow the use of unaligned VSX
7910 load/store on P8/P9.
7912 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7914 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
7916 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
7917 swap associated with both a load and a store.
7919 2018-01-02 Andrew Waterman <andrew@sifive.com>
7921 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
7922 * config/riscv/riscv.md (clear_cache): Use it.
7924 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
7926 * web.c: Remove out-of-date comment.
7928 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7930 * expr.c (fixup_args_size_notes): Check that any existing
7931 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
7932 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
7933 (emit_single_push_insn): ...here.
7935 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7937 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
7938 (const_vector_encoded_nelts): New function.
7939 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
7940 (const_vector_int_elt, const_vector_elt): Declare.
7941 * emit-rtl.c (const_vector_int_elt_1): New function.
7942 (const_vector_elt): Likewise.
7943 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
7944 of CONST_VECTOR_ELT.
7946 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7948 * expr.c: Include rtx-vector-builder.h.
7949 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
7950 directly on the tree encoding.
7951 (const_vector_from_tree): Likewise.
7952 * optabs.c: Include rtx-vector-builder.h.
7953 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
7954 sequence of "u" values.
7955 * vec-perm-indices.c: Include rtx-vector-builder.h.
7956 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
7957 directly on the vec_perm_indices encoding.
7959 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7961 * doc/rtl.texi (const_vector): Describe new encoding scheme.
7962 * Makefile.in (OBJS): Add rtx-vector-builder.o.
7963 * rtx-vector-builder.h: New file.
7964 * rtx-vector-builder.c: Likewise.
7965 * rtl.h (rtx_def::u2): Add a const_vector field.
7966 (CONST_VECTOR_NPATTERNS): New macro.
7967 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
7968 (CONST_VECTOR_DUPLICATE_P): Likewise.
7969 (CONST_VECTOR_STEPPED_P): Likewise.
7970 (CONST_VECTOR_ENCODED_ELT): Likewise.
7971 (const_vec_duplicate_p): Check for a duplicated vector encoding.
7972 (unwrap_const_vec_duplicate): Likewise.
7973 (const_vec_series_p): Check for a non-duplicated vector encoding.
7974 Say that the function only returns true for integer vectors.
7975 * emit-rtl.c: Include rtx-vector-builder.h.
7976 (gen_const_vec_duplicate_1): Delete.
7977 (gen_const_vector): Call gen_const_vec_duplicate instead of
7978 gen_const_vec_duplicate_1.
7979 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
7980 (gen_const_vec_duplicate): Use rtx_vector_builder.
7981 (gen_const_vec_series): Likewise.
7982 (gen_rtx_CONST_VECTOR): Likewise.
7983 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
7984 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7985 Build a new vector rather than modifying a CONST_VECTOR in-place.
7986 (handle_special_swappables): Update call accordingly.
7987 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
7988 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7989 Build a new vector rather than modifying a CONST_VECTOR in-place.
7990 (handle_special_swappables): Update call accordingly.
7992 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7994 * simplify-rtx.c (simplify_const_binary_operation): Use
7995 CONST_VECTOR_ELT instead of XVECEXP.
7997 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7999 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
8000 the selector elements to be different from the data elements
8001 if the selector is a VECTOR_CST.
8002 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
8003 ssizetype for the selector.
8005 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8007 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
8008 before testing each element individually.
8009 * tree-vect-generic.c (lower_vec_perm): Likewise.
8011 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8013 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
8014 * selftest-run-tests.c (selftest::run_tests): Call it.
8015 * vector-builder.h (vector_builder::operator ==): New function.
8016 (vector_builder::operator !=): Likewise.
8017 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
8018 (vec_perm_indices::all_from_input_p): New function.
8019 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8020 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
8021 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
8022 instead of reading the VECTOR_CST directly. Detect whether both
8023 vector inputs are the same before constructing the vec_perm_indices,
8024 and update the number of inputs argument accordingly. Use the
8025 utility functions added above. Only construct sel2 if we need to.
8027 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8029 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
8030 the broadcast of the low byte.
8031 (expand_mult_highpart): Use an explicit encoding for the permutes.
8032 * optabs-query.c (can_mult_highpart_p): Likewise.
8033 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
8034 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8035 (vectorizable_bswap): Likewise.
8036 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
8037 explicit encoding for the power-of-2 permutes.
8038 (vect_permute_store_chain): Likewise.
8039 (vect_grouped_load_supported): Likewise.
8040 (vect_permute_load_chain): Likewise.
8042 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8044 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
8045 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
8046 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
8047 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8048 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
8049 (vect_gen_perm_mask_any): Likewise.
8051 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8053 * int-vector-builder.h: New file.
8054 * vec-perm-indices.h: Include int-vector-builder.h.
8055 (vec_perm_indices): Redefine as an int_vector_builder.
8056 (auto_vec_perm_indices): Delete.
8057 (vec_perm_builder): Redefine as a stand-alone class.
8058 (vec_perm_indices::vec_perm_indices): New function.
8059 (vec_perm_indices::clamp): Likewise.
8060 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
8061 (vec_perm_indices::new_vector): New function.
8062 (vec_perm_indices::new_expanded_vector): Update for new
8063 vec_perm_indices class.
8064 (vec_perm_indices::rotate_inputs): New function.
8065 (vec_perm_indices::all_in_range_p): Operate directly on the
8066 encoded form, without computing elided elements.
8067 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
8068 encoding. Update for new vec_perm_indices class.
8069 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
8070 the given vec_perm_builder.
8071 (expand_vec_perm_var): Update vec_perm_builder constructor.
8072 (expand_mult_highpart): Use vec_perm_builder instead of
8073 auto_vec_perm_indices.
8074 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
8075 vec_perm_indices instead of auto_vec_perm_indices. Use a single
8076 or double series encoding as appropriate.
8077 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
8078 vec_perm_indices instead of auto_vec_perm_indices.
8079 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8080 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8081 (vect_permute_store_chain): Likewise.
8082 (vect_grouped_load_supported): Likewise.
8083 (vect_permute_load_chain): Likewise.
8084 (vect_shift_permute_load_chain): Likewise.
8085 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8086 (vect_transform_slp_perm_load): Likewise.
8087 (vect_schedule_slp_instance): Likewise.
8088 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8089 (vectorizable_mask_load_store): Likewise.
8090 (vectorizable_bswap): Likewise.
8091 (vectorizable_store): Likewise.
8092 (vectorizable_load): Likewise.
8093 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
8094 vec_perm_indices instead of auto_vec_perm_indices. Use
8095 tree_to_vec_perm_builder to read the vector from a tree.
8096 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
8097 vec_perm_builder instead of a vec_perm_indices.
8098 (have_whole_vector_shift): Use vec_perm_builder and
8099 vec_perm_indices instead of auto_vec_perm_indices. Leave the
8100 truncation to calc_vec_perm_mask_for_shift.
8101 (vect_create_epilog_for_reduction): Likewise.
8102 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
8103 from auto_vec_perm_indices to vec_perm_indices.
8104 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8105 instead of changing individual elements.
8106 (aarch64_vectorize_vec_perm_const): Use new_vector to install
8107 the vector in d.perm.
8108 * config/arm/arm.c (expand_vec_perm_d::perm): Change
8109 from auto_vec_perm_indices to vec_perm_indices.
8110 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8111 instead of changing individual elements.
8112 (arm_vectorize_vec_perm_const): Use new_vector to install
8113 the vector in d.perm.
8114 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
8115 Update vec_perm_builder constructor.
8116 (rs6000_expand_interleave): Likewise.
8117 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
8118 (rs6000_expand_interleave): Likewise.
8120 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8122 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
8123 to qimode could truncate the indices.
8124 * optabs.c (expand_vec_perm_var): Likewise.
8126 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8128 * Makefile.in (OBJS): Add vec-perm-indices.o.
8129 * vec-perm-indices.h: New file.
8130 * vec-perm-indices.c: Likewise.
8131 * target.h (vec_perm_indices): Replace with a forward class
8133 (auto_vec_perm_indices): Move to vec-perm-indices.h.
8134 * optabs.h: Include vec-perm-indices.h.
8135 (expand_vec_perm): Delete.
8136 (selector_fits_mode_p, expand_vec_perm_var): Declare.
8137 (expand_vec_perm_const): Declare.
8138 * target.def (vec_perm_const_ok): Replace with...
8139 (vec_perm_const): ...this new hook.
8140 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
8141 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
8142 * doc/tm.texi: Regenerate.
8143 * optabs.def (vec_perm_const): Delete.
8144 * doc/md.texi (vec_perm_const): Likewise.
8145 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
8146 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
8147 expand_vec_perm for constant permutation vectors. Assert that
8148 the mode of variable permutation vectors is the integer equivalent
8149 of the mode that is being permuted.
8150 * optabs-query.h (selector_fits_mode_p): Declare.
8151 * optabs-query.c: Include vec-perm-indices.h.
8152 (selector_fits_mode_p): New function.
8153 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
8154 is defined, instead of checking whether the vec_perm_const_optab
8155 exists. Use targetm.vectorize.vec_perm_const instead of
8156 targetm.vectorize.vec_perm_const_ok. Check whether the indices
8157 fit in the vector mode before using a variable permute.
8158 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
8159 vec_perm_indices instead of an rtx.
8160 (expand_vec_perm): Replace with...
8161 (expand_vec_perm_const): ...this new function. Take the selector
8162 as a vec_perm_indices rather than an rtx. Also take the mode of
8163 the selector. Update call to shift_amt_for_vec_perm_mask.
8164 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
8165 Use vec_perm_indices::new_expanded_vector to expand the original
8166 selector into bytes. Check whether the indices fit in the vector
8167 mode before using a variable permute.
8168 (expand_vec_perm_var): Make global.
8169 (expand_mult_highpart): Use expand_vec_perm_const.
8170 * fold-const.c: Includes vec-perm-indices.h.
8171 * tree-ssa-forwprop.c: Likewise.
8172 * tree-vect-data-refs.c: Likewise.
8173 * tree-vect-generic.c: Likewise.
8174 * tree-vect-loop.c: Likewise.
8175 * tree-vect-slp.c: Likewise.
8176 * tree-vect-stmts.c: Likewise.
8177 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
8179 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
8180 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
8181 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
8182 (aarch64_vectorize_vec_perm_const): ...this new function.
8183 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8184 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8185 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
8186 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
8187 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8188 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8189 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
8191 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
8192 check for NEON modes.
8193 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
8194 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
8195 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
8196 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
8198 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
8199 the old VEC_PERM_CONST conditions.
8200 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
8201 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
8202 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
8203 (ia64_vectorize_vec_perm_const_ok): Merge into...
8204 (ia64_vectorize_vec_perm_const): ...this new function.
8205 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
8206 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
8207 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
8208 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
8209 * config/mips/mips.c (mips_expand_vec_perm_const)
8210 (mips_vectorize_vec_perm_const_ok): Merge into...
8211 (mips_vectorize_vec_perm_const): ...this new function.
8212 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
8213 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
8214 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
8215 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
8216 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
8217 (rs6000_expand_vec_perm_const): Delete.
8218 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
8220 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8221 (altivec_expand_vec_perm_const_le): Take each operand individually.
8222 Operate on constant selectors rather than rtxes.
8223 (altivec_expand_vec_perm_const): Likewise. Update call to
8224 altivec_expand_vec_perm_const_le.
8225 (rs6000_expand_vec_perm_const): Delete.
8226 (rs6000_vectorize_vec_perm_const_ok): Delete.
8227 (rs6000_vectorize_vec_perm_const): New function.
8228 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8229 an element count and rtx array.
8230 (rs6000_expand_extract_even): Update call accordingly.
8231 (rs6000_expand_interleave): Likewise.
8232 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
8233 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
8234 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
8235 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
8236 (rs6000_expand_vec_perm_const): Delete.
8237 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8238 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8239 (altivec_expand_vec_perm_const_le): Take each operand individually.
8240 Operate on constant selectors rather than rtxes.
8241 (altivec_expand_vec_perm_const): Likewise. Update call to
8242 altivec_expand_vec_perm_const_le.
8243 (rs6000_expand_vec_perm_const): Delete.
8244 (rs6000_vectorize_vec_perm_const_ok): Delete.
8245 (rs6000_vectorize_vec_perm_const): New function. Remove stray
8246 reference to the SPE evmerge intructions.
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/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
8252 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
8254 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8256 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8258 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
8259 vector mode and that that mode matches the mode of the data
8261 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
8262 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
8263 directly using expand_vec_perm_1 when forcing selectors into
8265 (expand_vec_perm_var): New function, split out from expand_vec_perm.
8267 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8269 * optabs-query.h (can_vec_perm_p): Delete.
8270 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
8271 * optabs-query.c (can_vec_perm_p): Split into...
8272 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
8273 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
8274 particular selector is valid.
8275 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8276 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8277 (vect_grouped_load_supported): Likewise.
8278 (vect_shift_permute_load_chain): Likewise.
8279 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8280 (vect_transform_slp_perm_load): Likewise.
8281 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8282 (vectorizable_bswap): Likewise.
8283 (vect_gen_perm_mask_checked): Likewise.
8284 * fold-const.c (fold_ternary_loc): Likewise. Don't take
8285 implementations of variable permutation vectors into account
8286 when deciding which selector to use.
8287 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
8288 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
8289 with a false third argument.
8290 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
8291 to test whether the constant selector is valid and can_vec_perm_var_p
8292 to test whether a variable selector is valid.
8294 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8296 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
8297 * optabs-query.c (can_vec_perm_p): Likewise.
8298 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
8299 instead of vec_perm_indices.
8300 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
8301 (vect_gen_perm_mask_checked): Likewise,
8302 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
8303 (vect_gen_perm_mask_checked): Likewise,
8305 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8307 * optabs-query.h (qimode_for_vec_perm): Declare.
8308 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
8309 (qimode_for_vec_perm): ...this new function.
8310 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
8312 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8314 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
8315 does not have a conditional at the top.
8317 2018-01-02 Richard Biener <rguenther@suse.de>
8319 * ipa-inline.c (big_speedup_p): Fix expression.
8321 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8324 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
8327 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8331 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
8332 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
8333 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
8334 cond_taken_branch_cost 3->4.
8336 2018-01-01 Jakub Jelinek <jakub@redhat.com>
8338 PR tree-optimization/83581
8339 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
8340 TODO_cleanup_cfg if any changes have been made.
8343 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
8344 convert_modes if target mode has the right side, but different mode
8348 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
8349 last argument when extracting from CONCAT. If either from_real or
8350 from_imag is NULL, use expansion through memory. If result is not
8351 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
8352 the parts directly to inner mode, if even that fails, use expansion
8356 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
8357 check for bswap in mode rather than HImode and use that in expand_unop
8360 Copyright (C) 2018 Free Software Foundation, Inc.
8362 Copying and distribution of this file, with or without modification,
8363 are permitted in any medium without royalty provided the copyright
8364 notice and this notice are preserved.