1 2018-03-08 Richard Biener <rguenther@suse.de>
3 PR tree-optimization/84746
4 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
5 (phi_translate): Pass in destination ANTIC_OUT set.
6 (phi_translate_1): Likewise. For a simplified result lookup
7 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
8 (phi_translate_set): Adjust.
9 (do_pre_regular_insertion): Likewise.
10 (do_pre_partial_partial_insertion): Likewise.
12 2018-03-08 Martin Liska <mliska@suse.cz>
15 * doc/gcov.texi: Document usage of profile files.
16 * gcov-io.h: Document changes in the format.
18 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
22 * dwarf2out.c (struct dw_line_info_table): Update comments for
24 (FORCE_RESET_NEXT_VIEW): New.
25 (FORCE_RESETTING_VIEW_P): New.
26 (RESETTING_VIEW_P): Check for -1 too.
27 (ZERO_VIEW_P): Likewise.
28 (new_line_info_table): Force-reset next view.
29 (dwarf2out_begin_function): Likewise.
30 (dwarf2out_source_line): Simplify zero_view_p initialization.
31 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
32 view directly. Omit view when omitting .loc at line 0.
34 2018-03-08 Jakub Jelinek <jakub@redhat.com>
36 PR tree-optimization/84740
37 * tree-switch-conversion.c (process_switch): Call build_constructors
38 only if info.phi_count is non-zero.
40 PR tree-optimization/84739
41 * tree-tailcall.c (find_tail_calls): Check call arguments against
42 DECL_ARGUMENTS (current_function_decl) rather than
43 DECL_ARGUMENTS (func) when checking for tail recursion.
45 2018-03-07 Jakub Jelinek <jakub@redhat.com>
47 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
48 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
49 Volker Reichelt's entry and add entries for people that perform
50 GCC fuzzy testing and report numerous bugs.
52 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
55 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
56 readonly data in sdata, if that is disabled.
57 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
58 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
59 -mreadonly-in-sdata option.
61 2018-03-07 Martin Sebor <msebor@redhat.com>
63 PR tree-optimization/84468
64 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
65 basic block when looking for nul assignment.
67 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
70 * except.h (output_function_exception_table): Adjust prototype.
71 * except.c (output_function_exception_table): Remove FNNAME parameter
72 and add SECTION parameter. Ouput one part of the table at a time.
73 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
74 the first part of the exception table and emit unwind directives.
75 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
76 (i386_pe_seh_cold_init): Likewise.
77 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
78 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
79 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
80 (ix86_output_call_insn): Emit a nop in one more case for SEH.
81 * config/i386/winnt.c: Include except.h.
82 (struct seh_frame_state): Add reg_offset, after_prologue and
83 in_cold_section fields.
84 (i386_pe_seh_end_prologue): Set seh->after_prologue.
85 (i386_pe_seh_cold_init): New function.
86 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
87 to seh->in_cold_section.
88 (seh_emit_push): Record the offset of the push.
89 (seh_emit_save): Record the offet of the save.
90 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
91 Test seh->after_prologue to disregard the epilogue.
92 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
93 (i386_pe_end_cold_function): New function.
95 2018-03-07 Jakub Jelinek <jakub@redhat.com>
98 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
99 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
102 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
104 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
105 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
108 * multiple_target.c: Include tree-inline.h and intl.h.
109 (expand_target_clones): Diagnose and fail if node->definition and
110 !tree_versionable_function_p (node->decl).
112 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
114 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
116 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
117 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
118 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
120 2018-03-06 Jakub Jelinek <jakub@redhat.com>
123 * combine.c (try_combine): Use reg_or_subregno instead of handling
124 just paradoxical SUBREGs and REGs.
126 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
128 * config/arc/arc.c (arc_finalize_pic): Remove function.
129 (arc_must_save_register): We use single base PIC register, remove
130 checks to save/restore the PIC register.
131 (arc_expand_prologue): Likewise.
132 * config/arc/arc-protos.h (arc_set_default_type_attributes):
134 (arc_verify_short): Likewise.
135 (arc_attr_type): Likewise.
136 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
137 (walk_stores): Likewise.
138 (arc_address_cost): Make it static.
139 (arc_verify_short): Likewise.
140 (branch_dest): Likewise.
141 (arc_attr_type): Likewise.
142 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
143 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
144 (arc_final_prescan_insn): Remove inserting the nops due to
145 hardware hazards. It is done in reorg step.
146 (insn_length_variant_t): Remove.
147 (insn_length_parameters_t): Likewise.
148 (arc_insn_length_parameters): Likewise.
149 (arc_get_insn_variants): Likewise.
150 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
152 2018-03-06 Jakub Jelinek <jakub@redhat.com>
155 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
158 PR tree-optimization/84687
159 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
161 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
163 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
165 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
166 Rename to ppc_speculation_barrier.
167 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
168 __builtin_ppc_speculation_barrier.
170 2018-03-05 Jakub Jelinek <jakub@redhat.com>
173 * combine.c (combine_simplify_rtx): Don't try to simplify if
174 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
177 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
179 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
180 to 32 bytes when compiling for POWER9.
182 2018-03-05 Jakub Jelinek <jakub@redhat.com>
185 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
186 regparm >= 3 with no arg reg available also for calls with
187 flag_force_indirect_call. Pass decl to ix86_function_regparm.
190 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
192 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
194 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
197 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
199 2018-03-05 Richard Biener <rguenther@suse.de>
201 PR tree-optimization/84486
202 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
203 When inserting a __builtin_assume_aligned call set the LHS
204 SSA name alignment info accordingly.
206 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
208 PR tree-optimization/84114
209 * config/aarch64/aarch64.c (aarch64_reassociation_width)
210 Avoid reassociation of FLOAT_MODE addition.
212 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
214 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
215 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
216 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
217 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
218 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
219 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
220 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
222 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
223 __builtin_ia32_wbinvd): New builtins.
224 (SPECIAL_ARGS2): New.
225 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
226 (SPECIAL_ARGS2): New.
227 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
228 (ix86_valid_target_attribute_inner_p): Ditto.
229 (ix86_init_mmx_sse_builtins): Add special_args2.
230 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
231 TARGET_WBNOINVD_P): New.
232 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
233 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
234 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
235 * config/i386/immintrin.h (_wbinvd): New intrinsic.
236 * config/i386/pconfigintrin.h: New file.
237 * config/i386/wbnoinvdintrin.h: Ditto.
238 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
239 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
241 2018-03-05 Richard Biener <rguenther@suse.de>
243 PR tree-optimization/84670
244 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
246 (BB_VISITED_WITH_VISITED_SUCCS): New define.
247 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
248 (compute_antic_aux): Only assert the number of values in ANTIC_IN
249 doesn't grow if all successors (recursively) were visited at least
252 2018-03-05 Richard Biener <rguenther@suse.de>
254 PR tree-optimization/84650
255 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
256 if executed in the loop pipeline.
258 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
260 * doc/configfiles.texi (Configuration Files): Move info about
261 conditionalizing $target-protos.h to...
262 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
263 differs from $target-protos.h.
265 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
266 Chung-Ju Wu <jasonwucj@gmail.com>
268 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
269 * config/nds32/nds32-multiple.md (setmemsi): Define.
270 * config/nds32/nds32-memory-manipulation.c
271 (nds32_gen_dup_4_byte_to_word_value): New.
272 (emit_setmem_word_loop): New.
273 (emit_setmem_byte_loop): New.
274 (nds32_expand_setmem_loop): New.
275 (nds32_expand_setmem_loop_v3m): New.
276 (nds32_expand_setmem_unroll): New.
277 (nds32_expand_setmem): New.
279 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
280 Chung-Ju Wu <jasonwucj@gmail.com>
282 * config/nds32/nds32-memory-manipulation.c
283 (nds32_emit_load_store): New.
284 (nds32_emit_post_inc_load_store): New.
285 (nds32_emit_mem_move): New.
286 (nds32_emit_mem_move_block): New.
287 (nds32_expand_movmemsi_loop_unknown_size): New.
288 (nds32_expand_movmemsi_loop_known_size): New.
289 (nds32_expand_movmemsi_loop): New.
290 (nds32_expand_movmemsi_unroll): New.
291 (nds32_expand_movmemqi): Rename ...
292 (nds32_expand_movmemsi): ... to this.
293 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
294 (movmemsi): ... to this.
295 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
296 (nds32_expand_movmemsi): ... to this.
298 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
299 Monk Chiang <sh.chiang04@gmail.com>
300 Chung-Ju Wu <jasonwucj@gmail.com>
302 * config/nds32/nds32-protos.h
303 (nds32_expand_load_multiple): New arguments.
304 (nds32_expand_store_multiple): Ditto.
305 (nds32_valid_multiple_load_store): Rename ...
306 (nds32_valid_multiple_load_store_p): ... to this.
307 * config/nds32/nds32-memory-manipulation.c
308 (nds32_expand_load_multiple): Refine implementation.
309 (nds32_expand_store_multiple): Ditto.
310 * config/nds32/nds32-multiple.md
311 (load_multiple): Update nds32_expand_load_multiple interface.
312 (store_multiple): Update nds32_expand_store_multiple interface.
313 * config/nds32/nds32-predicates.c
314 (nds32_valid_multiple_load_store): Rename ...
315 (nds32_valid_multiple_load_store_p): ... to this and refine
317 * config/nds32/predicates.md
318 (nds32_load_multiple_and_update_address_operation): New predicate.
319 (nds32_store_multiple_and_update_address_operation): New predicate.
321 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
322 Chung-Ju Wu <jasonwucj@gmail.com>
324 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
325 (combo): New attribute.
326 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
328 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
330 * config/nds32/nds32.opt: Change -mcmodel= default value.
332 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
333 Monk Chiang <sh.chiang04@gmail.com>
334 Chung-Ju Wu <jasonwucj@gmail.com>
336 * config/nds32/constants.md (unspec_element): New enum.
337 * config/nds32/constraints.md (Umw): New constraint.
338 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
339 * config/nds32/nds32-intrinsic.md: Likewise.
340 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
341 (nds32_valid_smw_lwm_base_p): New.
342 (nds32_output_smw_single_word): New.
343 (nds32_output_lmw_single_word): New.
344 (nds32_expand_unaligned_load): New.
345 (nds32_expand_unaligned_store): New.
346 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
347 (nds32_output_smw_single_word): Declare.
348 (nds32_output_lmw_single_word): Declare.
349 (nds32_expand_unaligned_load): Declare.
350 (nds32_expand_unaligned_store): Declare.
351 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
352 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
353 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
354 NDS32_BUILTIN_UASTORE_DW.
355 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
358 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
359 Kito Cheng <kito.cheng@gmail.com>
360 Chung-Ju Wu <jasonwucj@gmail.com>
362 * config/nds32/nds32-intrinsic.c
363 (nds32_expand_builtin_null_ftype_reg): Delete.
364 (nds32_expand_builtin_reg_ftype_imm): Ditto.
365 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
366 (nds32_read_argument): New.
367 (nds32_legitimize_target): Ditto.
368 (nds32_legitimize_argument): Ditto.
369 (nds32_check_constant_argument): Ditto.
370 (nds32_expand_unop_builtin): Ditto.
371 (nds32_expand_unopimm_builtin): Ditto.
372 (nds32_expand_binop_builtin): Ditto.
373 (nds32_builtin_decl_impl): Ditto.
374 (builtin_description): Ditto.
375 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
376 (nds32_init_builtins_impl): Ditto.
377 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
378 (nds32_builtin_decl): New.
379 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
380 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
382 2018-03-02 Jeff Law <law@redhat.com>
384 * reorg.c (stop_search_p): Handle DEBUG_INSN.
385 (redundant_insn, fill_simple_delay_slots): Likewise.
386 (fill_slots_from_thread): Likewise.
387 * resource.c (mark_referenced_resources): Likewise.
388 (mark_set_resources, find_dead_or_set_registers): Likewise.
390 2018-03-02 Jakub Jelinek <jakub@redhat.com>
392 * substring-locations.h (format_warning_va): Formatting fix for
394 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
396 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
397 * substring-locations.c: Include intl.h.
398 (format_warning_va): Turned into small wrapper around
399 format_warning_n_va, renamed to ...
400 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
401 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
403 (format_warning_at_substring_n): New function.
404 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
405 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
406 format_warning_at_substring with just a shorter name instead of
407 const function pointer.
408 (fmtwarn_n): New function.
409 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
410 appropriate, get rid of all the fmtstr temporaries, move conditionals
411 with G_() wrapped string literals directly into fmtwarn arguments,
412 cast dir.len to (int), formatting fixes.
414 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
416 * doc/invoke.texi: Remove "Cilk Plus" references.
418 2018-03-02 Jakub Jelinek <jakub@redhat.com>
419 Richard Biener <rguenther@suse.de>
422 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
423 for error or warning attributes if CALL_FROM_THUNK_P is set.
426 2018-03-02 Jakub Jelinek <jakub@redhat.com>
429 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
430 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
433 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
434 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
436 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
437 instead of -1U in last predictors element's probability member.
439 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
442 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
443 arguments if they are comparable.
445 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
447 PR tree-optimization/84634
448 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
449 masks and masked_loop_p with a single loop_masks, making sure it's
450 null for bb vectorization.
452 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
454 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
455 (vect_analyze_data_ref_access): Use loop->safe_len rather than
456 loop->force_vectorize to check whether there is no alias.
458 2018-03-02 Jakub Jelinek <jakub@redhat.com>
461 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
463 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
465 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
466 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
467 instead of a loop around prev_real_insn.
468 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
472 * config/i386/i386.c (ix86_print_operand): Use conditional
473 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
476 2018-03-02 Richard Biener <rguenther@suse.de>
478 PR tree-optimization/84427
479 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
480 (bitmap_set_subtract_values): Rewrite to handle multiple
483 (prune_clobbered_mems): Likewise.
484 (phi_translate): Take edge instead of pred/phiblock.
485 (phi_translate_1): Likewise.
486 (phi_translate_set): Likewise. Insert all translated
487 exprs for a value into the set, keeping possibly multiple
488 expressions per value.
489 (compute_antic_aux): Adjust for phi_translate changes.
490 When intersecting union the expressions and prune those
491 not in the final value set, keeping possibly multiple
492 expressions per value. Do not use value-insertion
493 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
494 all expressions. Add verification that the value-sets
495 only shrink during iteration.
496 (compute_partial_antic_aux): Adjust for the phi_translate changes.
497 (do_pre_regular_insertion): Likewise.
498 (do_pre_partial_partial_insertion): Likewise.
500 2018-03-02 Richard Biener <rguenther@suse.de>
503 * config/darwin.c (saved_debug_info_level): New static global.
504 (darwin_asm_lto_start): Disable debug info generation for LTO out.
505 (darwin_asm_lto_end): Restore debug info generation settings.
507 2018-03-01 Martin Liska <mliska@suse.cz>
510 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
513 2018-03-01 Richard Biener <rguenther@suse.de>
516 * dwarf2out.c (gen_variable_die): Properly handle late VLA
517 type annotation with LTO when debug was disabled at compile-time.
519 2018-03-01 Matthew Fortune <mfortune@gmail.com>
521 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
523 (mips_final_postscan_insn): Likewise.
525 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
527 PR rtl-optimization/84528
528 * alias.c (init_alias_target): Add commentary.
529 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
530 a unique base value if the frame pointer is not eliminated
531 to the stack pointer.
533 2018-03-01 Tom de Vries <tom@codesourcery.com>
535 PR rtl-optimization/83327
536 * lra-int.h (hard_regs_spilled_into): Declare.
537 * lra.c (hard_regs_spilled_into): Define.
538 (init_reg_info): Init hard_regs_spilled_into.
539 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
540 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
541 (process_bb_lives): Handle hard_regs_spilled_into.
542 (lra_create_live_ranges_1): Before doing liveness propagation, clear
543 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
545 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
547 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
548 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
549 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
550 * config/rs6000/aix72.h: New file.
552 2018-02-28 Jakub Jelinek <jakub@redhat.com>
554 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
555 instead of warning_at with conditional singular and plural messages
559 * stor-layout.c (update_alignment_for_field): For
560 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
561 && !DECL_PACKED (field), do the alignment update, just use
562 only desired_align instead of MAX (type_align, desired_align)
564 (place_field): Don't do known_align < desired_align handling
565 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
566 is non-NULL, instead do it after rli->prev_field handling and
567 only if not within a bitfield word. For DECL_PACKED (field)
568 use type_align of BITS_PER_UNIT.
570 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
572 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
573 superfluous parentheses and trailing spaces.
575 2018-02-28 Richard Biener <rguenther@suse.de>
577 PR tree-optimization/84584
578 * graphite-scop-detection.c (scop_detection::add_scop): Discard
579 SCoPs with fake exit edge.
581 2018-02-28 Martin Liska <mliska@suse.cz>
584 * timevar.c (timer::print): Fix format to properly print 100%
587 2018-02-28 Richard Biener <rguenther@suse.de>
590 * genmatch.c (capture_info::walk_match): Do not mark
591 captured expressions without operands as expr_p given
592 they act more like predicates and should be subject to
593 "lost tail" side-effect preserving.
595 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
597 PR rtl-optimization/81611
598 * auto-inc-dec.c (attempt_change): Move dead note from
599 mem_insn if it's the next use of regno
600 (find_address): Take address use of reg holding
601 non-incremented value. Add parm to limit search to the named
603 (merge_in_block): Attempt to use a mem insn that is the next
604 use of the original regno.
606 2018-02-27 Martin Sebor <msebor@redhat.com>
609 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
610 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
612 2018-02-27 Martin Sebor <msebor@redhat.com>
615 * diagnostic-core.h (warning_n, error_n, inform_n): Change
616 n argument to unsigned HOST_WIDE_INT.
617 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
618 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
619 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
620 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
622 2018-02-27 Richard Biener <rguenther@suse.de>
624 PR tree-optimization/84512
625 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
626 Do not use the estimate returned from record_stmt_cost for
627 the scalar iteration cost but sum properly using add_stmt_cost.
629 2018-02-27 Richard Biener <rguenther@suse.de>
631 PR tree-optimization/84466
632 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
633 Adjust last change to less strictly validate use operands.
635 2018-02-27 Martin Liska <mliska@suse.cz>
637 PR gcov-profile/84548
638 * gcov.c (process_file): Allow partial overlap and consider it
639 also as group functions.
640 (output_lines): Properly calculate range of lines for a group.
642 2018-02-27 Martin Liska <mliska@suse.cz>
644 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
645 'ggc' suffixes. Change first column width.
646 (timer::print): Fix formatting of the column.
648 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
650 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
651 preserve inline entry blocks for the sake of debug inline
652 entry point markers alone.
653 (remove_unused_locals): Suggest in comments a better place to
654 force the preservation of inline entry blocks that are
655 otherwise unused, but do not preserve them.
657 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
659 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
661 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
664 * config/i386/constraints.md (Bs): Replace
665 ix86_indirect_branch_register with
666 TARGET_INDIRECT_BRANCH_REGISTER.
668 * config/i386/i386.md (indirect_jump): Likewise.
669 (tablejump): Likewise.
670 (*sibcall_memory): Likewise.
671 (*sibcall_value_memory): Likewise.
672 Peepholes of indirect call and jump via memory: Likewise.
673 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
674 (*sibcall_value_GOT_32): Likewise.
675 * config/i386/predicates.md (indirect_branch_operand): Likewise.
676 (GOT_memory_operand): Likewise.
677 (call_insn_operand): Likewise.
678 (sibcall_insn_operand): Likewise.
679 (GOT32_symbol_operand): Likewise.
680 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
682 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
684 PR rtl-optimization/83496
685 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
686 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
687 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
688 redundant insn, if any.
689 (relax_delay_slots): Likewise.
690 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
692 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
694 PR tree-optimization/83965
695 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
696 that grouped statements are part of a reduction chain. Return
697 true if the statement is not marked as a reduction itself but
699 (vect_recog_dot_prod_pattern): Don't check whether the statement
700 is part of a group here.
701 (vect_recog_sad_pattern): Likewise.
702 (vect_recog_widen_sum_pattern): Likewise.
704 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
707 * final.c (rest_of_clean_state): Also look for calls inside sequences.
709 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
712 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
714 (ix86_output_indirect_function_return): New prototype.
715 (ix86_split_simple_return_pop_internal): Likewise.
716 * config/i386/i386.c (indirect_return_via_cx): New.
717 (indirect_return_via_cx_bnd): Likewise.
718 (indirect_thunk_name): Handle return va CX_REG.
719 (output_indirect_thunk_function): Create alias for
720 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
721 (ix86_output_indirect_jmp): Remove the bool argument.
722 (ix86_output_indirect_function_return): New function.
723 (ix86_split_simple_return_pop_internal): Likewise.
724 * config/i386/i386.md (*indirect_jump): Don't pass false
725 to ix86_output_indirect_jmp.
726 (*tablejump_1): Likewise.
727 (simple_return_pop_internal): Change it to define_insn_and_split.
728 Call ix86_split_simple_return_pop_internal to split it for
730 (simple_return_indirect_internal): Call
731 ix86_output_indirect_function_return instead of
732 ix86_output_indirect_jmp.
734 2018-02-26 Jakub Jelinek <jakub@redhat.com>
737 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
738 memset and value initialization afterwards.
740 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
742 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
744 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
747 * common/config/aarch64/aarch64-common.c
748 (aarch_option_optimization_table[]): Switch
749 off fomit-frame-pointer
751 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
752 Chung-Ju Wu <jasonwucj@gmail.com>
754 * config/nds32/nds32-multiple.md (load_multiple): Disallow
756 (store_multiple): Ditto.
758 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
760 * config.gcc: Add --with-cpu support for nds32 target.
761 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
762 * config/nds32/nds32.opt: Add -mcpu= option.
764 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
766 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
767 isel=yes): Warn for these deprecated options.
769 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
771 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
772 ISA_2_5_MASKS_EMBEDDED.
774 2018-02-23 Jakub Jelinek <jakub@redhat.com>
776 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
777 p->max as pointers rather than using iterative_hash_expr.
779 2018-02-23 Carl Love <cel@us.ibm.com>
781 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
782 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
784 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
785 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
786 P8V_BUILTIN_VEC_VUNSIGNED2.
788 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
791 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
792 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
794 * lra-constraints.c (curr_insn_transform): Set up
795 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
797 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
798 * lra-eliminations.c (spill_pseudos): Ditto.
799 (process_insn_for_elimination): Ditto.
800 * lra-lives.c (reg_early_clobber_p): Use the new macros.
801 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
802 LRA_NON_CLOBBERED_ALT.
804 2018-02-22 Martin Sebor <msebor@redhat.com>
806 PR tree-optimization/84480
807 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
808 to maybe_diag_stxncpy_trunc. Call it.
809 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
810 from gimple_fold_builtin_strcpy. Print inlining stack.
811 (handle_builtin_stxncpy): Print inlining stack.
812 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
814 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
817 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
818 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
819 and -fcheck-pointer-bounds are used together.
820 (indirect_thunk_prefix): New enum.
821 (indirect_thunk_need_prefix): New function.
822 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
823 "_nt" instead of "_bnd" for NOTRACK prefix.
824 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
825 (output_indirect_thunk_function): Likewise.
827 (ix86_code_end): Update output_indirect_thunk_function calls.
828 (ix86_output_indirect_branch_via_reg): Replace
829 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
830 (ix86_output_indirect_branch_via_push): Likewise.
831 (ix86_output_function_return): Likewise.
832 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
833 incompatible with -fcf-protection=branch and
834 -fcheck-pointer-bounds.
836 2018-02-22 Steve Ellcey <sellcey@cavium.com>
839 * config/aarch64/aarch64.c (aarch64_print_address_internal):
840 Change gcc_assert call to output_operand_lossage.
842 2018-02-22 Steve Ellcey <sellcey@cavium.com>
844 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
846 2018-02-22 DJ Delorie <dj@redhat.com>
847 Sebastian Perta <sebastian.perta@renesas.com>
848 Oleg Endo <olegendo@gcc.gnu.org>
850 * config/rx/rx.c (rx_rtx_costs): New function.
851 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
853 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
855 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
857 2018-02-22 Martin Liska <mliska@suse.cz>
860 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
861 Add "native" as a possible value.
863 2018-02-22 Martin Liska <mliska@suse.cz>
866 * config/i386/i386.c (ix86_option_override_internal):
867 Add "native" as a possible value for -march and -mtune.
869 2018-02-22 Jakub Jelinek <jakub@redhat.com>
872 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
873 to all type variants.
875 PR tree-optimization/84503
876 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
877 width as info->bitpos + info->bitsize - start.
878 (merged_store_group::merge_overlapping): Simplify width computation.
879 (check_no_overlap): New function.
880 (imm_store_chain_info::try_coalesce_bswap): Compute expected
881 start + width and last_order of the group, fail if check_no_overlap
883 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
884 to group if check_no_overlap fails.
886 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
888 * config/rs6000/altivec.md: Delete contraint arguments to
889 define_expand, define_split, and define_peephole2, and in
890 define_insn_and_split if always unused.
891 * config/rs6000/darwin.md: Ditto.
892 * config/rs6000/dfp.md: Ditto.
893 * config/rs6000/rs6000.md: Ditto.
894 * config/rs6000/sync.md: Ditto.
895 * config/rs6000/vector.md: Ditto.
896 * config/rs6000/vsx.md: Ditto.
898 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
900 * config/rs6000/altivec.md: Write output control strings as braced
901 blocks instead of double-quoted strings.
902 * config/rs6000/darwin.md: Ditto.
903 * config/rs6000/rs6000.md: Ditto.
904 * config/rs6000/vector.md: Ditto.
905 * config/rs6000/vsx.md: Ditto.
907 2018-02-21 Jason Merrill <jason@redhat.com>
909 PR c++/84314 - ICE with templates and fastcall attribute.
910 * attribs.c (build_type_attribute_qual_variant): Remove assert.
912 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
914 * ipa-cp.c (determine_versionability): Fix comment typos.
916 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
919 * ipa-cp.c (determine_versionability): Do not version functions caling
922 2018-02-21 Martin Liska <mliska@suse.cz>
925 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
926 Add "native" as a possible value.
927 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
928 the macro when native cpu detection is available.
930 2018-02-21 Martin Liska <mliska@suse.cz>
933 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
934 Add "native" as a possible value.
935 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
936 when native cpu detection is available.
938 2018-02-21 Jakub Jelinek <jakub@redhat.com>
939 Martin Sebor <msebor@redhat.com>
941 PR tree-optimization/84478
942 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
944 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
945 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
946 support which is conservatively correct, for 2 only stay conservative
947 for maxlen. Formatting and comment capitalization fixes. Add STRICT
948 argument to the 2 argument get_range_strlen, adjust 6 arg
949 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
951 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
952 (gimple_fold_builtin_strlen): Pass true as last argument to
955 2018-02-20 Martin Sebor <msebor@redhat.com>
958 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
959 (builtin_memref::set_base_and_offset): Same. Handle inner references.
960 (builtin_memref::builtin_memref): Factor out parts into
961 set_base_and_offset and call it.
963 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
966 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
967 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
968 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
969 search at the associated MODE_INT.
971 2018-02-20 Jeff Law <law@redhat.com>
974 PR tree-optimization/81592
976 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
977 for range data rather than using global data.
978 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
979 range data rather than using global data.
980 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
981 pass it to children as needed.
982 (struct directive::fmtresult): Similarly.
983 (struct directive::set_width): Similarly.
984 (struct directive::set_precision): Similarly.
985 (format_integer, format_directive, parse_directive): Similarly.
986 (format_none): Accept unnamed vr_values parameter.
987 (format_percent, format_floating, format_character): Similarly.
988 (format_string, format_plain): Similarly.
989 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
990 the EVRP range analyzer for range data rather than using global data.
991 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
992 gimple-ssa-evrp-analyze.h
993 (class sprintf_dom_walker): Add after_dom_children member function.
994 Add evrp_range_analyzer member.
995 (sprintf_dom_walker::before_dom_children): Call into the EVRP
996 range analyzer as needed.
997 (sprintf_dom_walker::after_dom_children): New member function.
998 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
1000 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
1001 (evrp_range_analyzer::pop_to_marker): Likewise.
1003 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
1005 PR tree-optimization/84419
1006 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
1007 with the required type if its current type is compatible but
1010 2018-02-20 Jakub Jelinek <jakub@redhat.com>
1013 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
1014 after vectorization.
1016 2018-02-20 Martin Liska <mliska@suse.cz>
1019 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
1020 possible values if we don't have a hint.
1022 2018-02-20 Martin Liska <mliska@suse.cz>
1026 * final.c (shorten_branches): Build align_tab array with one
1028 * opts.c (finish_options): Add alignment option limit check.
1029 (MAX_CODE_ALIGN): Likewise.
1030 (MAX_CODE_ALIGN_VALUE): Likewise.
1031 * doc/invoke.texi: Document maximum allowed option value for
1032 all -falign-* options.
1034 2018-02-19 Jakub Jelinek <jakub@redhat.com>
1037 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
1038 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
1039 * var-tracking.c (emit_note_insn_var_location): Remove all references
1040 to NOTE_INSN_CALL_ARG_LOCATION.
1041 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
1042 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
1043 Use copy_rtx_if_shared.
1044 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
1045 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
1046 (dwarf2out_var_location): Remove handling of
1047 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
1049 * final.c (final_scan_insn): Remove all references to
1050 NOTE_INSN_CALL_ARG_LOCATION.
1051 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
1052 before dumping final insns.
1053 * except.c (emit_note_eh_region_end): Remove all references to
1054 NOTE_INSN_CALL_ARG_LOCATION.
1055 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
1056 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
1057 * config/arc/arc.c (hwloop_optimize): Likewise.
1058 * config/arm/arm.c (create_fix_barrier): Likewise.
1059 * config/s390/s390.c (s390_chunkify_start): Likewise.
1060 * config/sh/sh.c (find_barrier): Likewise.
1061 * config/i386/i386.c (rest_of_insert_endbranch,
1062 ix86_seh_fixup_eh_fallthru): Likewise.
1063 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
1064 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
1065 * config/frv/frv.c (frv_function_prologue): Likewise.
1066 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
1068 (note_outside_basic_block_p): Remove all references to
1069 NOTE_INSN_CALL_ARG_LOCATION.
1070 * gengtype.c (adjust_field_rtx_def): Likewise.
1071 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
1073 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
1074 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
1077 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
1080 PR tree-optimization/84452
1081 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
1082 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
1085 2018-02-19 Martin Liska <mliska@suse.cz>
1088 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
1090 2018-02-19 Martin Liska <mliska@suse.cz>
1091 Richard Sandiford <richard.sandiford@linaro.org>
1093 PR tree-optimization/82491
1094 * gimple-fold.c (get_base_constructor): Make earlier bail out
1097 2018-02-19 Carl Love <cel@us.ibm.com>
1099 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
1100 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
1102 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
1103 P8V_BUILTIN_VEC_NEG.
1105 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
1107 * config/rl78/rl78.md (movdf): New define expand.
1109 2018-02-19 Martin Liska <mliska@suse.cz>
1112 * doc/invoke.texi: Fix typo.
1113 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
1115 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
1117 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
1118 handle rs6000_single_float and rs6000_double_float specially for
1121 2018-02-16 Jeff Law <law@redhat.com>
1123 * config/rx/rx.c (add_pop_cfi_notes): New function.;
1126 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1129 * ipa-inline.c (inline_small_functions): Fix a typo.
1131 2018-02-16 Nathan Sidwell <nathan@acm.org>
1133 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
1135 2018-02-16 Carl Love <cel@us.ibm.com>
1137 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
1138 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
1139 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
1140 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
1141 expansion to P8V_BUILTIN_VEC_FLOAT2.
1143 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
1145 PR rtl-optimization/70023
1146 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
1147 src_regno into account.
1149 2018-02-16 Carl Love <cel@us.ibm.com>
1151 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
1152 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
1153 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
1154 * config/rs6000/rs6000.c: Remove case statements for
1155 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
1156 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
1157 and P9V_BUILTIN_VEC_VINSERT4B.
1158 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
1159 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
1160 * config/rs6000/vsx.md:
1161 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
1164 2018-02-16 Carl Love <cel@us.ibm.com>
1166 * config/rs6000/altivec.h: Add builtin names vec_extract4b
1168 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
1170 * config/rs6000/rs6000-c.c: Add the definitions for
1171 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
1172 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
1173 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
1174 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
1175 definition for insert4b and define insn *insert3b_internal.
1176 * doc/extend.texi: Add documentation for vec_extract4b.
1178 2018-02-16 Nathan Sidwell <nathan@acm.org>
1180 * doc/extend.texi (Backwards Compatibility): Mention friend
1181 injection. Note for-scope is deprecated.
1182 * doc/invoke.texi (-ffriend-injection): Deprecate.
1184 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
1186 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
1187 that moved to I2, also allow destinations that are a paradoxical
1188 subreg (instead of a normal reg).
1190 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
1193 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
1196 2018-02-16 Richard Biener <rguenther@suse.de>
1198 PR tree-optimization/84037
1199 PR tree-optimization/84016
1201 * config/i386/i386.c (ix86_builtin_vectorization_cost):
1202 Adjust vec_construct for the fact we need additional higher latency
1203 128bit inserts for AVX256 and AVX512 vector builds.
1204 (ix86_add_stmt_cost): Scale vector construction cost for
1207 2018-02-16 Richard Biener <rguenther@suse.de>
1209 PR tree-optimization/84417
1210 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
1211 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
1212 (non_rewritable_lvalue_p): Likewise, use poly-ints.
1214 2018-02-16 Martin Liska <mliska@suse.cz>
1217 * internal-fn.def (ASAN_CHECK): Set proper flags.
1218 (ASAN_MARK): Likewise.
1220 2018-02-16 Julia Koval <julia.koval@intel.com>
1222 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
1223 from PTA_CANNONLAKE.
1225 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1228 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
1229 Use ++iter rather than iter++ for std::list iterators.
1230 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
1231 defer deleting them until all nodes in the forest are processed. Do
1232 free even leaf nodes. Change to_process into auto_vec.
1235 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
1236 * vec.h (vec_default_construct): Use memset instead of placement new
1237 if BROKEN_VALUE_INITIALIZATION is defined.
1238 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
1239 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
1242 PR rtl-optimization/83723
1243 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
1244 * lra.c (lra_substitute_pseudo): Likewise. If true, use
1245 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
1247 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
1249 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
1251 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
1253 PR rtl-optimization/81443
1254 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
1255 from inner REGs to paradoxical SUBREGs.
1257 2018-02-16 Richard Biener <rguenther@suse.de>
1259 PR tree-optimization/84399
1260 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1261 For operands we can analyze at their definition make sure we can
1262 analyze them at each use as well.
1264 2018-02-16 Richard Biener <rguenther@suse.de>
1266 PR tree-optimization/84190
1267 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
1268 volatile accesses if the decl isn't volatile.
1270 2018-02-15 Jason Merrill <jason@redhat.com>
1272 PR c++/84314 - ICE with templates and fastcall attribute.
1273 * attribs.c (build_type_attribute_qual_variant): Don't clobber
1274 TYPE_CANONICAL on an existing type.
1276 2018-02-15 Jakub Jelinek <jakub@redhat.com>
1278 PR tree-optimization/84383
1279 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
1280 dstoff nor call operand_equal_p if dstbase is NULL.
1282 PR tree-optimization/84334
1283 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
1284 also a CONSTANT_CLASS_P, punt.
1286 2018-02-14 Jim Wilson <jimw@sifive.com>
1288 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
1289 first SMALL_OPERAND check. New local min_second_step. Move assert
1290 to where locals are set. Add TARGET_RVC support.
1291 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
1293 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
1295 * doc/invoke.texi: Correct -Wformat-overflow code sample.
1297 2018-02-14 Martin Sebor <msebor@redhat.com>
1299 PR tree-optimization/83698
1300 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
1301 arrays constrain the offset range to their bounds.
1302 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
1303 (builtin_access::overlap): Avoid setting the size of overlap if it's
1305 (maybe_diag_overlap): Also consider arrays when deciding what values
1306 of offsets to include in diagnostics.
1308 2018-02-14 Martin Sebor <msebor@redhat.com>
1311 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
1312 that correspond to the kind of a declaration.
1314 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
1317 * config/pa/pa.md: Load address of PIC label using the linkage table
1318 if the label is nonlocal.
1320 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
1322 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
1323 warning message if user requests -maltivec=be.
1324 * doc/invoke.texi: Document deprecation of -maltivec=be.
1326 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
1329 * config/rs6000/rs6000-c.c: Update definitions for
1330 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
1331 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
1333 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1336 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
1337 add _get_ssp intrinsics. Remove argument from
1338 __builtin_ia32_rdssp[d|q].
1339 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
1340 * config/i386/i386-builtin.def: Remove argument from
1341 __builtin_ia32_rdssp[d|q].
1342 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
1343 ix86_expand_special_args_builtin for _rdssp[d|q].
1344 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
1345 Clear register before usage.
1346 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
1347 Add documentation for new _get_ssp and _inc_ssp intrinsics.
1349 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
1351 PR tree-optimization/84357
1352 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
1353 operand 1 of an ARRAY_REF too.
1355 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
1358 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
1359 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1361 (set_of_reg): New struct.
1362 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
1363 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
1364 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1366 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
1367 Split into bitclr, bitset, bitinvert patterns if appropriate.
1368 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
1369 use rx_fuse_in_memory_bitop.
1370 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
1371 to named insn, correct maximum insn length.
1373 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
1376 * machmode.def: Define a complex mode for PARTIAL_INT.
1377 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
1379 * doc/rtl.texi: Document CSPImode.
1380 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
1382 (msp430_hard_regno_nregs_with_padding): Likewise.
1384 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
1387 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
1389 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
1391 PR rtl-optimization/84169
1392 * combine.c (try_combine): New variable split_i2i3. Set it to true if
1393 we generated a parallel as new i3 and we split that to new i2 and i3
1394 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
1395 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
1396 those to i2, not i1. Partially rewrite this scan code.
1398 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1401 * stor-layout.c (place_field): For variable length fields, adjust
1402 offset_align afterwards not just based on the field's alignment,
1403 but also on the size.
1406 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
1407 of exps and logs in the use_exp2 case.
1409 2018-02-13 Jeff Law <law@redhat.com>
1411 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
1414 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
1417 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
1421 * common.opt (gas-loc-support, gas-locview-support): New.
1422 (ginline-points, ginternal-reset-location-views): New.
1423 * doc/invoke.texi: Document them. Use @itemx where intended.
1424 (gvariable-location-views): Adjust.
1425 * target.def (reset_location_view): New.
1426 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
1427 (TARGET_RESET_LOCATION_VIEW): New.
1428 * doc/tm.texi: Rebuilt.
1429 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
1430 (dwarf2out_default_as_locview_support): New.
1431 (output_asm_line_debug_info): Use option variables.
1432 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
1433 (output_loc_list): Likewise.
1434 (add_high_low_attributes): Check option variables.
1435 Don't output entry view attribute in strict mode.
1436 (gen_inlined_subroutine_die): Check option variables.
1437 (dwarf2out_inline_entry): Likewise.
1438 (init_sections_and_labels): Likewise.
1439 (dwarf2out_early_finish): Likewise.
1440 (maybe_reset_location_view): New, from...
1441 (dwarf2out_var_location): ... here. Call it.
1442 * debug.h (dwarf2out_default_as_loc_support): Declare.
1443 (dwarf2out_default_as_locview_support): Declare.
1444 * hooks.c (hook_int_rtx_insn_0): New.
1445 * hooks.h (hook_int_rtx_insn_0): Declare.
1446 * toplev.c (process_options): Take -gas-loc-support and
1447 -gas-locview-support from dwarf2out. Enable
1448 -gvariable-location-views by default only with locview
1449 assembler support. Enable -ginternal-reset-location-views by
1450 default only if the target defines the corresponding hook.
1451 Enable -ginline-points by default if location views are
1452 enabled; force it disabled if statement frontiers are
1454 * tree-inline.c (expand_call_inline): Check option variables.
1455 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
1457 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
1459 PR tree-optimization/84321
1460 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
1461 handling. Also check whether the anti-range contains any values
1462 that satisfy the mask; switch to a VR_RANGE if not.
1464 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
1467 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
1469 2018-02-13 Martin Jambor <mjambor@suse.cz>
1472 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
1473 of call statements, also set location of a load to a temporary.
1475 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
1477 * config/rl78/rl78.c (add_vector_labels): New function.
1478 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
1479 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
1480 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
1481 which checks that no arguments are passed.
1482 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
1483 * doc/extend.texi: Documentation for the new attribute.
1485 2018-02-13 Andreas Schwab <schwab@suse.de>
1487 * config/riscv/linux.h (CPP_SPEC): Define.
1489 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1492 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
1493 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
1494 OPTION_MASK_ISA_AES as first argument to def_builtin_const
1495 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
1496 instead of OPTION_MASK_ISA_PCLMUL as first argument to
1497 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
1498 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
1499 temporarily for AES and PCLMUL builtins.
1501 PR tree-optimization/84339
1502 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
1503 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
1507 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
1508 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
1509 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
1511 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
1513 * omp-simd-clone.h: New file.
1514 * omp-simd-clone.c: Include omp-simd-clone.h.
1515 (expand_simd_clones): No longer static.
1516 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
1517 cgraph.h and omp-simd-clone.h.
1518 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
1519 (vect_recog_widen_shift_pattern): Formatting fix.
1520 (vect_pattern_recog_1): Don't check optab for calls.
1523 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
1524 operands[2] into a REG before using gen_lowpart on it.
1526 2018-02-12 Jeff Law <law@redhat.com>
1529 * config/sh/sh.c (find_barrier): Consider a sibling call
1532 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
1533 successfully back substituting a reg.
1535 2018-02-12 Richard Biener <rguenther@suse.de>
1537 PR tree-optimization/84037
1538 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
1539 parameter, move visited init to caller.
1540 (vect_slp_analyze_operations): Separate cost from validity
1541 check, initialize visited once for all instances.
1542 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
1544 * tree-vect-stmts.c (vect_model_simple_cost): Make early
1546 (vect_model_promotion_demotion_cost): Likewise.
1547 (vectorizable_bswap): Guard cost modeling with !slp_node
1548 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
1550 (vectorizable_call): Likewise.
1551 (vectorizable_conversion): Likewise.
1552 (vectorizable_assignment): Likewise.
1553 (vectorizable_shift): Likewise.
1554 (vectorizable_operation): Likewise.
1555 (vectorizable_store): Likewise.
1556 (vectorizable_load): Likewise.
1557 (vectorizable_condition): Likewise.
1558 (vectorizable_comparison): Likewise.
1560 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
1563 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
1564 (ASAN_MARK): Fix fnspec to account for return value, change pointer
1565 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
1567 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1570 * params.def (inline-min-speedup): Increase from 8 to 15.
1571 (max-inline-insns-auto): Decrease from 40 to 30.
1572 * ipa-split.c (consider_split): Add some buffer for function to
1573 be considered inlining candidate.
1574 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
1577 2018-02-12 Richard Biener <rguenther@suse.de>
1579 PR tree-optimization/84037
1580 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
1581 matched stmts if we cannot swap the non-matched ones.
1583 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1585 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
1586 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
1587 _mm_maskz_scalef_round_ss): New intrinsics.
1588 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
1589 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
1590 __builtin_ia32_scalefss_round): Remove.
1591 (__builtin_ia32_scalefsd_mask_round,
1592 __builtin_ia32_scalefss_mask_round): New intrinsics.
1593 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
1594 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
1595 ((match_operand:VF_128 2 "<round_nimm_predicate>"
1596 "<round_constraint>")): Changed to ...
1597 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
1598 "<round_scalar_constraint>")): ... this.
1599 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
1600 %0, %1, %2<round_op3>}"): Changed to ...
1601 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
1602 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
1603 %2<round_scalar_mask_op3>}"): ... this.
1604 * config/i386/subst.md (round_scalar_nimm_predicate): New.
1606 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1608 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
1609 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
1610 (_mm_maskz_sqrt_round_ss): New intrinsics.
1611 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
1612 (__builtin_ia32_sqrtsd_mask_round)
1613 (__builtin_ia32_sqrtss_mask_round): New builtins.
1614 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
1615 (__builtin_ia32_sqrtss_round): Remove.
1616 (__builtin_ia32_sqrtsd_mask_round)
1617 (__builtin_ia32_sqrtss_mask_round): New builtins.
1618 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
1619 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
1620 ((match_operand:VF_128 1 "vector_operand"
1621 "xBm,<round_constraint>")): Changed to ...
1622 ((match_operand:VF_128 1 "vector_operand"
1623 "xBm,<round_scalar_constraint>")): ... this.
1624 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
1625 %0, %2, %<iptr>1<round_op3>}): Changed to ...
1626 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
1627 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
1628 %<iptr>1<round_scalar_mask_op3>}): ... this.
1629 ((set_attr "prefix" "<round_prefix>")): Changed to ...
1630 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
1632 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
1635 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
1636 Cast vec_cmpeq result to correct type.
1637 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
1638 Cast vec_cmpgt result to correct type.
1640 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
1642 * final.c (final_scan_insn_1): Renamed from...
1643 (final_scan_insn): ... this. New wrapper, to recover
1644 seen from the outermost call in recursive ones.
1645 * config/sparc/sparc.c (output_return): Drop seen from call.
1646 (output_sibcall): Likewise.
1647 * config/visium/visium.c (output_branch): Likewise.
1649 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
1651 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
1654 2018-02-10 Alan Modra <amodra@gmail.com>
1657 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
1658 Specify LR as an input.
1660 2018-02-10 Jakub Jelinek <jakub@redhat.com>
1663 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
1664 remove_member_access_dummy_vars): New functions.
1665 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
1666 lower_omp_1, execute_lower_omp): Use them.
1668 PR rtl-optimization/84308
1669 * shrink-wrap.c (spread_components): Release todo vector.
1671 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
1673 PR rtl-optimization/57193
1674 * ira-color.c (struct allocno_color_data): Add member
1675 conflict_allocno_hard_prefs.
1676 (update_conflict_allocno_hard_prefs): New.
1677 (bucket_allocno_compare_func): Add a preference based on
1678 conflict_allocno_hard_prefs.
1679 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
1680 (color_allocnos): Remove a dead code. Initiate
1681 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
1683 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1686 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
1687 constraint from =wa to wa. Avoid a subreg on the output operand,
1688 instead use a pseudo and subreg it in a move.
1689 (p9_xxbrd_<mode>): Changed to ...
1690 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
1691 (p9_xxbrd_v2df): New expander.
1692 (p9_xxbrw_<mode>): Changed to ...
1693 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
1694 (p9_xxbrw_v4sf): New expander.
1696 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1698 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
1700 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
1703 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
1704 multiply in 32-bit mode.
1705 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
1706 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
1709 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1711 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
1712 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
1713 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
1714 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
1716 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
1719 * dwarf2out.c (is_trivial_indirect_ref): New function.
1720 (dwarf2out_late_global_decl): Do not generate a location
1721 attribute for variables that have a non-trivial DECL_VALUE_EXPR
1722 and that are not defined in the current unit.
1724 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1726 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
1727 instead of a libcall for UNORDERED.
1729 2018-02-09 Tamar Christina <tamar.christina@arm.com>
1732 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
1733 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
1735 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1738 * config/s390/s390.c (s390_set_current_function): Invoke
1739 s390_indirect_branch_settings also if fndecl didn't change.
1741 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1743 * config/rs6000/rs6000.md (blockage): Set length to zero.
1745 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1747 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
1749 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1752 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
1753 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
1757 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
1758 PARALLEL incoming that failed vt_get_decl_and_offset check.
1761 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
1762 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
1764 (get_variable_section): For decls in named .bss* sections pass true as
1765 second argument to bss_initializer_p.
1767 2018-02-09 Marek Polacek <polacek@redhat.com>
1768 Jakub Jelinek <jakub@redhat.com>
1771 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
1772 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
1773 Sync some changes from cxx_fold_indirect_ref.
1775 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1777 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
1779 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
1780 (BLOCK_INLINE_ENTRY_LABEL): New.
1781 (dwarf2out_var_location): Disregard inline entry markers.
1782 (inline_entry_data): New struct.
1783 (inline_entry_data_hasher): New hashtable type.
1784 (inline_entry_data_hasher::hash): New.
1785 (inline_entry_data_hasher::equal): New.
1786 (inline_entry_data_table): New variable.
1787 (add_high_low_attributes): Add DW_AT_entry_pc and
1788 DW_AT_GNU_entry_view attributes if a pending entry is found
1789 in inline_entry_data_table. Add old entry_pc attribute only
1790 if debug nonbinding markers are disabled.
1791 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
1792 markers are enabled.
1793 (block_within_block_p, dwarf2out_inline_entry): New.
1794 (dwarf2out_finish): Check that no entries remained in
1795 inline_entry_data_table.
1796 * final.c (reemit_insn_block_notes): Handle inline entry notes.
1797 (final_scan_insn, notice_source_line): Likewise.
1798 (rest_of_clean_state): Skip inline entry markers.
1799 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
1801 * gimple.c (gimple_build_debug_inline_entry): New.
1802 * gimple.h (enum gimple_debug_subcode): Add
1803 GIMPLE_DEBUG_INLINE_ENTRY.
1804 (gimple_build_debug_inline_entry): Declare.
1805 (gimple_debug_inline_entry_p): New.
1806 (gimple_debug_nonbind_marker_p): Adjust.
1807 * insn-notes.def (INLINE_ENTRY): New.
1808 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
1809 inline entry marker notes.
1810 (print_insn): Likewise.
1811 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
1812 (INSN_DEBUG_MARKER_KIND): Likewise.
1813 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
1814 * tree-inline.c (expand_call_inline): Build and insert
1815 debug_inline_entry stmt.
1816 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
1817 inline entry blocks early, if nonbind markers are enabled.
1818 (dump_scope_block): Dump fragment info.
1819 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
1820 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
1821 (gimple_build_debug_inline_entry): New.
1822 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
1823 Enable/disable inline entry points too.
1824 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
1825 (DEBUG_INSN): Describe inline entry markers.
1827 * common.opt (gvariable-location-views): New.
1828 (gvariable-location-views=incompat5): New.
1829 * config.in: Rebuilt.
1830 * configure: Rebuilt.
1831 * configure.ac: Test assembler for view support.
1832 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
1833 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
1834 * dwarf2out.c (var_loc_view): New typedef.
1835 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
1836 (dwarf2out_locviews_in_attribute): New.
1837 (dwarf2out_locviews_in_loclist): New.
1838 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
1839 (enum dw_line_info_opcode): Add LI_adv_address.
1840 (struct dw_line_info_table): Add view.
1841 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
1842 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
1843 (zero_view_p): New variable.
1844 (ZERO_VIEW_P): New macro.
1845 (output_asm_line_debug_info): New.
1846 (struct var_loc_node): Add view.
1847 (add_AT_view_list, AT_loc_list): New.
1848 (add_var_loc_to_decl): Add view param. Test it against last.
1849 (new_loc_list): Add view params. Record them.
1850 (AT_loc_list_ptr): Handle loc and view lists.
1851 (view_list_to_loc_list_val_node): New.
1852 (print_dw_val): Handle dw_val_class_view_list.
1853 (size_of_die): Likewise.
1854 (value_format): Likewise.
1855 (loc_list_has_views): New.
1856 (gen_llsym): Set vl_symbol too.
1857 (maybe_gen_llsym, skip_loc_list_entry): New.
1858 (dwarf2out_maybe_output_loclist_view_pair): New.
1859 (output_loc_list): Output view list or entries too.
1860 (output_view_list_offset): New.
1861 (output_die): Handle dw_val_class_view_list.
1862 (output_dwarf_version): New.
1863 (output_compilation_unit_header): Use it.
1864 (output_skeleton_debug_sections): Likewise.
1865 (output_rnglists, output_line_info): Likewise.
1866 (output_pubnames, output_aranges): Update version comments.
1867 (output_one_line_info_table): Output view numbers in asm comments.
1868 (dw_loc_list): Determine current endview, pass it to new_loc_list.
1869 Call maybe_gen_llsym.
1870 (loc_list_from_tree_1): Adjust.
1871 (add_AT_location_description): Create view list attribute if
1872 needed, check it's absent otherwise.
1873 (convert_cfa_to_fb_loc_list): Adjust.
1874 (maybe_emit_file): Call output_asm_line_debug_info for test.
1875 (dwarf2out_var_location): Reset views as needed. Precompute
1876 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
1877 attribute. Set view.
1878 (new_line_info_table): Reset next view.
1879 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
1880 (dwarf2out_source_line): Likewise. Output view resets and labels to
1881 the assembler, or select appropriate line info opcodes.
1882 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
1883 (optimize_string_length): Catch it. Adjust.
1884 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
1885 dw_val_class_view_list, and remove it if no longer needed.
1886 (hash_loc_list): Hash view numbers.
1887 (loc_list_hasher::equal): Compare them.
1888 (optimize_location_lists): Check whether a view list symbol is
1889 needed, and whether the locview attribute is present, and
1890 whether they match. Remove the locview attribute if no longer
1892 (index_location_lists): Call skip_loc_list_entry for test.
1893 (dwarf2out_finish): Call output_asm_line_debug_info for test.
1894 Use output_dwarf_version.
1895 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
1896 (struct dw_val_node): Add val_view_list.
1897 * final.c (SEEN_NEXT_VIEW): New.
1898 (set_next_view_needed): New.
1899 (clear_next_view_needed): New.
1900 (maybe_output_next_view): New.
1901 (final_start_function): Rename to...
1902 (final_start_function_1): ... this. Take pointer to FIRST,
1903 add SEEN parameter. Emit param bindings in the initial view.
1904 (final_start_function): Reintroduce SEEN-less interface.
1905 (final): Rename to...
1906 (final_1): ... this. Take SEEN parameter. Output final pending
1907 next view at the end.
1908 (final): Reintroduce seen-less interface.
1909 (final_scan_insn): Output pending next view before switching
1910 sections or ending a block. Mark the next view as needed when
1911 outputting variable locations. Notify debug backend of section
1912 changes, and of location view changes.
1913 (rest_of_handle_final): Adjust.
1914 * toplev.c (process_options): Autodetect value for debug variable
1915 location views option. Warn on incompat5 without -gdwarf-5.
1916 * doc/invoke.texi (gvariable-location-views): New.
1917 (gvariable-location-views=incompat5): New.
1918 (gno-variable-location-views): New.
1920 2018-02-08 David Malcolm <dmalcolm@redhat.com>
1922 PR tree-optimization/84136
1923 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
1924 that the result of find_edge is non-NULL.
1926 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1929 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
1930 storing integer register in SImode. Fix cost of 256 and 512
1931 byte aligned SSE register store.
1933 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1935 * config/i386/i386.c (ix86_multiplication_cost): Fix
1936 multiplication cost for TARGET_AVX512DQ.
1938 2018-02-08 Marek Polacek <polacek@redhat.com>
1940 PR tree-optimization/84238
1941 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
1944 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1946 PR tree-optimization/84265
1947 * tree-vect-stmts.c (vectorizable_store): Don't treat
1948 VMAT_CONTIGUOUS accesses as grouped.
1949 (vectorizable_load): Likewise.
1951 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1953 PR tree-optimization/81635
1954 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
1955 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
1956 (test_round_for_mask): New functions.
1957 (wide_int_cc_tests): Call test_round_for_mask.
1958 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
1959 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
1960 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
1961 range returned by get_range_info.
1963 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1966 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
1967 * symtab.c: Include builtins.h
1968 (symtab_node::output_to_lto_symbol_table_p): Move here
1969 from lto-streamer-out.c:output_symbol_p.
1970 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
1971 (output_symbol_p): Move all logic to symtab.c
1972 (produce_symtab): Update.
1974 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1976 * config/s390/s390-opts.h (enum indirect_branch): Define.
1977 * config/s390/s390-protos.h (s390_return_addr_from_memory)
1978 (s390_indirect_branch_via_thunk)
1979 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
1980 (enum s390_indirect_branch_type): Define.
1981 * config/s390/s390.c (struct s390_frame_layout, struct
1982 machine_function): Remove.
1983 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
1984 (indirect_branch_table_label_no, indirect_branch_table_name):
1986 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
1987 (enum s390_indirect_branch_option): Define.
1988 (s390_return_addr_from_memory): New function.
1989 (s390_handle_string_attribute): New function.
1990 (s390_attribute_table): Add new attribute handler.
1991 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
1992 (s390_indirect_branch_via_thunk): New function.
1993 (s390_indirect_branch_via_inline_thunk): New function.
1994 (s390_function_ok_for_sibcall): When jumping via thunk disallow
1995 sibling call optimization for non z10 compiles.
1996 (s390_emit_call): Force indirect branch target to be a single
1997 register. Add r1 clobber for non-z10 compiles.
1998 (s390_emit_epilogue): Emit return jump via return_use expander.
1999 (s390_reorg): Handle JUMP_INSNs as execute targets.
2000 (s390_option_override_internal): Perform validity checks for the
2001 new command line options.
2002 (s390_indirect_branch_attrvalue): New function.
2003 (s390_indirect_branch_settings): New function.
2004 (s390_set_current_function): Invoke s390_indirect_branch_settings.
2005 (s390_output_indirect_thunk_function): New function.
2006 (s390_code_end): Implement target hook.
2007 (s390_case_values_threshold): Implement target hook.
2008 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
2010 * config/s390/s390.h (struct s390_frame_layout)
2011 (struct machine_function): Move here from s390.c.
2012 (TARGET_INDIRECT_BRANCH_NOBP_RET)
2013 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
2014 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
2015 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
2016 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
2017 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
2018 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
2019 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
2020 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
2021 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
2022 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
2023 (mnemonic attribute): Add values which aren't recognized
2025 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
2026 pattern for branch conversion. Fix mnemonic attribute.
2027 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
2028 indirect branch via thunk if requested.
2029 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
2030 ("*indirect_jump"): Disable for branch conversion using out of
2032 ("indirect_jump_via_thunk<mode>_z10")
2033 ("indirect_jump_via_thunk<mode>")
2034 ("indirect_jump_via_inlinethunk<mode>_z10")
2035 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
2036 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
2037 ("casesi_jump_via_inlinethunk<mode>_z10")
2038 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
2039 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
2040 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
2041 ("*indirect2_jump"): Disable for branch conversion.
2042 ("casesi_jump"): Turn into expander and expand patterns for branch
2044 ("return_use"): New expander.
2045 ("*return"): Emit return via thunk and rename it to ...
2046 ("*return<mode>"): ... this one.
2047 * config/s390/s390.opt: Add new options and and enum for the
2050 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2052 * lra-constraints.c (match_reload): Unconditionally use
2053 gen_lowpart_SUBREG, rather than selecting between that
2054 and equivalent gen_rtx_SUBREG code.
2056 2018-02-08 Richard Biener <rguenther@suse.de>
2058 PR tree-optimization/84233
2059 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
2060 changed flag instead of boguously re-using phi_inserted.
2062 2018-02-08 Martin Jambor <mjambor@suse.cz>
2064 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
2065 static local variables.
2067 2018-02-08 Richard Biener <rguenther@suse.de>
2069 PR tree-optimization/84278
2070 * tree-vect-stmts.c (vectorizable_store): When looking for
2071 smaller vector types to perform grouped strided loads/stores
2072 make sure the mode is supported by the target.
2073 (vectorizable_load): Likewise.
2075 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2077 * config/aarch64/aarch64.c (aarch64_components_for_bb):
2078 Increase LDP/STP opportunities by adding adjacent callee-saves.
2080 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2082 PR rtl-optimization/84068
2083 PR rtl-optimization/83459
2084 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
2086 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
2088 PR tree-optimization/84224
2089 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
2090 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
2093 2018-02-07 Iain Sandoe <iain@codesourcery.com>
2096 * config/rs6000/altivec.md (*restore_world): Remove LR use.
2097 * config/rs6000/predicates.md (restore_world_operation): Adjust op
2098 count, remove one USE.
2100 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
2102 * doc/install.texi (Configuration): Document the
2103 --with-long-double-format={ibm,ieee} PowerPC configuration
2107 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
2108 Convert from define_expand to be define_insn_and_split. Rework
2109 float/double/_Float128 conversions to QI/HI/SImode to work with
2110 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
2111 conversions to QI/HImode types did a store and then a load to
2112 truncate the value. For conversions to VSX registers, don't split
2113 the insn, instead emit the code directly. Use the code iterator
2114 any_fix to combine signed and unsigned conversions.
2115 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
2116 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
2117 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2118 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
2119 (fix_<mode>di2_hw): Likewise.
2120 (fixuns_<mode>di2_hw): Likewise.
2121 (fix_<mode>si2_hw): Likewise.
2122 (fixuns_<mode>si2_hw): Likewise.
2123 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
2124 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2125 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
2126 fix<uns>_trunc<SFDF:mode>si2_p8.
2127 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
2129 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
2130 (fix<uns>_<mode>_mem): Likewise.
2131 (fctiw<u>z_<mode>_mem): Likewise.
2132 (fix<uns>_<mode>_mem): Likewise.
2133 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
2134 the register allocator from doing a direct move to the GPRs to do
2135 a store, and instead use the ISA 3.0 store byte/half-word from
2136 vector register instruction. For IEEE 128-bit floating point,
2137 also optimize stores of 32-bit ints.
2138 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
2140 2018-02-07 Alan Hayward <alan.hayward@arm.com>
2142 * genextract.c (push_pathstr_operand): New function to support
2144 (walk_rtx): Call push_pathstr_operand.
2145 (print_path): Support [a-zA-Z].
2147 2018-02-07 Richard Biener <rguenther@suse.de>
2149 PR tree-optimization/84037
2150 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
2151 (cse_and_gimplify_to_preheader): Declare.
2152 (vect_get_place_in_interleaving_chain): Likewise.
2153 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
2155 (_loop_vec_info::~_loop_vec_info): Delete it.
2156 (cse_and_gimplify_to_preheader): New function.
2157 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
2158 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
2159 (vectorizable_load): Likewise. For grouped stores always base
2160 the IV on the first element.
2161 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
2162 condition before gimplifying.
2164 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2166 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
2167 *DIV_EXPR and *MOD_EXPR.
2169 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
2172 * config/i386/i386.c (ix86_option_override_internal): Mask out
2173 the CF_SET bit when checking -fcf-protection.
2175 2018-02-07 Tom de Vries <tom@codesourcery.com>
2178 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
2181 2018-02-07 Richard Biener <rguenther@suse.de>
2183 PR tree-optimization/84204
2184 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
2187 PR tree-optimization/84205
2188 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
2189 special-case isl_ast_op_zdiv_r.
2191 PR tree-optimization/84223
2192 * graphite-scop-detection.c (gather_bbs::before_dom_children):
2193 Only add conditions from within the region.
2194 (gather_bbs::after_dom_children): Adjust.
2196 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
2199 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
2200 * config/avr/avr.md: Only post-reload split REG-REG moves if
2201 either register is GENERAL_REG_P.
2203 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2205 PR tree-optimization/84235
2206 * tree-ssa-scopedtables.c
2207 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
2208 if the subtraction is performed in floating point type where NaNs are
2209 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
2210 build 1. Formatting fix.
2212 2018-02-06 Jakub Jelinek <jakub@redhat.com>
2215 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
2216 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
2217 and skip it regardless of bb boundaries. Use CALL_P macro,
2218 don't test INSN_P (insn) together with CALL_P or JUMP_P check
2219 unnecessarily, formatting fix.
2221 2018-02-06 Michael Collison <michael.collison@arm.com>
2223 * config/arm/thumb2.md:
2224 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
2225 (*thumb_mov_notscc): Ditto.
2227 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
2230 * config/rs6000/rs6000.md (su code attribute): Use "u" for
2231 unsigned_fix, not "s".
2233 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2235 * configure.ac (gcc_fn_eh_frame_ro): New function.
2236 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
2237 correct .eh_frame permissions.
2238 * configure: Regenerate.
2240 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
2242 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
2245 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2247 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2248 Display warning message for -mno-speculate-indirect-jumps.
2250 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
2252 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
2254 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
2256 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
2258 PR tree-optimization/84225
2259 * tree-eh.c (find_trapping_overflow): Only call
2260 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
2262 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2265 * config/i386/i386.c: Reimplement the check of possible options
2266 -mibt/-mshstk conbination. Change error messages.
2267 * doc/invoke.texi: Fix a typo: remove extra '='.
2269 2018-02-06 Marek Polacek <polacek@redhat.com>
2271 PR tree-optimization/84228
2272 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
2274 2018-02-06 Tamar Christina <tamar.christina@arm.com>
2277 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
2278 emitted arch directives.
2279 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
2280 __ARM_FEATURE_COPROC before changing architectures.
2282 2018-02-06 Richard Biener <rguenther@suse.de>
2284 * config/i386/i386.c (print_reg): Fix typo.
2285 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
2287 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
2289 * configure: Regenerate.
2291 2018-02-05 Martin Sebor <msebor@redhat.com>
2293 PR tree-optimization/83369
2294 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
2297 2018-02-05 Martin Liska <mliska@suse.cz>
2299 * doc/invoke.texi: Cherry-pick upstream r323995.
2301 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
2303 * ira.c (ira_init_register_move_cost): Adjust comment.
2305 2018-02-05 Martin Liska <mliska@suse.cz>
2307 PR gcov-profile/84137
2308 * doc/gcov.texi: Fix typo in documentation.
2310 2018-02-05 Martin Liska <mliska@suse.cz>
2312 PR gcov-profile/83879
2313 * doc/gcov.texi: Document necessity of --dynamic-list-data when
2314 using dlopen functionality.
2316 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
2318 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
2319 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
2320 _mm_maskz_range_ss, _mm_mask_range_round_ss,
2321 _mm_maskz_range_round_ss): New intrinsics.
2322 (__builtin_ia32_rangesd128_round)
2323 (__builtin_ia32_rangess128_round): Remove.
2324 (__builtin_ia32_rangesd128_mask_round,
2325 __builtin_ia32_rangess128_mask_round): New builtins.
2326 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
2327 __builtin_ia32_rangess128_round): Remove.
2328 (__builtin_ia32_rangesd128_mask_round,
2329 __builtin_ia32_rangess128_mask_round): New builtins.
2330 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
2331 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
2332 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
2333 "<round_saeonly_constraint>")): Changed to ...
2334 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
2335 "<round_saeonly_scalar_constraint>")): ... this.
2336 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
2337 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
2338 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
2339 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
2340 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
2342 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
2344 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
2346 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
2347 Remove all values except native, 8540 and 8548.
2349 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
2351 * config/i386/i386.c (ix86_output_function_return): Pass
2352 INVALID_REGNUM, instead of -1, as invalid register number to
2353 indirect_thunk_name and output_indirect_thunk.
2355 2018-02-02 Julia Koval <julia.koval@intel.com>
2357 * config.gcc: Add -march=icelake.
2358 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
2359 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
2360 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
2361 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
2362 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
2363 (processor_target_table): Add icelake.
2364 (ix86_option_override_internal): Handle new PTAs.
2365 (get_builtin_code_for_version): Handle icelake.
2366 (M_INTEL_COREI7_ICELAKE): New.
2367 (fold_builtin_cpu): Handle icelake.
2368 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
2369 * doc/invoke.texi: Add -march=icelake.
2371 2018-02-02 Julia Koval <julia.koval@intel.com>
2373 * config/i386/i386.c (ix86_option_override_internal): Change flags type
2374 to wide_int_bitmask.
2375 * wide-int-bitmask.h: New.
2377 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2380 * config/i386/i386.md: Replace Pmode with word_mode in
2381 builtin_setjmp_setup and builtin_longjmp to support x32.
2383 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
2387 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
2389 (rs6000_supported_cpu_names): New static variable.
2390 (linux_cpu_translation_table): Likewise.
2391 (elf_platform) <cpu>: Define new static variable and use it.
2392 Translate kernel AT_PLATFORM name to canonical name if needed.
2393 Error if platform name is unknown.
2395 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
2398 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
2400 2018-02-01 Jeff Law <law@redhat.com>
2403 * config/i386/i386.c (release_scratch_register_on_entry): Add new
2404 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
2405 the scratch if RELEASE_VIA_POP is false.
2406 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
2407 If we have to save a temporary register, decrement SIZE appropriately.
2408 Pass new arguments to release_scratch_register_on_entry.
2409 (ix86_adjust_stack_and_probe): Likewise.
2410 (ix86_emit_probe_stack_range): Pass new arguments to
2411 release_scratch_register_on_entry.
2413 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
2415 PR rtl-optimization/84157
2416 * combine.c (change_zero_ext): Use REG_P predicate in
2417 front of HARD_REGISTER_P predicate.
2419 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
2421 * config/avr/avr.c (avr_option_override): Move disabling of
2422 -fdelete-null-pointer-checks to...
2423 * common/config/avr/avr-common.c (avr_option_optimization_table):
2426 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2428 PR tree-optimization/81635
2429 * tree-data-ref.c (split_constant_offset_1): For types that
2430 wrap on overflow, try to use range info to prove that wrapping
2433 2018-02-01 Renlin Li <renlin.li@arm.com>
2436 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
2438 (aarch64_register_move_cost): Likewise.
2439 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
2441 (REG_CLASS_NAMES): Likewise.
2442 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
2443 TAILCALL_ADDR_REGS. Remove IP registers.
2444 * config/aarch64/aarch64.md (Ucs): Update register constraint.
2446 2018-02-01 Richard Biener <rguenther@suse.de>
2448 * domwalk.h (dom_walker::dom_walker): Add additional constructor
2449 for specifying RPO order and allow NULL for that.
2450 * domwalk.c (dom_walker::dom_walker): Likewise.
2451 (dom_walker::walk): Handle NULL RPO order.
2452 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
2454 (rewrite_update_dom_walker): Likewise.
2455 (mark_def_dom_walker): Likewise.
2457 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2459 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
2460 (aarch64_maybe_expand_sve_subreg_move): Declare.
2461 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
2462 * config/aarch64/predicates.md (aarch64_any_register_operand): New
2464 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
2465 that are semantically a reverse operation.
2466 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
2467 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
2468 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
2470 (aarch64_can_change_mode_class): For big-endian, forbid changes
2471 between two SVE modes if they have different element sizes.
2473 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2475 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
2476 the TImode handling for big-endian targets.
2478 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2480 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
2481 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
2483 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
2484 Remove BSWAP handing for big-endian targets and use the form of
2485 LD1RQ appropariate for the mode.
2487 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2489 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
2490 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
2493 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2496 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
2497 check for operands that need to go through aarch64_sve_reload_be.
2499 2018-02-01 Jakub Jelinek <jakub@redhat.com>
2501 PR tree-optimization/81661
2502 PR tree-optimization/84117
2503 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
2504 * tree-eh.c: Include gimplify.h.
2505 (find_trapping_overflow, replace_trapping_overflow,
2506 rewrite_to_non_trapping_overflow): New functions.
2507 * tree-vect-loop.c: Include tree-eh.h.
2508 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
2509 * tree-data-ref.c: Include tree-eh.h.
2510 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
2512 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
2514 PR rtl-optimization/84123
2515 * combine.c (change_zero_ext): Check if hard register satisfies
2516 can_change_dest_mode before calling gen_lowpart_SUBREG.
2518 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
2521 * ira.c (ira_init_register_move_cost): Remove assert.
2523 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2525 PR rtl-optimization/84071
2526 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
2527 * doc/tm.texi: Regenerate.
2529 2018-01-31 Richard Biener <rguenther@suse.de>
2531 PR tree-optimization/84132
2532 * tree-data-ref.c (analyze_miv_subscript): Properly
2533 check whether evolution_function_is_affine_multivariate_p
2534 before calling gcd_of_steps_may_divide_p.
2536 2018-01-31 Julia Koval <julia.koval@intel.com>
2539 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
2540 * config/i386/i386.md (rdpid_rex64) New.
2541 (rdpid): Make 32bit only.
2543 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
2546 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
2547 an IDENTIFIER_NODE for FUNCTION_TYPE's.
2549 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2552 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
2554 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
2556 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2558 PR rtl-optimization/84071
2559 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
2560 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
2562 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2564 * config/arc/arc.c (arc_handle_aux_attribute): New function.
2565 (arc_attribute_table): Add 'aux' attribute.
2566 (arc_in_small_data_p): Consider aux like variables.
2567 (arc_is_aux_reg_p): New function.
2568 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
2569 (arc_get_aux_arg): New function.
2570 (prepare_move_operands): Handle aux-register access.
2571 (arc_handle_aux_attribute): New function.
2572 * doc/extend.texi (ARC Variable attributes): Add subsection.
2574 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2576 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
2577 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
2578 (arc_attribute_table): Add 'uncached' attribute.
2579 (arc_print_operand): Print '.di' flag for uncached memory
2581 (arc_in_small_data_p): Do not consider for small data the uncached
2583 (arc_is_uncached_mem_p): New function.
2584 * config/arc/predicates.md (compact_store_memory_operand): Check
2585 for uncached memory accesses.
2586 (nonvol_nonimm_operand): Likewise.
2587 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
2589 2018-01-31 Jakub Jelinek <jakub@redhat.com>
2592 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
2593 falign-loops=): Add Optimization flag.
2595 2018-01-30 Jeff Law <law@redhat.com>
2598 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
2599 INT_REGISTERS_SAVED. Check it prior to calling
2600 get_scratch_register_on_entry.
2601 (ix86_adjust_stack_and_probe): Similarly.
2602 (ix86_emit_probe_stack_range): Similarly.
2603 (ix86_expand_prologue): Corresponding changes.
2605 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2608 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
2609 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
2611 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
2614 * lra-constraints.c (curr_insn_transform): Process AND in the
2617 2018-01-30 Jakub Jelinek <jakub@redhat.com>
2619 PR rtl-optimization/83986
2620 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
2621 dependence against last_pending_memory_flush in addition to
2624 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
2626 PR tree-optimization/81611
2627 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
2630 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2633 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
2636 2018-01-30 Richard Biener <rguenther@suse.de>
2637 Jakub Jelinek <jakub@redhat.com>
2639 PR tree-optimization/84111
2640 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
2641 inner loops added during recursion, as they don't have up-to-date
2644 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2647 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
2648 (can_inline_edge_by_limits_p): ... here.
2649 (can_early_inline_edge_p, check_callers,
2650 update_caller_keys, update_callee_keys, recursive_inlining,
2651 add_new_edges_to_heap, speculation_useful_p,
2652 inline_small_functions,
2653 inline_small_functions, flatten_function,
2654 inline_to_all_callers_1): Update.
2656 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2658 * profile-count.c (profile_count::combine_with_ipa_count): Handle
2661 2018-01-30 Richard Biener <rguenther@suse.de>
2663 PR tree-optimization/83008
2664 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
2665 invariant and constant vector uses in stmts when they need
2668 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2671 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
2672 * configure: Regenerate.
2674 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2676 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
2678 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
2679 Use gen_rtx_REG rather than gen_lowpart.
2681 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2683 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
2684 rather than 0 when creating partial subregs.
2686 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2688 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
2691 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
2694 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
2695 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
2696 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
2697 flags. This restores the settings used before the 2017-07-24.
2698 Turning off pre increment/decrement/modify allows IVOPTS to
2699 optimize DF/SF loops where the index is an int.
2701 2018-01-29 Richard Biener <rguenther@suse.de>
2702 Kelvin Nilsen <kelvin@gcc.gnu.org>
2705 * tree-vect-stmts.c (vectorizable_call): Don't call
2706 targetm.vectorize_builtin_md_vectorized_function if callee is
2709 2018-01-22 Carl Love <cel@us.ibm.com>
2711 * doc/extend.tex: Fix typo in second arg in
2712 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
2714 2018-01-29 Richard Biener <rguenther@suse.de>
2716 PR tree-optimization/84086
2717 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
2718 (flush_ssaname_freelist): When SSA names were released reset
2719 the SCEV hash table.
2721 2018-01-29 Richard Biener <rguenther@suse.de>
2723 PR tree-optimization/84057
2724 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
2725 removed paths when removing edges.
2727 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
2729 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
2730 -mfunction-return=@var{choice}.
2732 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
2735 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
2737 (layout::print_source_line): Likewise.
2738 (test_get_line_width_without_trailing_whitespace): Add test cases.
2740 2018-01-27 Jakub Jelinek <jakub@redhat.com>
2743 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
2746 2018-01-26 Jim Wilson <jimw@sifive.com>
2748 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
2750 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
2753 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2755 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
2756 and CMP + SUB-immediate -> SUBS.
2758 2018-01-26 Martin Sebor <msebor@redhat.com>
2760 PR tree-optimization/83896
2761 * tree-ssa-strlen.c (get_string_len): Rename...
2762 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
2763 Avoid assuming length is constant.
2764 (handle_char_store): Use HOST_WIDE_INT for string length.
2766 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
2769 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
2770 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
2772 2018-01-26 Richard Biener <rguenther@suse.de>
2774 PR rtl-optimization/84003
2775 * dse.c (record_store): Only record redundant stores when
2776 the earlier store aliases at least all accesses the later one does.
2778 2018-01-26 Jakub Jelinek <jakub@redhat.com>
2780 PR rtl-optimization/83985
2781 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
2782 REG_CFA_RESTORE insns.
2783 (delete_unmarked_insns): Don't ignore separate shrink wrapping
2784 REG_CFA_RESTORE insns here.
2787 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
2788 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
2790 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2792 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
2793 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
2794 (arc_init): Likewise.
2795 (arc_override_options): Likewise.
2796 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
2798 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
2800 * config/arc/arc.h (TARGET_DBNZ): Define.
2801 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
2802 properly set the tune attribute.
2803 (dbnz): Use TARGET_DBNZ guard.
2804 * config/arc/arc.opt (mtune): Add core3 option.
2806 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2808 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
2809 recognize new pic like addresses.
2810 (arc_delegitimize_address): Clean up.
2812 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2814 * config/arc/arc-arches.def: Option mrf16 valid for all
2816 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
2817 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
2818 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
2819 * config/arc/arc-tables.opt: Regenerate.
2820 * config/arc/arc.c (arc_conditional_register_usage): Handle
2821 reduced register file case.
2822 (arc_file_start): Set must have build attributes.
2823 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
2825 * config/arc/arc.opt (mrf16): Add new option.
2826 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
2827 * config/arc/genmultilib.awk: Handle new mrf16 option.
2828 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
2829 * config/arc/t-multilib: Regenerate.
2830 * doc/invoke.texi (ARC Options): Document mrf16 option.
2832 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2834 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
2835 * config/arc/arc.c (arc_handle_secure_attribute): New function.
2836 (arc_attribute_table): Add 'secure_call' attribute.
2837 (arc_print_operand): Print secure call operand.
2838 (arc_function_ok_for_sibcall): Don't optimize tail calls when
2840 (arc_is_secure_call_p): New function. * config/arc/arc.md
2841 (call_i): Add support for sjli instruction.
2842 (call_value_i): Likewise.
2843 * config/arc/constraints.md (Csc): New constraint.
2845 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2846 John Eric Martin <John.Martin@emmicro-us.com>
2848 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
2849 * config/arc/arc.c (_arc_jli_section): New struct.
2850 (arc_jli_section): New type.
2851 (rc_jli_sections): New static variable.
2852 (arc_handle_jli_attribute): New function.
2853 (arc_attribute_table): Add jli_always and jli_fixed attribute.
2854 (arc_file_end): New function.
2855 (TARGET_ASM_FILE_END): Define.
2856 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
2857 (arc_add_jli_section): New function.
2858 (jli_call_scan): Likewise.
2859 (arc_reorg): Call jli_call_scan.
2860 (arc_output_addsi): Remove 'S' from printing asm operand.
2861 (arc_is_jli_call_p): New function.
2862 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
2864 (movhi_insn): Likewise.
2865 (movsi_insn): Likewise.
2866 (movsi_set_cc_insn): Likewise.
2867 (loadqi_update): Likewise.
2868 (load_zeroextendqisi_update): Likewise.
2869 (load_signextendqisi_update): Likewise.
2870 (loadhi_update): Likewise.
2871 (load_zeroextendhisi_update): Likewise.
2872 (load_signextendhisi_update): Likewise.
2873 (loadsi_update): Likewise.
2874 (loadsf_update): Likewise.
2875 (movsicc_insn): Likewise.
2876 (bset_insn): Likewise.
2877 (bxor_insn): Likewise.
2878 (bclr_insn): Likewise.
2879 (bmsk_insn): Likewise.
2880 (bicsi3_insn): Likewise.
2881 (cmpsi_cc_c_insn): Likewise.
2882 (movsi_ne): Likewise.
2883 (movsi_cond_exec): Likewise.
2884 (clrsbsi2): Likewise.
2895 (call_i): Remove 'S' asm letter, add jli instruction.
2896 (call_value_i): Likewise.
2897 * config/arc/arc.op (mjli-always): New option.
2898 * config/arc/constraints.md (Cji): New constraint.
2899 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
2901 (subsf3_fpx): Likewise.
2902 (mulsf3_fpx): Likewise.
2903 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
2905 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
2907 * doc/invoke.texi (ARC): Document mjli-always option.
2909 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
2911 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
2912 avoid addition with 0 and use incw and decw where possible.
2914 2018-01-26 Richard Biener <rguenther@suse.de>
2916 PR tree-optimization/81082
2917 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
2918 association if it requires casting to unsigned.
2919 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
2920 from fold_plusminus_mult_expr to catch important cases late when
2921 range info is available.
2923 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2925 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
2926 * configure.ac (hidden_linkonce): New test.
2927 * configure: Regenerate.
2928 * config.in: Regenerate.
2930 2018-01-26 Julia Koval <julia.koval@intel.com>
2932 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
2933 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
2934 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
2935 _mm_mask_bitshuffle_epi64_mask): Fix type.
2936 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
2937 USI_FTYPE_V4DI_V4DI_USI): Remove.
2938 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
2939 __builtin_ia32_vpshufbitqmb256_mask,
2940 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
2941 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
2942 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
2944 2018-01-26 Alan Modra <amodra@gmail.com>
2947 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
2948 UNSPEC_VBPERMQ. Sort other unspecs.
2950 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
2952 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
2954 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
2957 * predict.c (drop_profile): Do not push/pop cfun; update also
2959 (handle_missing_profiles): Fix logic looking for zero profiles.
2961 2018-01-25 Jakub Jelinek <jakub@redhat.com>
2964 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
2965 on functions with #pragma omp declare simd or functions with simd
2967 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
2968 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
2969 Remove trailing \n from warning_at calls.
2971 2018-01-25 Tom de Vries <tom@codesourcery.com>
2974 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
2975 for neutered workers.
2977 2018-01-24 Joseph Myers <joseph@codesourcery.com>
2980 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
2981 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
2983 2018-01-24 Jeff Law <law@redhat.com>
2986 * i386.c (get_probe_interval): Move to earlier point.
2987 (ix86_compute_frame_layout): If -fstack-clash-protection and
2988 the frame is larger than the probe interval, then use pushes
2989 to save registers rather than reg->mem moves.
2990 (ix86_expand_prologue): Remove conditional for int_registers_saved
2993 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
2996 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
2997 min/max for never referenced object.
2999 2018-01-24 Jakub Jelinek <jakub@redhat.com>
3002 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
3004 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
3005 attributes from DECL_ATTRIBUTES (decl) without affecting
3006 DECL_ATTRIBUTES (current_function_decl).
3007 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
3008 functions with non-NULL DECL_ABSTRACT_ORIGIN.
3010 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
3012 PR tree-optimization/83979
3013 * fold-const.c (fold_comparison): Use constant_boolean_node
3014 instead of boolean_{true,false}_node.
3016 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
3018 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
3021 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3023 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3024 Simplify the clause that sets the length attribute.
3025 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3026 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
3027 clause that sets the length attribute.
3028 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3030 2018-01-24 Tom de Vries <tom@codesourcery.com>
3033 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
3034 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
3035 Add strict parameter.
3036 (prevent_branch_around_nothing): Insert dummy insn between branch to
3037 label and label with no ptx insn inbetween.
3038 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
3040 2018-01-24 Tom de Vries <tom@codesourcery.com>
3043 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3044 for neutered threads in warp.
3045 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
3047 2018-01-24 Richard Biener <rguenther@suse.de>
3049 PR tree-optimization/83176
3050 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
3053 2018-01-24 Richard Biener <rguenther@suse.de>
3055 PR tree-optimization/82819
3056 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
3057 code generating pluses that are no-ops in the target precision.
3059 2018-01-24 Richard Biener <rguenther@suse.de>
3062 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
3064 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3066 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
3067 to merge probabilities.
3068 * predict.c (probably_never_executed): Also mark as cold functions
3069 with global 0 profile and guessed local profile.
3070 * profile-count.c (profile_probability::combine_with_count): New
3072 * profile-count.h (profile_probability::operator*,
3073 profile_probability::operator*=, profile_probability::operator/,
3074 profile_probability::operator/=): Reduce precision to adjusted
3075 and set value to guessed on contradictory divisions.
3076 (profile_probability::combine_with_freq): Remove.
3077 (profile_probability::combine_wiht_count): Declare.
3078 (profile_count::force_nonzero):: Set to adjusted.
3079 (profile_count::probability_in):: Set quality to adjusted.
3080 * tree-ssa-tail-merge.c (replace_block_by): Use
3083 2018-01-23 Andrew Waterman <andrew@sifive.com>
3084 Jim Wilson <jimw@sifive.com>
3086 * config/riscv/riscv.c (riscv_stack_boundary): New.
3087 (riscv_option_override): Set riscv_stack_boundary. Handle
3088 riscv_preferred_stack_boundary_arg.
3089 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
3090 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
3091 (STACK_BOUNDARY): Set to riscv_stack_boundary.
3092 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
3093 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
3094 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
3096 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
3099 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
3100 of struct ix86_frame.
3101 (ix86_expand_epilogue): Likewise. Add a local variable for
3102 the reg_save_offset field in struct ix86_frame.
3104 2018-01-23 Bin Cheng <bin.cheng@arm.com>
3106 PR tree-optimization/82604
3107 * tree-loop-distribution.c (enum partition_kind): New enum item
3108 PKIND_PARTIAL_MEMSET.
3109 (partition_builtin_p): Support above new enum item.
3110 (generate_code_for_partition): Ditto.
3111 (compute_access_range): Differentiate cases that equality can be
3112 proven at all loops, the innermost loops or no loops.
3113 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
3114 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
3115 (finalize_partitions, distribute_loop): Don't fuse partition of
3116 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
3117 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
3120 2018-01-23 Martin Liska <mliska@suse.cz>
3122 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
3123 order to ignore the predictor.
3124 (PRED_POLYMORPHIC_CALL): Likewise.
3125 (PRED_RECURSIVE_CALL): Likewise.
3127 2018-01-23 Martin Liska <mliska@suse.cz>
3129 * tree-profile.c (tree_profiling): Print function header to
3130 aware reader which function we are working on.
3131 * value-prof.c (gimple_find_values_to_profile): Do not print
3132 not interesting value histograms.
3134 2018-01-23 Martin Liska <mliska@suse.cz>
3136 * profile-count.h (enum profile_quality): Add
3137 profile_uninitialized as the first value. Do not number values
3138 as they are zero based.
3139 (profile_count::verify): Update sanity check.
3140 (profile_probability::verify): Likewise.
3142 2018-01-23 Nathan Sidwell <nathan@acm.org>
3144 * doc/invoke.texi (ffor-scope): Deprecate.
3146 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3148 PR tree-optimization/83510
3149 * domwalk.c (set_all_edges_as_executable): New function.
3150 (dom_walker::dom_walker): Convert bool param
3151 "skip_unreachable_blocks" to enum reachability. Move setup of
3152 edge flags to set_all_edges_as_executable and only do it when
3153 reachability is REACHABLE_BLOCKS.
3154 * domwalk.h (enum dom_walker::reachability): New enum.
3155 (dom_walker::dom_walker): Convert bool param
3156 "skip_unreachable_blocks" to enum reachability.
3157 (set_all_edges_as_executable): New decl.
3158 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
3159 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
3161 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
3162 but converting true to REACHABLE_BLOCKS.
3163 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
3165 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
3166 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
3167 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
3169 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
3170 if check_all_array_refs will be called.
3172 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3174 * tree.c (selftest::test_location_wrappers): Add more test
3177 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3179 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
3180 (selftest::test_bit_in_range): Likewise.
3182 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3185 * doc/sourcebuild.texi (vect_float): Say that the selector
3186 only describes the situation when -funsafe-math-optimizations is on.
3187 (vect_float_strict): Document.
3189 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3191 PR tree-optimization/83965
3192 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
3193 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
3194 instead of checking only for a reduction.
3195 (vect_recog_widen_sum_pattern): Likewise.
3197 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3199 * predict.c (probably_never_executed): Only use precise profile info.
3200 (compute_function_frequency): Skip after inlining hack since we now
3201 have quality checking.
3203 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3205 * profile-count.h (profile_probability::very_unlikely,
3206 profile_probability::unlikely, profile_probability::even): Set
3207 precision to guessed.
3209 2018-01-23 Richard Biener <rguenther@suse.de>
3211 PR tree-optimization/83963
3212 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
3213 Properly terminate dominator walk when crossing the exit edge not
3214 when visiting its source block.
3216 2018-01-23 Jakub Jelinek <jakub@redhat.com>
3219 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
3220 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
3222 2018-01-22 Jakub Jelinek <jakub@redhat.com>
3224 PR tree-optimization/83957
3225 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
3226 semicolon after for body surrounded by braces.
3228 PR tree-optimization/83081
3229 * profile-count.h (profile_probability::split): New method.
3230 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
3231 Use profile_probability::split.
3232 (do_compare_rtx_and_jump): Fix adjustment of probabilities
3233 when splitting a single conditional jump into 2.
3235 2018-01-22 David Malcolm <dmalcolm@redhat.com>
3237 PR tree-optimization/69452
3238 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
3241 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3243 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
3244 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
3245 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
3247 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3249 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
3250 * config/rl78/rl78.md (movdi): New define_expand.
3251 * config/rl78/rl78.c (rl78_split_movdi): New function.
3253 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
3256 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
3258 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
3259 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
3260 128-bit to produce an UNSPEC move to get the double word with the
3261 signbit and then a shift directly to do signbit.
3262 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
3263 implementation with a new version that just does either a direct
3264 move or a regular move. Move memory interface to separate insns.
3265 Move insns so they are next to the expander.
3266 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
3267 with signbit move. Split big and little endian case.
3268 (signbit<mode>2_dm_mem_le): Likewise.
3269 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
3270 (signbit<mode>2_dm2): Likewise.
3272 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3274 * config/rl78/rl78.md (anddi3): New define_expand.
3276 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3278 * config/rl78/rl78.md (umindi3): New define_expand.
3280 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3282 * config/rl78/rl78.md (smindi3): New define_expand.
3284 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3286 * config/rl78/rl78.md (smaxdi3): New define_expand.
3288 2018-01-22 Carl Love <cel@us.ibm.com>
3290 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
3291 LVX_V1TI): Add macro expansion.
3292 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
3293 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
3294 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
3295 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
3296 Change check to determine if the instruction is a byte reversing
3297 entry. Fix typo in comment.
3298 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
3299 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
3300 Add def_builtin calls for new builtins.
3301 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
3302 Add define_insn expansion.
3304 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3306 * config/rl78/rl78.md (umaxdi3): New define_expand.
3308 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3310 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
3311 for non-QImode registers.
3313 2018-01-22 Richard Biener <rguenther@suse.de>
3315 PR tree-optimization/83963
3316 * graphite-scop-detection.c (scop_detection::get_sese): Delay
3317 including the loop exit block.
3318 (scop_detection::merge_sese): Likewise.
3319 (scop_detection::add_scop): Do it here instead.
3321 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3323 * doc/sourcebuild.texi (arm_softfloat): Document.
3325 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
3328 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
3329 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
3330 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
3332 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3333 David Edelsohn <dje.gcc@gmail.com>
3336 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3337 Change "crset eq" to "crset 2".
3338 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3339 (*call_indirect_aix<mode>_nospec): Likewise.
3340 (*call_value_indirect_aix<mode>_nospec): Likewise.
3341 (*call_indirect_elfv2<mode>_nospec): Likewise.
3342 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
3343 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
3344 change assembly output from . to $.
3345 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3346 (indirect_jump<mode>_nospec): Change assembly output from . to $.
3347 (*tablejump<mode>_internal1_nospec): Likewise.
3349 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
3352 * config/sh/sh_optimize_sett_clrt.cc:
3353 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
3355 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
3357 PR tree-optimization/83940
3358 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
3359 offset_dt to vect_constant_def rather than vect_unknown_def_type.
3360 (vect_check_load_store_mask): Add a mask_dt_out parameter and
3361 use it to pass back the definition type.
3362 (vect_check_store_rhs): Likewise rhs_dt_out.
3363 (vect_build_gather_load_calls): Add a mask_dt argument and use
3364 it instead of a call to vect_is_simple_use.
3365 (vectorizable_store): Update calls to vect_check_load_store_mask
3366 and vect_check_store_rhs. Use the dt returned by the latter instead
3367 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
3368 instead of calls to vect_is_simple_use. Pass the scalar rather
3369 than the vector operand to vect_is_simple_use when handling
3370 second and subsequent copies of an rhs value.
3371 (vectorizable_load): Update calls to vect_check_load_store_mask
3372 and vect_build_gather_load_calls. Use the cached mask_dt and
3373 gs_info.offset_dt instead of calls to vect_is_simple_use.
3375 2018-01-20 Jakub Jelinek <jakub@redhat.com>
3378 * tree-emutls.c: Include gimplify.h.
3379 (lower_emutls_2): New function.
3380 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
3381 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
3382 it before further processing.
3385 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
3386 UINTVAL (trueop1) instead of INTVAL (op1).
3388 2018-01-19 Jakub Jelinek <jakub@redhat.com>
3392 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
3393 INCOMING_FRAME_SP_OFFSET if not defined.
3394 (scan_trace): Add ENTRY argument. If true and
3395 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
3396 emit a note to adjust the CFA offset.
3397 (create_cfi_notes): Adjust scan_trace callers.
3398 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
3399 INCOMING_FRAME_SP_OFFSET in the CIE.
3400 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
3401 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
3403 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
3404 * doc/tm.texi: Regenerated.
3406 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3408 PR rtl-optimization/83147
3409 * lra-constraints.c (remove_inheritance_pseudos): Use
3410 lra_substitute_pseudo_within_insn.
3412 2018-01-19 Tom de Vries <tom@codesourcery.com>
3413 Cesar Philippidis <cesar@codesourcery.com>
3416 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
3418 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
3421 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
3422 spaces for function labels.
3424 2018-01-19 Martin Liska <mliska@suse.cz>
3426 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
3427 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
3428 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
3429 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
3430 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
3431 (PRED_CONST_RETURN): Change from 69 to 65.
3432 (PRED_NULL_RETURN): Change from 91 to 71.
3433 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
3434 (PRED_LOOP_GUARD): Change from 66 to 73.
3436 2018-01-19 Martin Liska <mliska@suse.cz>
3438 * predict.c (predict_insn_def): Add new assert.
3439 (struct branch_predictor): Change type to signed integer.
3440 (test_prediction_value_range): Amend test to cover
3442 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
3443 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
3444 (PRED_LOOP_ITERATIONS_MAX): Likewise.
3445 (PRED_LOOP_IV_COMPARE): Likewise.
3446 * predict.h (PROB_UNINITIALIZED): Define new constant.
3448 2018-01-19 Martin Liska <mliska@suse.cz>
3450 * predict.c (dump_prediction): Add new format for
3451 analyze_brprob.py script which is enabled with -details
3453 * profile-count.h (precise_p): New function.
3455 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3457 PR tree-optimization/83922
3458 * tree-vect-loop.c (vect_verify_full_masking): Return false if
3459 there are no statements that need masking.
3460 (vect_active_double_reduction_p): New function.
3461 (vect_analyze_loop_operations): Use it when handling phis that
3462 are not in the loop header.
3464 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3466 PR tree-optimization/83914
3467 * tree-vect-loop.c (vectorizable_induction): Don't convert
3468 init_expr or apply the peeling adjustment for inductions
3469 that are nested within the vectorized loop.
3471 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3473 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
3476 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3480 * function.h (gimplify_parameters): Add gimple_seq * argument.
3481 * function.c: Include gimple.h and options.h.
3482 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
3483 for the added local temporaries if needed.
3484 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
3485 if there are any parameter cleanups, wrap whole body into a
3486 try/finally with the cleanups.
3488 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
3491 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
3492 Use GET_MODE_CLASS for scalar floating point.
3494 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3498 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
3499 Fix call of call_cgraph_insertion_hooks.
3501 2018-01-18 Martin Sebor <msebor@redhat.com>
3503 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
3505 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3508 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
3511 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
3514 * common.opt: (-ffile-prefix-map): New option.
3515 * opts.c (common_handle_option): Defer it.
3516 * opts-global.c (handle_common_deferred_options): Handle it.
3517 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
3518 * file-prefix-map.h: New file.
3519 (remap_debug_filename, add_debug_prefix_map): ...here.
3520 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
3521 * final.c (debug_prefix_map, add_debug_prefix_map
3522 remap_debug_filename): Move to...
3523 * file-prefix-map.c: New file.
3524 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
3525 generalize, get rid of alloca(), use strrchr() instead of strchr().
3526 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
3527 Implement in terms of add_prefix_map().
3528 (remap_macro_filename, remap_debug_filename): Implement in term of
3530 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
3531 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
3532 * dbxout.c: Include file-prefix-map.h.
3533 * varasm.c: Likewise.
3534 * vmsdbgout.c: Likewise.
3535 * xcoffout.c: Likewise.
3536 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
3537 * doc/cppopts.texi (-fmacro-prefix-map): Document.
3538 * doc/invoke.texi (-ffile-prefix-map): Document.
3539 (-fdebug-prefix-map): Update description.
3541 2018-01-18 Martin Liska <mliska@suse.cz>
3543 * config/i386/i386.c (indirect_thunk_name): Document that also
3545 (output_indirect_thunk): Document why both instructions
3546 (pause and lfence) are generated.
3548 2018-01-18 Richard Biener <rguenther@suse.de>
3550 PR tree-optimization/83887
3551 * graphite-scop-detection.c
3552 (scop_detection::get_nearest_dom_with_single_entry): Remove.
3553 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
3554 (scop_detection::merge_sese): Re-implement with a flood-fill
3555 algorithm that properly finds a SESE region if it exists.
3557 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3560 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
3561 pointer_diff optimizations use view_convert instead of convert.
3563 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3565 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3566 Generate different code for -mno-speculate-indirect-jumps.
3567 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3568 (*call_indirect_aix<mode>): Disable for
3569 -mno-speculate-indirect-jumps.
3570 (*call_indirect_aix<mode>_nospec): New define_insn.
3571 (*call_value_indirect_aix<mode>): Disable for
3572 -mno-speculate-indirect-jumps.
3573 (*call_value_indirect_aix<mode>_nospec): New define_insn.
3574 (*sibcall_nonlocal_sysv<mode>): Generate different code for
3575 -mno-speculate-indirect-jumps.
3576 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3578 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3580 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
3581 long double type, set the flags for noting the default long double
3582 type, even if we don't pass or return a long double type.
3584 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
3587 * ipa-inline.c (flatten_function): Do not overwrite final inlining
3590 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
3592 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
3593 support for merge[hl].
3594 (fold_mergehl_helper): New helper function.
3595 (tree-vector-builder.h): New #include for tree_vector_builder usage.
3596 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
3597 (altivec_vmrglw_direct): Add xxmrglw insn.
3599 2018-01-17 Andrew Waterman <andrew@sifive.com>
3601 * config/riscv/riscv.c (riscv_conditional_register_usage): If
3602 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
3604 2018-01-17 David Malcolm <dmalcolm@redhat.com>
3607 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
3608 call the lto_location_cache before reading the
3609 DECL_SOURCE_LOCATION of the types.
3611 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3612 Richard Sandiford <richard.sandiford@linaro.org>
3614 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
3615 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
3616 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
3617 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
3618 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
3620 * config/aarch64/constraints.md (aarch64_movti_operand):
3622 * config/aarch64/predicates.md (Uti): Add new constraint.
3624 2018-01-17 Carl Love <cel@us.ibm.com>
3626 * config/rs6000/vsx.md (define_expand xl_len_r,
3627 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
3628 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
3630 (define_expand, define_insn): Move the shift left from the
3631 define_insn to the define_expand for lxvl and stxvl instructions.
3632 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
3633 and XL_LEN_R definitions to PURE.
3635 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
3637 * config/i386/i386.c (indirect_thunk_name): Declare regno
3638 as unsigned int. Compare regno with INVALID_REGNUM.
3639 (output_indirect_thunk): Ditto.
3640 (output_indirect_thunk_function): Ditto.
3641 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
3642 in the call to output_indirect_thunk_function.
3644 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
3647 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
3648 rather than the size of inner_type to determine the stack slot size
3649 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
3651 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
3654 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
3657 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
3659 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
3660 endian Linux systems to optionally enable multilibs for selecting
3661 the long double type if the user configured an explicit type.
3662 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
3663 have no long double multilibs if not defined.
3664 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
3665 warn if the user used -mabi={ieee,ibm}longdouble and we built
3666 multilibs for long double.
3667 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
3668 appropriate multilib option.
3669 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
3671 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
3672 for building long double multilibs.
3673 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
3675 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
3677 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
3680 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
3682 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
3685 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
3688 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
3691 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3693 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
3694 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
3696 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
3698 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
3699 different rtl trees depending on TARGET_64BIT.
3700 (rs6000_gen_lvx): Likewise.
3702 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3704 * config/visium/visium.md (nop): Tweak comment.
3705 (hazard_nop): Likewise.
3707 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3709 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
3710 -mspeculate-indirect-jumps.
3711 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
3712 for -mno-speculate-indirect-jumps.
3713 (*call_indirect_elfv2<mode>_nospec): New define_insn.
3714 (*call_value_indirect_elfv2<mode>): Disable for
3715 -mno-speculate-indirect-jumps.
3716 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
3717 (indirect_jump): Emit different RTL for
3718 -mno-speculate-indirect-jumps.
3719 (*indirect_jump<mode>): Disable for
3720 -mno-speculate-indirect-jumps.
3721 (*indirect_jump<mode>_nospec): New define_insn.
3722 (tablejump): Emit different RTL for
3723 -mno-speculate-indirect-jumps.
3724 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
3725 (tablejumpsi_nospec): New define_expand.
3726 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
3727 (tablejumpdi_nospec): New define_expand.
3728 (*tablejump<mode>_internal1): Disable for
3729 -mno-speculate-indirect-jumps.
3730 (*tablejump<mode>_internal1_nospec): New define_insn.
3731 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
3734 2018-01-16 Artyom Skrobov tyomitch@gmail.com
3736 * caller-save.c (insert_save): Drop unnecessary parameter. All
3739 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3740 Richard Biener <rguenth@suse.de>
3743 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
3744 return early, inline manually is_gimple_sizepos. Make sure if we
3745 call gimplify_expr we don't end up with a gimple constant.
3746 * tree.c (variably_modified_type_p): Don't return true for
3747 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
3748 * gimplify.h (is_gimple_sizepos): Remove.
3750 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3752 PR tree-optimization/83857
3753 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
3754 vectorizable_live_operation for pure SLP statements.
3755 (vectorizable_live_operation): Handle PHIs.
3757 2018-01-16 Richard Biener <rguenther@suse.de>
3759 PR tree-optimization/83867
3760 * tree-vect-stmts.c (vect_transform_stmt): Precompute
3761 nested_in_vect_loop_p since the scalar stmt may get invalidated.
3763 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3766 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
3767 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
3768 If off is not INTEGER_CST, issue a may not be aligned warning
3769 rather than isn't aligned. Use isn%'t rather than isn't.
3770 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
3772 <case MULT_EXPR>: Improve the case when bottom and one of the
3773 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
3774 operand, in that case check if the other operand is multiple of
3775 bottom divided by the INTEGER_CST operand.
3777 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3780 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
3781 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
3782 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
3783 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
3784 * config/pa/pa.c (pa_function_arg_advance): Likewise.
3785 (pa_function_arg, pa_arg_partial_bytes): Likewise.
3786 (pa_function_arg_size): New function.
3788 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3790 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
3791 in a separate statement.
3793 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3795 PR tree-optimization/83847
3796 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
3797 group gathers and scatters.
3799 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3801 PR rtl-optimization/86620
3802 * params.def (max-sched-ready-insns): Bump minimum value to 1.
3804 PR rtl-optimization/83213
3805 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
3806 to last if both are JUMP_INSNs.
3808 PR tree-optimization/83843
3809 * gimple-ssa-store-merging.c
3810 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
3811 store_immediate_info for bswap/nop orig_stores.
3813 2018-01-15 Andrew Waterman <andrew@sifive.com>
3815 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
3817 <UDIV>: Increase cost if !TARGET_DIV.
3819 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
3821 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
3822 (define_attr "cr_logical_3op"): New.
3823 (cceq_ior_compare): Adjust.
3824 (cceq_ior_compare_complement): Adjust.
3825 (*cceq_rev_compare): Adjust.
3826 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
3827 (is_cracked_insn): Adjust.
3828 (insn_must_be_first_in_group): Adjust.
3829 * config/rs6000/40x.md: Adjust.
3830 * config/rs6000/440.md: Adjust.
3831 * config/rs6000/476.md: Adjust.
3832 * config/rs6000/601.md: Adjust.
3833 * config/rs6000/603.md: Adjust.
3834 * config/rs6000/6xx.md: Adjust.
3835 * config/rs6000/7450.md: Adjust.
3836 * config/rs6000/7xx.md: Adjust.
3837 * config/rs6000/8540.md: Adjust.
3838 * config/rs6000/cell.md: Adjust.
3839 * config/rs6000/e300c2c3.md: Adjust.
3840 * config/rs6000/e500mc.md: Adjust.
3841 * config/rs6000/e500mc64.md: Adjust.
3842 * config/rs6000/e5500.md: Adjust.
3843 * config/rs6000/e6500.md: Adjust.
3844 * config/rs6000/mpc.md: Adjust.
3845 * config/rs6000/power4.md: Adjust.
3846 * config/rs6000/power5.md: Adjust.
3847 * config/rs6000/power6.md: Adjust.
3848 * config/rs6000/power7.md: Adjust.
3849 * config/rs6000/power8.md: Adjust.
3850 * config/rs6000/power9.md: Adjust.
3851 * config/rs6000/rs64.md: Adjust.
3852 * config/rs6000/titan.md: Adjust.
3854 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3856 * config/i386/predicates.md (indirect_branch_operand): Rewrite
3857 ix86_indirect_branch_register logic.
3859 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3861 * config/i386/constraints.md (Bs): Update
3862 ix86_indirect_branch_register check. Don't check
3863 ix86_indirect_branch_register with GOT_memory_operand.
3865 * config/i386/predicates.md (GOT_memory_operand): Don't check
3866 ix86_indirect_branch_register here.
3867 (GOT32_symbol_operand): Likewise.
3869 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3871 * config/i386/predicates.md (constant_call_address_operand):
3872 Rewrite ix86_indirect_branch_register logic.
3873 (sibcall_insn_operand): Likewise.
3875 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3877 * config/i386/constraints.md (Bs): Replace
3878 ix86_indirect_branch_thunk_register with
3879 ix86_indirect_branch_register.
3881 * config/i386/i386.md (indirect_jump): Likewise.
3882 (tablejump): Likewise.
3883 (*sibcall_memory): Likewise.
3884 (*sibcall_value_memory): Likewise.
3885 Peepholes of indirect call and jump via memory: Likewise.
3886 * config/i386/i386.opt: Likewise.
3887 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3888 (GOT_memory_operand): Likewise.
3889 (call_insn_operand): Likewise.
3890 (sibcall_insn_operand): Likewise.
3891 (GOT32_symbol_operand): Likewise.
3893 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3896 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
3897 type rather than type addr's type points to.
3898 (expand_omp_atomic_mutex): Likewise.
3899 (expand_omp_atomic): Likewise.
3901 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3904 * config/i386/i386.c (output_indirect_thunk_function): Use
3905 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
3906 for __x86_return_thunk.
3908 2018-01-15 Richard Biener <rguenther@suse.de>
3911 * expmed.c (extract_bit_field_1): Fix typo.
3913 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3916 * config/arm/iterators.md (VF): New mode iterator.
3917 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
3918 Remove integer-related logic from pattern.
3919 (neon_vabd<mode>_3): Likewise.
3921 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3924 * common.opt (fstrict-overflow): No longer an alias.
3925 (fwrapv-pointer): New option.
3926 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
3927 also for pointer types based on flag_wrapv_pointer.
3928 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
3929 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
3930 opts->x_flag_wrapv got set.
3931 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
3932 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
3933 POINTER_TYPE_OVERFLOW_UNDEFINED.
3934 * match.pd: Likewise in address comparison pattern.
3935 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
3937 2018-01-15 Richard Biener <rguenther@suse.de>
3940 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
3941 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
3942 Reset type names to their identifier if their TYPE_DECL doesn't
3943 have linkage (and thus is used for ODR and devirt).
3944 (save_debug_info_for_decl): Remove.
3945 (save_debug_info_for_type): Likewise.
3946 (add_tree_to_fld_list): Adjust.
3947 * tree-pretty-print.c (dump_generic_node): Make dumping of
3948 type names more robust.
3950 2018-01-15 Richard Biener <rguenther@suse.de>
3952 * BASE-VER: Bump to 8.0.1.
3954 2018-01-14 Martin Sebor <msebor@redhat.com>
3957 * builtins.c (check_access): Avoid warning when the no-warning bit
3960 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
3962 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
3963 * ira-color (allocno_hard_regs_compare): Likewise.
3965 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
3968 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
3969 Use .pushsection/.popsection.
3971 2018-01-14 Martin Sebor <msebor@redhat.com>
3974 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
3976 2018-01-14 Jakub Jelinek <jakub@redhat.com>
3978 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
3979 entry from extra_headers.
3980 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
3981 extra_headers, make the list bitwise identical to the i?86-*-* one.
3983 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3985 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
3986 -mcmodel=large with -mindirect-branch=thunk,
3987 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
3988 -mfunction-return=thunk-extern.
3989 * doc/invoke.texi: Document -mcmodel=large is incompatible with
3990 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
3991 -mfunction-return=thunk and -mfunction-return=thunk-extern.
3993 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3995 * config/i386/i386.c (print_reg): Print the name of the full
3996 integer register without '%'.
3997 (ix86_print_operand): Handle 'V'.
3998 * doc/extend.texi: Document 'V' modifier.
4000 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4002 * config/i386/constraints.md (Bs): Disallow memory operand for
4003 -mindirect-branch-register.
4005 * config/i386/predicates.md (indirect_branch_operand): Likewise.
4006 (GOT_memory_operand): Likewise.
4007 (call_insn_operand): Likewise.
4008 (sibcall_insn_operand): Likewise.
4009 (GOT32_symbol_operand): Likewise.
4010 * config/i386/i386.md (indirect_jump): Call convert_memory_address
4011 for -mindirect-branch-register.
4012 (tablejump): Likewise.
4013 (*sibcall_memory): Likewise.
4014 (*sibcall_value_memory): Likewise.
4015 Disallow peepholes of indirect call and jump via memory for
4016 -mindirect-branch-register.
4017 (*call_pop): Replace m with Bw.
4018 (*call_value_pop): Likewise.
4019 (*sibcall_pop_memory): Replace m with Bs.
4020 * config/i386/i386.opt (mindirect-branch-register): New option.
4021 * doc/invoke.texi: Document -mindirect-branch-register option.
4023 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4025 * config/i386/i386-protos.h (ix86_output_function_return): New.
4026 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
4027 set function_return_type.
4028 (indirect_thunk_name): Add ret_p to indicate thunk for function
4030 (output_indirect_thunk_function): Pass false to
4031 indirect_thunk_name.
4032 (ix86_output_indirect_branch_via_reg): Likewise.
4033 (ix86_output_indirect_branch_via_push): Likewise.
4034 (output_indirect_thunk_function): Create alias for function
4035 return thunk if regno < 0.
4036 (ix86_output_function_return): New function.
4037 (ix86_handle_fndecl_attribute): Handle function_return.
4038 (ix86_attribute_table): Add function_return.
4039 * config/i386/i386.h (machine_function): Add
4040 function_return_type.
4041 * config/i386/i386.md (simple_return_internal): Use
4042 ix86_output_function_return.
4043 (simple_return_internal_long): Likewise.
4044 * config/i386/i386.opt (mfunction-return=): New option.
4045 (indirect_branch): Mention -mfunction-return=.
4046 * doc/extend.texi: Document function_return function attribute.
4047 * doc/invoke.texi: Document -mfunction-return= option.
4049 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4051 * config/i386/i386-opts.h (indirect_branch): New.
4052 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
4053 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
4054 with local indirect jump when converting indirect call and jump.
4055 (ix86_set_indirect_branch_type): New.
4056 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
4057 (indirectlabelno): New.
4058 (indirect_thunk_needed): Likewise.
4059 (indirect_thunk_bnd_needed): Likewise.
4060 (indirect_thunks_used): Likewise.
4061 (indirect_thunks_bnd_used): Likewise.
4062 (INDIRECT_LABEL): Likewise.
4063 (indirect_thunk_name): Likewise.
4064 (output_indirect_thunk): Likewise.
4065 (output_indirect_thunk_function): Likewise.
4066 (ix86_output_indirect_branch_via_reg): Likewise.
4067 (ix86_output_indirect_branch_via_push): Likewise.
4068 (ix86_output_indirect_branch): Likewise.
4069 (ix86_output_indirect_jmp): Likewise.
4070 (ix86_code_end): Call output_indirect_thunk_function if needed.
4071 (ix86_output_call_insn): Call ix86_output_indirect_branch if
4073 (ix86_handle_fndecl_attribute): Handle indirect_branch.
4074 (ix86_attribute_table): Add indirect_branch.
4075 * config/i386/i386.h (machine_function): Add indirect_branch_type
4076 and has_local_indirect_jump.
4077 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
4079 (tablejump): Likewise.
4080 (*indirect_jump): Use ix86_output_indirect_jmp.
4081 (*tablejump_1): Likewise.
4082 (simple_return_indirect_internal): Likewise.
4083 * config/i386/i386.opt (mindirect-branch=): New option.
4084 (indirect_branch): New.
4087 (thunk-inline): Likewise.
4088 (thunk-extern): Likewise.
4089 * doc/extend.texi: Document indirect_branch function attribute.
4090 * doc/invoke.texi: Document -mindirect-branch= option.
4092 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
4095 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
4097 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
4099 * ipa-inline.c (want_inline_small_function_p): Return false if
4100 inlining has already failed with CIF_FINAL_ERROR.
4101 (update_caller_keys): Call want_inline_small_function_p before
4103 (update_callee_keys): Likewise.
4105 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4107 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
4109 (rs6000_quadword_masked_address_p): Likewise.
4110 (quad_aligned_load_p): Likewise.
4111 (quad_aligned_store_p): Likewise.
4112 (const_load_sequence_p): Add comment to describe the outer-most loop.
4113 (mimic_memory_attributes_and_flags): New function.
4114 (rs6000_gen_stvx): Likewise.
4115 (replace_swapped_aligned_store): Likewise.
4116 (rs6000_gen_lvx): Likewise.
4117 (replace_swapped_aligned_load): Likewise.
4118 (replace_swapped_load_constant): Capitalize argument name in
4119 comment describing this function.
4120 (rs6000_analyze_swaps): Add a third pass to search for vector loads
4121 and stores that access quad-word aligned addresses and replace
4122 with stvx or lvx instructions when appropriate.
4123 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
4124 New function prototype.
4125 (rs6000_quadword_masked_address_p): Likewise.
4126 (rs6000_gen_lvx): Likewise.
4127 (rs6000_gen_stvx): Likewise.
4128 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
4129 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
4130 when memory address is aligned.
4131 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
4132 this split to select lvx instruction when memory address is aligned.
4133 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
4134 instruction when memory address is aligned.
4135 (*vsx_le_perm_load_v16qi): Likewise.
4136 (four unnamed splitters): Modify to select the stvx instruction
4137 when memory is aligned.
4139 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
4141 * predict.c (determine_unlikely_bbs): Handle correctly BBs
4142 which appears in the queue multiple times.
4144 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4145 Alan Hayward <alan.hayward@arm.com>
4146 David Sherwood <david.sherwood@arm.com>
4148 * tree-vectorizer.h (vec_lower_bound): New structure.
4149 (_loop_vec_info): Add check_nonzero and lower_bounds.
4150 (LOOP_VINFO_CHECK_NONZERO): New macro.
4151 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
4152 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
4153 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
4154 fields. Make seg_len the distance travelled, not including the
4156 (dr_direction_indicator): Declare.
4157 (dr_zero_step_indicator): Likewise.
4158 (dr_known_forward_stride_p): Likewise.
4159 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
4161 (runtime_alias_check_p): Allow runtime alias checks with
4163 (operator ==): Compare access_size and align.
4164 (prune_runtime_alias_test_list): Rework for new distinction between
4165 the access_size and seg_len.
4166 (create_intersect_range_checks_index): Likewise. Cope with polynomial
4168 (get_segment_min_max): New function.
4169 (create_intersect_range_checks): Use it.
4170 (dr_step_indicator): New function.
4171 (dr_direction_indicator): Likewise.
4172 (dr_zero_step_indicator): Likewise.
4173 (dr_known_forward_stride_p): Likewise.
4174 * tree-loop-distribution.c (data_ref_segment_size): Return
4175 DR_STEP * (niters - 1).
4176 (compute_alias_check_pairs): Update call to the dr_with_seg_len
4178 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
4179 (vect_preserves_scalar_order_p): New function, split out from...
4180 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
4181 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
4182 (vect_vfa_access_size): New function.
4183 (vect_vfa_align): Likewise.
4184 (vect_compile_time_alias): Take access_size_a and access_b arguments.
4185 (dump_lower_bound): New function.
4186 (vect_check_lower_bound): Likewise.
4187 (vect_small_gap_p): Likewise.
4188 (vectorizable_with_step_bound_p): Likewise.
4189 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
4190 depencies if the vectorization factor is 1. Convert the checks
4191 for nonzero steps into checks on the bounds of DR_STEP. Try using
4192 a bunds check for variable steps if the minimum required step is
4193 relatively small. Update calls to the dr_with_seg_len
4194 constructor and to vect_compile_time_alias.
4195 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
4197 (vect_loop_versioning): Call it.
4198 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
4200 (vect_estimate_min_profitable_iters): Account for any bounds checks.
4202 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4203 Alan Hayward <alan.hayward@arm.com>
4204 David Sherwood <david.sherwood@arm.com>
4206 * doc/sourcebuild.texi (vect_scatter_store): Document.
4207 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
4209 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
4211 * genopinit.c (main): Add supports_vec_scatter_store and
4212 supports_vec_scatter_store_cached to target_optabs.
4213 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
4214 IFN_MASK_SCATTER_STORE.
4215 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
4217 * internal-fn.h (internal_store_fn_p): Declare.
4218 (internal_fn_stored_value_index): Likewise.
4219 * internal-fn.c (scatter_store_direct): New macro.
4220 (expand_scatter_store_optab_fn): New function.
4221 (direct_scatter_store_optab_supported_p): New macro.
4222 (internal_store_fn_p): New function.
4223 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
4224 IFN_MASK_SCATTER_STORE.
4225 (internal_fn_mask_index): Likewise.
4226 (internal_fn_stored_value_index): New function.
4227 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
4229 * optabs-query.h (supports_vec_scatter_store_p): Declare.
4230 * optabs-query.c (supports_vec_scatter_store_p): New function.
4231 * tree-vectorizer.h (vect_get_store_rhs): Declare.
4232 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
4233 true for scatter stores.
4234 (vect_gather_scatter_fn_p): Handle scatter stores too.
4235 (vect_check_gather_scatter): Consider using scatter stores if
4236 supports_vec_scatter_store_p.
4237 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
4239 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4240 internal_fn_stored_value_index.
4241 (check_load_store_masking): Handle scatter stores too.
4242 (vect_get_store_rhs): Make public.
4243 (vectorizable_call): Use internal_store_fn_p.
4244 (vectorizable_store): Handle scatter store internal functions.
4245 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
4246 when deciding whether the end of the group has been reached.
4247 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
4248 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
4249 (mask_scatter_store<mode>): New insns.
4251 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4252 Alan Hayward <alan.hayward@arm.com>
4253 David Sherwood <david.sherwood@arm.com>
4255 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
4256 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
4257 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
4259 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
4260 Use vect_truncate_gather_scatter_offset if we can't treat the
4261 operation as a normal gather load or scatter store.
4262 (get_group_load_store_type): Take the gather_scatter_info
4263 as argument. Try using a gather load or scatter store for
4264 single-element groups.
4265 (get_load_store_type): Update calls to get_group_load_store_type
4266 and vect_use_strided_gather_scatters_p.
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_create_data_ref_ptr): Take an extra
4273 optional tree argument.
4274 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
4276 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
4277 but continue to use the current value as a fallback.
4278 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
4279 to compare the updates.
4280 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
4281 (get_load_store_type): Use it when handling a strided access.
4282 (vect_get_strided_load_store_ops): New function.
4283 (vect_get_data_ptr_increment): Likewise.
4284 (vectorizable_load): Handle strided gather loads. Always pass
4285 a step to vect_create_data_ref_ptr and bump_vector_ptr.
4287 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4288 Alan Hayward <alan.hayward@arm.com>
4289 David Sherwood <david.sherwood@arm.com>
4291 * doc/md.texi (gather_load@var{m}): Document.
4292 (mask_gather_load@var{m}): Likewise.
4293 * genopinit.c (main): Add supports_vec_gather_load and
4294 supports_vec_gather_load_cached to target_optabs.
4295 * optabs-tree.c (init_tree_optimization_optabs): Use
4296 ggc_cleared_alloc to allocate target_optabs.
4297 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
4298 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
4300 * internal-fn.h (internal_load_fn_p): Declare.
4301 (internal_gather_scatter_fn_p): Likewise.
4302 (internal_fn_mask_index): Likewise.
4303 (internal_gather_scatter_fn_supported_p): Likewise.
4304 * internal-fn.c (gather_load_direct): New macro.
4305 (expand_gather_load_optab_fn): New function.
4306 (direct_gather_load_optab_supported_p): New macro.
4307 (direct_internal_fn_optab): New function.
4308 (internal_load_fn_p): Likewise.
4309 (internal_gather_scatter_fn_p): Likewise.
4310 (internal_fn_mask_index): Likewise.
4311 (internal_gather_scatter_fn_supported_p): Likewise.
4312 * optabs-query.c (supports_at_least_one_mode_p): New function.
4313 (supports_vec_gather_load_p): Likewise.
4314 * optabs-query.h (supports_vec_gather_load_p): Declare.
4315 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
4316 and memory_type field.
4317 (NUM_PATTERNS): Bump to 15.
4318 * tree-vect-data-refs.c: Include internal-fn.h.
4319 (vect_gather_scatter_fn_p): New function.
4320 (vect_describe_gather_scatter_call): Likewise.
4321 (vect_check_gather_scatter): Try using internal functions for
4322 gather loads. Recognize existing calls to a gather load function.
4323 (vect_analyze_data_refs): Consider using gather loads if
4324 supports_vec_gather_load_p.
4325 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
4326 (vect_get_gather_scatter_offset_type): Likewise.
4327 (vect_convert_mask_for_vectype): Likewise.
4328 (vect_add_conversion_to_patterm): Likewise.
4329 (vect_try_gather_scatter_pattern): Likewise.
4330 (vect_recog_gather_scatter_pattern): New pattern recognizer.
4331 (vect_vect_recog_func_ptrs): Add it.
4332 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4333 internal_fn_mask_index and internal_gather_scatter_fn_p.
4334 (check_load_store_masking): Take the gather_scatter_info as an
4335 argument and handle gather loads.
4336 (vect_get_gather_scatter_ops): New function.
4337 (vectorizable_call): Check internal_load_fn_p.
4338 (vectorizable_load): Likewise. Handle gather load internal
4340 (vectorizable_store): Update call to check_load_store_masking.
4341 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
4342 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
4343 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
4344 (aarch64_gather_scale_operand_d): New predicates.
4345 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
4346 (mask_gather_load<mode>): New insns.
4348 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4349 Alan Hayward <alan.hayward@arm.com>
4350 David Sherwood <david.sherwood@arm.com>
4352 * optabs.def (fold_left_plus_optab): New optab.
4353 * doc/md.texi (fold_left_plus_@var{m}): Document.
4354 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
4355 * internal-fn.c (fold_left_direct): Define.
4356 (expand_fold_left_optab_fn): Likewise.
4357 (direct_fold_left_optab_supported_p): Likewise.
4358 * fold-const-call.c (fold_const_fold_left): New function.
4359 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
4360 * tree-parloops.c (valid_reduction_p): New function.
4361 (gather_scalar_reductions): Use it.
4362 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
4363 (vect_finish_replace_stmt): Declare.
4364 * tree-vect-loop.c (fold_left_reduction_fn): New function.
4365 (needs_fold_left_reduction_p): New function, split out from...
4366 (vect_is_simple_reduction): ...here. Accept reductions that
4367 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
4368 (vect_force_simple_reduction): Also store the reduction type in
4369 the assignment's STMT_VINFO_REDUC_TYPE.
4370 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
4371 (merge_with_identity): New function.
4372 (vect_expand_fold_left): Likewise.
4373 (vectorize_fold_left_reduction): Likewise.
4374 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
4375 scalar phi in place for it. Check for target support and reject
4376 cases that would reassociate the operation. Defer the transform
4377 phase to vectorize_fold_left_reduction.
4378 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
4379 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
4380 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
4382 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4384 * tree-if-conv.c (predicate_mem_writes): Remove redundant
4385 call to ifc_temp_var.
4387 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4388 Alan Hayward <alan.hayward@arm.com>
4389 David Sherwood <david.sherwood@arm.com>
4391 * target.def (legitimize_address_displacement): Take the original
4392 offset as a poly_int.
4393 * targhooks.h (default_legitimize_address_displacement): Update
4395 * targhooks.c (default_legitimize_address_displacement): Likewise.
4396 * doc/tm.texi: Regenerate.
4397 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
4398 as an argument, moving assert of ad->disp == ad->disp_term to...
4399 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
4400 Try calling targetm.legitimize_address_displacement before expanding
4401 the address rather than afterwards, and adjust for the new interface.
4402 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
4403 Match the new hook interface. Handle SVE addresses.
4404 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
4407 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4409 * Makefile.in (OBJS): Add early-remat.o.
4410 * target.def (select_early_remat_modes): New hook.
4411 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
4412 * doc/tm.texi: Regenerate.
4413 * targhooks.h (default_select_early_remat_modes): Declare.
4414 * targhooks.c (default_select_early_remat_modes): New function.
4415 * timevar.def (TV_EARLY_REMAT): New timevar.
4416 * passes.def (pass_early_remat): New pass.
4417 * tree-pass.h (make_pass_early_remat): Declare.
4418 * early-remat.c: New file.
4419 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
4421 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
4423 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4424 Alan Hayward <alan.hayward@arm.com>
4425 David Sherwood <david.sherwood@arm.com>
4427 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
4428 vfm1 with a bound_epilog parameter.
4429 (vect_do_peeling): Update calls accordingly, and move the prologue
4430 call earlier in the function. Treat the base bound_epilog as 0 for
4431 fully-masked loops and retain vf - 1 for other loops. Add 1 to
4432 this base when peeling for gaps.
4433 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
4434 with fully-masked loops.
4435 (vect_estimate_min_profitable_iters): Handle the single peeled
4436 iteration in that case.
4438 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4439 Alan Hayward <alan.hayward@arm.com>
4440 David Sherwood <david.sherwood@arm.com>
4442 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
4443 single-element interleaving even if the size is not a power of 2.
4444 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
4445 accesses for single-element interleaving if the group size is
4448 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4449 Alan Hayward <alan.hayward@arm.com>
4450 David Sherwood <david.sherwood@arm.com>
4452 * doc/md.texi (fold_extract_last_@var{m}): Document.
4453 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
4454 * optabs.def (fold_extract_last_optab): New optab.
4455 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
4456 * internal-fn.c (fold_extract_direct): New macro.
4457 (expand_fold_extract_optab_fn): Likewise.
4458 (direct_fold_extract_optab_supported_p): Likewise.
4459 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
4460 * tree-vect-loop.c (vect_model_reduction_cost): Handle
4461 EXTRACT_LAST_REDUCTION.
4462 (get_initial_def_for_reduction): Do not create an initial vector
4463 for EXTRACT_LAST_REDUCTION reductions.
4464 (vectorizable_reduction): Leave the scalar phi in place for
4465 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
4466 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
4467 epilogue code for EXTRACT_LAST_REDUCTION and defer the
4468 transform phase to vectorizable_condition.
4469 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
4471 (vect_finish_stmt_generation): ...here.
4472 (vect_finish_replace_stmt): New function.
4473 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
4474 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
4476 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
4478 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4479 Alan Hayward <alan.hayward@arm.com>
4480 David Sherwood <david.sherwood@arm.com>
4482 * doc/md.texi (extract_last_@var{m}): Document.
4483 * optabs.def (extract_last_optab): New optab.
4484 * internal-fn.def (EXTRACT_LAST): New internal function.
4485 * internal-fn.c (cond_unary_direct): New macro.
4486 (expand_cond_unary_optab_fn): Likewise.
4487 (direct_cond_unary_optab_supported_p): Likewise.
4488 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
4489 loops using EXTRACT_LAST.
4490 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
4491 (extract_last_<mode>): ...this optab.
4492 (vec_extract<mode><Vel>): Update accordingly.
4494 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4495 Alan Hayward <alan.hayward@arm.com>
4496 David Sherwood <david.sherwood@arm.com>
4498 * target.def (empty_mask_is_expensive): New hook.
4499 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
4500 * doc/tm.texi: Regenerate.
4501 * targhooks.h (default_empty_mask_is_expensive): Declare.
4502 * targhooks.c (default_empty_mask_is_expensive): New function.
4503 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
4504 if the target says that empty masks are expensive.
4505 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
4507 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
4509 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4510 Alan Hayward <alan.hayward@arm.com>
4511 David Sherwood <david.sherwood@arm.com>
4513 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
4514 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
4515 (vect_use_loop_mask_for_alignment_p): New function.
4516 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
4517 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
4518 niters_skip argument. Make sure that the first niters_skip elements
4519 of the first iteration are inactive.
4520 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
4521 Update call to vect_set_loop_masks_directly.
4522 (get_misalign_in_elems): New function, split out from...
4523 (vect_gen_prolog_loop_niters): ...here.
4524 (vect_update_init_of_dr): Take a code argument that specifies whether
4525 the adjustment should be added or subtracted.
4526 (vect_update_init_of_drs): Likewise.
4527 (vect_prepare_for_masked_peels): New function.
4528 (vect_do_peeling): Skip prologue peeling if we're using a mask
4529 instead. Update call to vect_update_inits_of_drs.
4530 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4532 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
4533 alignment. Do not include the number of peeled iterations in
4534 the minimum threshold in that case.
4535 (vectorizable_induction): Adjust the start value down by
4536 LOOP_VINFO_MASK_SKIP_NITERS iterations.
4537 (vect_transform_loop): Call vect_prepare_for_masked_peels.
4538 Take the number of skipped iterations into account when calculating
4540 * tree-vect-stmts.c (vect_gen_while_not): New function.
4542 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4543 Alan Hayward <alan.hayward@arm.com>
4544 David Sherwood <david.sherwood@arm.com>
4546 * doc/sourcebuild.texi (vect_fully_masked): Document.
4547 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
4549 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
4551 (vect_analyze_loop_2): ...here. Don't check the vectorization
4552 factor against the number of loop iterations if the loop is
4555 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4556 Alan Hayward <alan.hayward@arm.com>
4557 David Sherwood <david.sherwood@arm.com>
4559 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
4560 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
4561 (dump_groups): Update accordingly.
4562 (iv_use::mem_type): New member variable.
4563 (address_p): New function.
4564 (record_use): Add a mem_type argument and initialize the new
4566 (record_group_use): Add a mem_type argument. Use address_p.
4567 Remove obsolete null checks of base_object. Update call to record_use.
4568 (find_interesting_uses_op): Update call to record_group_use.
4569 (find_interesting_uses_cond): Likewise.
4570 (find_interesting_uses_address): Likewise.
4571 (get_mem_type_for_internal_fn): New function.
4572 (find_address_like_use): Likewise.
4573 (find_interesting_uses_stmt): Try find_address_like_use before
4574 calling find_interesting_uses_op.
4575 (addr_offset_valid_p): Use the iv mem_type field as the type
4576 of the addressed memory.
4577 (add_autoinc_candidates): Likewise.
4578 (get_address_cost): Likewise.
4579 (split_small_address_groups_p): Use address_p.
4580 (split_address_groups): Likewise.
4581 (add_iv_candidate_for_use): Likewise.
4582 (autoinc_possible_for_pair): Likewise.
4583 (rewrite_groups): Likewise.
4584 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
4585 (determine_group_iv_cost): Update after split of USE_ADDRESS.
4586 (get_alias_ptr_type_for_ptr_address): New function.
4587 (rewrite_use_address): Rewrite address uses in calls that were
4588 identified by find_address_like_use.
4590 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4591 Alan Hayward <alan.hayward@arm.com>
4592 David Sherwood <david.sherwood@arm.com>
4594 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
4596 * gimple-expr.h (is_gimple_addressable: Likewise.
4597 * gimple-expr.c (is_gimple_address): Likewise.
4598 * internal-fn.c (expand_call_mem_ref): New function.
4599 (expand_mask_load_optab_fn): Use it.
4600 (expand_mask_store_optab_fn): Likewise.
4602 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4603 Alan Hayward <alan.hayward@arm.com>
4604 David Sherwood <david.sherwood@arm.com>
4606 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
4607 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
4608 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
4609 (cond_umax@var{mode}): Document.
4610 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
4611 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
4612 (cond_umin_optab, cond_umax_optab): New optabs.
4613 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
4614 (COND_IOR, COND_XOR): New internal functions.
4615 * internal-fn.h (get_conditional_internal_fn): Declare.
4616 * internal-fn.c (cond_binary_direct): New macro.
4617 (expand_cond_binary_optab_fn): Likewise.
4618 (direct_cond_binary_optab_supported_p): Likewise.
4619 (get_conditional_internal_fn): New function.
4620 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
4621 Cope with reduction statements that are vectorized as calls rather
4623 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
4624 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
4625 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
4626 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4627 (UNSPEC_COND_EOR): New unspecs.
4628 (optab): Add mappings for them.
4629 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
4630 (sve_int_op, sve_fp_op): New int attributes.
4632 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4633 Alan Hayward <alan.hayward@arm.com>
4634 David Sherwood <david.sherwood@arm.com>
4636 * optabs.def (while_ult_optab): New optab.
4637 * doc/md.texi (while_ult@var{m}@var{n}): Document.
4638 * internal-fn.def (WHILE_ULT): New internal function.
4639 * internal-fn.h (direct_internal_fn_supported_p): New override
4640 that takes two types as argument.
4641 * internal-fn.c (while_direct): New macro.
4642 (expand_while_optab_fn): New function.
4643 (convert_optab_supported_p): Likewise.
4644 (direct_while_optab_supported_p): New macro.
4645 * wide-int.h (wi::udiv_ceil): New function.
4646 * tree-vectorizer.h (rgroup_masks): New structure.
4647 (vec_loop_masks): New typedef.
4648 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
4650 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
4651 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
4652 (vect_max_vf): New function.
4653 (slpeel_make_loop_iterate_ntimes): Delete.
4654 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
4655 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
4656 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
4657 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
4658 internal-fn.h, stor-layout.h and optabs-query.h.
4659 (vect_set_loop_mask): New function.
4660 (add_preheader_seq): Likewise.
4661 (add_header_seq): Likewise.
4662 (interleave_supported_p): Likewise.
4663 (vect_maybe_permute_loop_masks): Likewise.
4664 (vect_set_loop_masks_directly): Likewise.
4665 (vect_set_loop_condition_masked): Likewise.
4666 (vect_set_loop_condition_unmasked): New function, split out from
4667 slpeel_make_loop_iterate_ntimes.
4668 (slpeel_make_loop_iterate_ntimes): Rename to..
4669 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
4670 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
4671 (vect_do_peeling): Update call accordingly.
4672 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
4674 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4675 mask_compare_type, can_fully_mask_p and fully_masked_p.
4676 (release_vec_loop_masks): New function.
4677 (_loop_vec_info): Use it to free the loop masks.
4678 (can_produce_all_loop_masks_p): New function.
4679 (vect_get_max_nscalars_per_iter): Likewise.
4680 (vect_verify_full_masking): Likewise.
4681 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
4682 retries, and free the mask rgroups before retrying. Check loop-wide
4683 reasons for disallowing fully-masked loops. Make the final decision
4684 about whether use a fully-masked loop or not.
4685 (vect_estimate_min_profitable_iters): Do not assume that peeling
4686 for the number of iterations will be needed for fully-masked loops.
4687 (vectorizable_reduction): Disable fully-masked loops.
4688 (vectorizable_live_operation): Likewise.
4689 (vect_halve_mask_nunits): New function.
4690 (vect_double_mask_nunits): Likewise.
4691 (vect_record_loop_mask): Likewise.
4692 (vect_get_loop_mask): Likewise.
4693 (vect_transform_loop): Handle the case in which the final loop
4694 iteration might handle a partial vector. Call vect_set_loop_condition
4695 instead of slpeel_make_loop_iterate_ntimes.
4696 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
4697 (check_load_store_masking): New function.
4698 (prepare_load_store_mask): Likewise.
4699 (vectorizable_store): Handle fully-masked loops.
4700 (vectorizable_load): Likewise.
4701 (supportable_widening_operation): Use vect_halve_mask_nunits for
4703 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
4704 (vect_gen_while): New function.
4705 * config/aarch64/aarch64.md (umax<mode>3): New expander.
4706 (aarch64_uqdec<mode>): New insn.
4708 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4709 Alan Hayward <alan.hayward@arm.com>
4710 David Sherwood <david.sherwood@arm.com>
4712 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
4713 (reduc_xor_scal_optab): New optabs.
4714 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
4715 (reduc_xor_scal_@var{m}): Document.
4716 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
4717 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
4719 * fold-const-call.c (fold_const_call): Handle them.
4720 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
4721 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
4722 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
4723 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
4724 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
4725 (UNSPEC_XORV): New unspecs.
4726 (optab): Add entries for them.
4727 (BITWISEV): New int iterator.
4728 (bit_reduc_op): New int attributes.
4730 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4731 Alan Hayward <alan.hayward@arm.com>
4732 David Sherwood <david.sherwood@arm.com>
4734 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
4735 * internal-fn.def (VEC_SHL_INSERT): New internal function.
4736 * optabs.def (vec_shl_insert_optab): New optab.
4737 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
4738 (duplicate_and_interleave): Likewise.
4739 * tree-vect-loop.c: Include internal-fn.h.
4740 (neutral_op_for_slp_reduction): New function, split out from
4741 get_initial_defs_for_reduction.
4742 (get_initial_def_for_reduction): Handle option 2 for variable-length
4743 vectors by loading the neutral value into a vector and then shifting
4744 the initial value into element 0.
4745 (get_initial_defs_for_reduction): Replace the code argument with
4746 the neutral value calculated by neutral_op_for_slp_reduction.
4747 Use gimple_build_vector for constant-length vectors.
4748 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
4749 but the first group_size elements have a neutral value.
4750 Use duplicate_and_interleave otherwise.
4751 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
4752 Update call to get_initial_defs_for_reduction. Handle SLP
4753 reductions for variable-length vectors by creating one vector
4754 result for each scalar result, with the elements associated
4755 with other scalar results stubbed out with the neutral value.
4756 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
4757 Require IFN_VEC_SHL_INSERT for double reductions on
4758 variable-length vectors, or SLP reductions that have
4759 a neutral value. Require can_duplicate_and_interleave_p
4760 support for variable-length unchained SLP reductions if there
4761 is no neutral value, such as for MIN/MAX reductions. Also require
4762 the number of vector elements to be a multiple of the number of
4763 SLP statements when doing variable-length unchained SLP reductions.
4764 Update call to vect_create_epilog_for_reduction.
4765 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
4766 and remove initial values.
4767 (duplicate_and_interleave): Make public.
4768 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
4769 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
4771 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4772 Alan Hayward <alan.hayward@arm.com>
4773 David Sherwood <david.sherwood@arm.com>
4775 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
4776 (can_duplicate_and_interleave_p): New function.
4777 (vect_get_and_check_slp_defs): Take the vector of statements
4778 rather than just the current one. Remove excess parentheses.
4779 Restriction rejectinon of vect_constant_def and vect_external_def
4780 for variable-length vectors to boolean types, or types for which
4781 can_duplicate_and_interleave_p is false.
4782 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
4783 (duplicate_and_interleave): New function.
4784 (vect_get_constant_vectors): Use gimple_build_vector for
4785 constant-length vectors and suitable variable-length constant
4786 vectors. Use duplicate_and_interleave for other variable-length
4787 vectors. Don't defer the update when inserting new statements.
4789 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4790 Alan Hayward <alan.hayward@arm.com>
4791 David Sherwood <david.sherwood@arm.com>
4793 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
4794 min_profitable_iters doesn't go negative.
4796 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4797 Alan Hayward <alan.hayward@arm.com>
4798 David Sherwood <david.sherwood@arm.com>
4800 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
4801 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
4802 * optabs.def (vec_mask_load_lanes_optab): New optab.
4803 (vec_mask_store_lanes_optab): Likewise.
4804 * internal-fn.def (MASK_LOAD_LANES): New internal function.
4805 (MASK_STORE_LANES): Likewise.
4806 * internal-fn.c (mask_load_lanes_direct): New macro.
4807 (mask_store_lanes_direct): Likewise.
4808 (expand_mask_load_optab_fn): Handle masked operations.
4809 (expand_mask_load_lanes_optab_fn): New macro.
4810 (expand_mask_store_optab_fn): Handle masked operations.
4811 (expand_mask_store_lanes_optab_fn): New macro.
4812 (direct_mask_load_lanes_optab_supported_p): Likewise.
4813 (direct_mask_store_lanes_optab_supported_p): Likewise.
4814 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
4816 (vect_load_lanes_supported): Likewise.
4817 * tree-vect-data-refs.c (strip_conversion): New function.
4818 (can_group_stmts_p): Likewise.
4819 (vect_analyze_data_ref_accesses): Use it instead of checking
4820 for a pair of assignments.
4821 (vect_store_lanes_supported): Take a masked_p parameter.
4822 (vect_load_lanes_supported): Likewise.
4823 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
4824 vect_store_lanes_supported and vect_load_lanes_supported.
4825 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4826 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
4827 parameter. Don't allow gaps for masked accesses.
4828 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
4829 and vect_load_lanes_supported.
4830 (get_load_store_type): Take a masked_p parameter and update
4831 call to get_group_load_store_type.
4832 (vectorizable_store): Update call to get_load_store_type.
4833 Handle IFN_MASK_STORE_LANES.
4834 (vectorizable_load): Update call to get_load_store_type.
4835 Handle IFN_MASK_LOAD_LANES.
4837 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4838 Alan Hayward <alan.hayward@arm.com>
4839 David Sherwood <david.sherwood@arm.com>
4841 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
4843 * config/aarch64/aarch64-protos.h
4844 (aarch64_sve_struct_memory_operand_p): Declare.
4845 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
4846 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
4847 (VPRED, vpred): Handle SVE structure modes.
4848 * config/aarch64/constraints.md (Utx): New constraint.
4849 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
4850 (aarch64_sve_struct_nonimmediate_operand): New predicates.
4851 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
4852 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
4853 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
4854 structure modes. Split into pieces after RA.
4855 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
4856 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
4858 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
4859 SVE structure modes.
4860 (aarch64_classify_address): Likewise.
4861 (sizetochar): Move earlier in file.
4862 (aarch64_print_operand): Handle SVE register lists.
4863 (aarch64_array_mode): New function.
4864 (aarch64_sve_struct_memory_operand_p): Likewise.
4865 (TARGET_ARRAY_MODE): Redefine.
4867 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4868 Alan Hayward <alan.hayward@arm.com>
4869 David Sherwood <david.sherwood@arm.com>
4871 * target.def (array_mode): New target hook.
4872 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
4873 * doc/tm.texi: Regenerate.
4874 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
4875 * hooks.c (hook_optmode_mode_uhwi_none): New function.
4876 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
4878 * stor-layout.c (mode_for_array): Likewise. Support polynomial
4881 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4882 Alan Hayward <alan.hayward@arm.com>
4883 David Sherwood <david.sherwood@arm.com>
4885 * fold-const.c (fold_binary_loc): Check the argument types
4886 rather than the result type when testing for a vector operation.
4888 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4890 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
4891 * doc/tm.texi: Regenerate.
4893 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4894 Alan Hayward <alan.hayward@arm.com>
4895 David Sherwood <david.sherwood@arm.com>
4897 * doc/invoke.texi (-msve-vector-bits=): Document new option.
4898 (sve): Document new AArch64 extension.
4899 * doc/md.texi (w): Extend the description of the AArch64
4900 constraint to include SVE vectors.
4901 (Upl, Upa): Document new AArch64 predicate constraints.
4902 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
4904 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
4905 (msve-vector-bits=): New option.
4906 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
4907 SVE when these are disabled.
4908 (sve): New extension.
4909 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
4910 modes. Adjust their number of units based on aarch64_sve_vg.
4911 (MAX_BITSIZE_MODE_ANY_MODE): Define.
4912 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
4913 aarch64_addr_query_type.
4914 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
4915 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
4916 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
4917 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
4918 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
4919 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
4920 (aarch64_simd_imm_zero_p): Delete.
4921 (aarch64_check_zero_based_sve_index_immediate): Declare.
4922 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4923 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4924 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4925 (aarch64_sve_float_mul_immediate_p): Likewise.
4926 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4928 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
4929 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
4930 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
4931 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
4932 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
4933 (aarch64_regmode_natural_size): Likewise.
4934 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
4935 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
4937 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
4938 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
4939 for VG and the SVE predicate registers.
4940 (V_ALIASES): Add a "z"-prefixed alias.
4941 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
4942 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
4943 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
4944 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
4945 (REG_CLASS_NAMES): Add entries for them.
4946 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
4947 and the predicate registers.
4948 (aarch64_sve_vg): Declare.
4949 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
4950 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
4951 (REGMODE_NATURAL_SIZE): Define.
4952 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
4954 * config/aarch64/aarch64.c: Include cfgrtl.h.
4955 (simd_immediate_info): Add a constructor for series vectors,
4956 and an associated step field.
4957 (aarch64_sve_vg): New variable.
4958 (aarch64_dbx_register_number): Handle VG and the predicate registers.
4959 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
4960 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
4961 (VEC_ANY_DATA, VEC_STRUCT): New constants.
4962 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
4963 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
4964 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
4965 (aarch64_get_mask_mode): New functions.
4966 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
4967 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
4968 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
4969 predicate modes and predicate registers. Explicitly restrict
4970 GPRs to modes of 16 bytes or smaller. Only allow FP registers
4971 to store a vector mode if it is recognized by
4972 aarch64_classify_vector_mode.
4973 (aarch64_regmode_natural_size): New function.
4974 (aarch64_hard_regno_caller_save_mode): Return the original mode
4976 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
4977 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
4978 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
4979 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
4981 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
4982 does not overlap dest if the function is frame-related. Handle
4984 (aarch64_split_add_offset): New function.
4985 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
4986 them aarch64_add_offset.
4987 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
4988 and update call to aarch64_sub_sp.
4989 (aarch64_add_cfa_expression): New function.
4990 (aarch64_expand_prologue): Pass extra temporary registers to the
4991 functions above. Handle the case in which we need to emit new
4992 DW_CFA_expressions for registers that were originally saved
4993 relative to the stack pointer, but now have to be expressed
4994 relative to the frame pointer.
4995 (aarch64_output_mi_thunk): Pass extra temporary registers to the
4997 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
4998 IP0 and IP1 values for SVE frames.
4999 (aarch64_expand_vec_series): New function.
5000 (aarch64_expand_sve_widened_duplicate): Likewise.
5001 (aarch64_expand_sve_const_vector): Likewise.
5002 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
5003 Handle SVE constants. Use emit_move_insn to move a force_const_mem
5004 into the register, rather than emitting a SET directly.
5005 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
5006 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
5007 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
5008 (offset_9bit_signed_scaled_p): New functions.
5009 (aarch64_replicate_bitmask_imm): New function.
5010 (aarch64_bitmask_imm): Use it.
5011 (aarch64_cannot_force_const_mem): Reject expressions involving
5012 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
5013 (aarch64_classify_index): Handle SVE indices, by requiring
5014 a plain register index with a scale that matches the element size.
5015 (aarch64_classify_address): Handle SVE addresses. Assert that
5016 the mode of the address is VOIDmode or an integer mode.
5017 Update call to aarch64_classify_symbol.
5018 (aarch64_classify_symbolic_expression): Update call to
5019 aarch64_classify_symbol.
5020 (aarch64_const_vec_all_in_range_p): New function.
5021 (aarch64_print_vector_float_operand): Likewise.
5022 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
5023 "vN" for FP registers with SVE modes. Handle (const ...) vectors
5024 and the FP immediates 1.0 and 0.5.
5025 (aarch64_print_address_internal): Handle SVE addresses.
5026 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
5027 (aarch64_regno_regclass): Handle predicate registers.
5028 (aarch64_secondary_reload): Handle big-endian reloads of SVE
5030 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
5031 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
5032 (aarch64_convert_sve_vector_bits): New function.
5033 (aarch64_override_options): Use it to handle -msve-vector-bits=.
5034 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5036 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
5037 Handle SVE vector and predicate modes. Accept VL-based constants
5038 that need only one temporary register, and VL offsets that require
5039 no temporary registers.
5040 (aarch64_conditional_register_usage): Mark the predicate registers
5041 as fixed if SVE isn't available.
5042 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
5043 Return true for SVE vector and predicate modes.
5044 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
5045 rather than an unsigned int. Handle SVE modes.
5046 (aarch64_preferred_simd_mode): Update call accordingly. Handle
5048 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
5050 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5051 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5052 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5053 (aarch64_sve_float_mul_immediate_p): New functions.
5054 (aarch64_sve_valid_immediate): New function.
5055 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
5056 Explicitly reject structure modes. Check for INDEX constants.
5057 Handle PTRUE and PFALSE constants.
5058 (aarch64_check_zero_based_sve_index_immediate): New function.
5059 (aarch64_simd_imm_zero_p): Delete.
5060 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
5061 vector modes. Accept constants in the range of CNT[BHWD].
5062 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
5063 ask for an Advanced SIMD mode.
5064 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
5065 (aarch64_simd_vector_alignment): Handle SVE predicates.
5066 (aarch64_vectorize_preferred_vector_alignment): New function.
5067 (aarch64_simd_vector_alignment_reachable): Use it instead of
5069 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
5070 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
5072 (MAX_VECT_LEN): Delete.
5073 (expand_vec_perm_d): Add a vec_flags field.
5074 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
5075 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
5076 (aarch64_evpc_ext): Don't apply a big-endian lane correction
5078 (aarch64_evpc_rev): Rename to...
5079 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
5080 (aarch64_evpc_rev_global): New function.
5081 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
5082 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
5084 (aarch64_evpc_sve_tbl): New function.
5085 (aarch64_expand_vec_perm_const_1): Update after rename of
5086 aarch64_evpc_rev. Handle SVE permutes too, trying
5087 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
5088 than aarch64_evpc_tbl.
5089 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
5090 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
5091 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
5092 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
5093 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
5094 (aarch64_expand_sve_vcond): New functions.
5095 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
5096 of aarch64_vector_mode_p.
5097 (aarch64_dwarf_poly_indeterminate_value): New function.
5098 (aarch64_compute_pressure_classes): Likewise.
5099 (aarch64_can_change_mode_class): Likewise.
5100 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
5101 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
5102 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
5103 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
5104 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
5105 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
5106 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
5107 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
5109 (Dn, Dl, Dr): Accept const as well as const_vector.
5110 (Dz): Likewise. Compare against CONST0_RTX.
5111 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
5112 of "vector" where appropriate.
5113 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
5114 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
5115 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
5116 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
5117 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
5118 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
5119 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
5120 (v_int_equiv): Extend to SVE modes.
5121 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
5123 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
5124 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
5125 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
5126 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
5127 (SVE_COND_FP_CMP): New int iterators.
5128 (perm_hilo): Handle the new unpack unspecs.
5129 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
5131 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
5132 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
5133 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
5134 (aarch64_equality_operator, aarch64_constant_vector_operand)
5135 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
5136 (aarch64_sve_nonimmediate_operand): Likewise.
5137 (aarch64_sve_general_operand): Likewise.
5138 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
5139 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
5140 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
5141 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
5142 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
5143 (aarch64_sve_float_arith_immediate): Likewise.
5144 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
5145 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
5146 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
5147 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
5148 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
5149 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
5150 (aarch64_sve_float_arith_operand): Likewise.
5151 (aarch64_sve_float_arith_with_sub_operand): Likewise.
5152 (aarch64_sve_float_mul_operand): Likewise.
5153 (aarch64_sve_vec_perm_operand): Likewise.
5154 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
5155 (aarch64_mov_operand): Accept const_poly_int and const_vector.
5156 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
5157 as well as const_vector.
5158 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
5159 in file. Use CONST0_RTX and CONSTM1_RTX.
5160 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
5161 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
5162 Use aarch64_simd_imm_zero.
5163 * config/aarch64/aarch64-sve.md: New file.
5164 * config/aarch64/aarch64.md: Include it.
5165 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
5166 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
5167 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
5168 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
5169 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
5170 (sve): New attribute.
5171 (enabled): Disable instructions with the sve attribute unless
5173 (movqi, movhi): Pass CONST_POLY_INT operaneds through
5174 aarch64_expand_mov_immediate.
5175 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
5176 CNT[BHSD] immediates.
5177 (movti): Split CONST_POLY_INT moves into two halves.
5178 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
5179 Split additions that need a temporary here if the destination
5180 is the stack pointer.
5181 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
5182 (*add<mode>3_poly_1): New instruction.
5183 (set_clobber_cc): New expander.
5185 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5187 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
5188 parameter and use it instead of GET_MODE_SIZE (innermode). Use
5189 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
5190 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
5191 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
5192 Change innermode from fixed_mode_size to machine_mode.
5193 (simplify_subreg): Update call accordingly. Handle a constant-sized
5194 subreg of a variable-length CONST_VECTOR.
5196 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5197 Alan Hayward <alan.hayward@arm.com>
5198 David Sherwood <david.sherwood@arm.com>
5200 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
5201 (add_offset_to_base): New function, split out from...
5202 (create_mem_ref): ...here. When handling a scale other than 1,
5203 check first whether the address is valid without the offset.
5204 Add it into the base if so, leaving the index and scale as-is.
5206 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5209 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
5210 fold_for_warn before checking if arg2 is INTEGER_CST.
5212 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
5214 * config/rs6000/predicates.md (load_multiple_operation): Delete.
5215 (store_multiple_operation): Delete.
5216 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
5217 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
5218 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
5219 guarded by TARGET_STRING.
5220 (rs6000_output_load_multiple): Delete.
5221 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
5222 OPTION_MASK_STRING / TARGET_STRING handling.
5223 (print_operand) <'N', 'O'>: Add comment that these are unused now.
5224 (const rs6000_opt_masks) <"string">: Change mask to 0.
5225 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
5226 (MASK_STRING): Delete.
5227 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
5229 (load_multiple): Delete.
5236 (store_multiple): Delete.
5243 (movmemsi_8reg): Delete.
5244 (corresponding unnamed define_insn): Delete.
5245 (movmemsi_6reg): Delete.
5246 (corresponding unnamed define_insn): Delete.
5247 (movmemsi_4reg): Delete.
5248 (corresponding unnamed define_insn): Delete.
5249 (movmemsi_2reg): Delete.
5250 (corresponding unnamed define_insn): Delete.
5251 (movmemsi_1reg): Delete.
5252 (corresponding unnamed define_insn): Delete.
5253 * config/rs6000/rs6000.opt (mno-string): New.
5254 (mstring): Replace by deprecation warning stub.
5255 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
5257 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5259 * regrename.c (regrename_do_replace): If replacing the same
5260 reg multiple times, try to reuse last created gen_raw_REG.
5263 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
5264 main to workaround a bug in GDB.
5266 2018-01-12 Tom de Vries <tom@codesourcery.com>
5269 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
5271 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
5273 PR rtl-optimization/80481
5274 * ira-color.c (get_cap_member): New function.
5275 (allocnos_conflict_by_live_ranges_p): Use it.
5276 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
5277 (setup_slot_coalesced_allocno_live_ranges): Ditto.
5279 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
5282 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
5283 (*saddl_se_1): Ditto.
5285 (*ssubl_se_1): Ditto.
5287 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5289 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
5290 rather than wi::to_widest for DR_INITs.
5291 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
5292 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
5293 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
5295 (vect_analyze_group_access_1): Note that here.
5297 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5299 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
5300 polynomial type sizes.
5302 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5304 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
5305 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
5306 (gimple_add_tmp_var): Likewise.
5308 2018-01-12 Martin Liska <mliska@suse.cz>
5310 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
5311 (gimple_alloc_sizes): Likewise.
5312 (dump_gimple_statistics): Use PRIu64 in printf format.
5313 * gimple.h: Change uint64_t to int.
5315 2018-01-12 Martin Liska <mliska@suse.cz>
5317 * tree-core.h: Use uint64_t instead of int.
5318 * tree.c (tree_node_counts): Likewise.
5319 (tree_node_sizes): Likewise.
5320 (dump_tree_statistics): Use PRIu64 in printf format.
5322 2018-01-12 Martin Liska <mliska@suse.cz>
5324 * Makefile.in: As qsort_chk is implemented in vec.c, add
5325 vec.o to linkage of gencfn-macros.
5326 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
5327 passing the info to record_node_allocation_statistics.
5328 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
5330 * ggc-common.c (struct ggc_usage): Add operator== and use
5331 it in operator< and compare function.
5332 * mem-stats.h (struct mem_usage): Likewise.
5333 * vec.c (struct vec_usage): Remove operator< and compare
5334 function. Can be simply inherited.
5336 2018-01-12 Martin Jambor <mjambor@suse.cz>
5339 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
5340 * tree-ssa-math-opts.c: Include domwalk.h.
5341 (convert_mult_to_fma_1): New function.
5342 (fma_transformation_info): New type.
5343 (fma_deferring_state): Likewise.
5344 (cancel_fma_deferring): New function.
5345 (result_of_phi): Likewise.
5346 (last_fma_candidate_feeds_initial_phi): Likewise.
5347 (convert_mult_to_fma): Added deferring logic, split actual
5348 transformation to convert_mult_to_fma_1.
5349 (math_opts_dom_walker): New type.
5350 (math_opts_dom_walker::after_dom_children): New method, body moved
5351 here from pass_optimize_widening_mul::execute, added deferring logic
5353 (pass_optimize_widening_mul::execute): Moved most of code to
5354 math_opts_dom_walker::after_dom_children.
5355 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
5356 * config/i386/i386.c (ix86_option_override_internal): Added
5357 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
5359 2018-01-12 Richard Biener <rguenther@suse.de>
5362 * dwarf2out.c (gen_variable_die): Do not reset old_die for
5363 inline instance vars.
5365 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
5368 * config/rx/rx.c (rx_is_restricted_memory_address):
5371 2018-01-12 Richard Biener <rguenther@suse.de>
5373 PR tree-optimization/80846
5374 * target.def (split_reduction): New target hook.
5375 * targhooks.c (default_split_reduction): New function.
5376 * targhooks.h (default_split_reduction): Declare.
5377 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
5378 target requests first reduce vectors by combining low and high
5380 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
5381 (get_vectype_for_scalar_type_and_size): Export.
5382 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
5383 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
5384 * doc/tm.texi: Regenerate.
5385 * config/i386/i386.c (ix86_split_reduction): Implement
5386 TARGET_VECTORIZE_SPLIT_REDUCTION.
5388 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5391 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
5392 in PIC mode except for TARGET_VXWORKS_RTP.
5393 * config/sparc/sparc.c: Include cfgrtl.h.
5394 (TARGET_INIT_PIC_REG): Define.
5395 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
5396 (sparc_pic_register_p): New predicate.
5397 (sparc_legitimate_address_p): Use it.
5398 (sparc_legitimize_pic_address): Likewise.
5399 (sparc_delegitimize_address): Likewise.
5400 (sparc_mode_dependent_address_p): Likewise.
5401 (gen_load_pcrel_sym): Remove 4th parameter.
5402 (load_got_register): Adjust call to above. Remove obsolete stuff.
5403 (sparc_expand_prologue): Do not call load_got_register here.
5404 (sparc_flat_expand_prologue): Likewise.
5405 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
5406 (sparc_use_pseudo_pic_reg): New function.
5407 (sparc_init_pic_reg): Likewise.
5408 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5409 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
5411 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
5413 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
5414 Add item for branch_cost.
5416 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5418 PR rtl-optimization/83565
5419 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
5420 not extend the result to a larger mode for rotate operations.
5421 (num_sign_bit_copies1): Likewise.
5423 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5426 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
5428 Use values-Xc.o for -pedantic.
5429 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
5431 2018-01-12 Martin Liska <mliska@suse.cz>
5434 * ipa-devirt.c (final_warning_record::grow_type_warnings):
5436 (possible_polymorphic_call_targets): Use it.
5437 (ipa_devirt): Likewise.
5439 2018-01-12 Martin Liska <mliska@suse.cz>
5441 * profile-count.h (enum profile_quality): Use 0 as invalid
5442 enum value of profile_quality.
5444 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
5446 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
5447 -mext-string options.
5449 2018-01-12 Richard Biener <rguenther@suse.de>
5451 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
5452 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
5453 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
5455 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
5457 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5459 * configure.ac (--with-long-double-format): Add support for the
5460 configuration option to change the default long double format on
5462 * config.gcc (powerpc*-linux*-*): Likewise.
5463 * configure: Regenerate.
5464 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
5465 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
5466 used without modification.
5468 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5470 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
5471 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
5472 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
5473 MISC_BUILTIN_SPEC_BARRIER.
5474 (rs6000_init_builtins): Likewise.
5475 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
5477 (speculation_barrier): New define_insn.
5478 * doc/extend.texi: Document __builtin_speculation_barrier.
5480 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5483 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
5484 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
5485 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
5487 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
5488 integral modes instead of "ss" and "sd".
5489 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
5490 vectors with 32-bit and 64-bit elements.
5491 (vecdupssescalarmodesuffix): New mode attribute.
5492 (vec_dup<mode>): Use it.
5494 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
5497 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
5498 frame if argument is passed on stack.
5500 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5503 * ree.c (combine_reaching_defs): Optimize also
5504 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
5505 reg2=any_extend(exp); reg1=reg2;, formatting fix.
5507 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5510 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
5512 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5515 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
5516 after they are computed.
5518 2018-01-11 Bin Cheng <bin.cheng@arm.com>
5520 PR tree-optimization/83695
5521 * gimple-loop-linterchange.cc
5522 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
5523 reset cached scev information after interchange.
5524 (pass_linterchange::execute): Remove call to scev_reset_htab.
5526 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5528 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
5529 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
5530 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
5531 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
5532 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
5533 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
5534 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
5535 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
5536 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
5537 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
5538 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
5539 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
5540 (V_lane_reg): Likewise.
5541 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
5543 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
5544 (vfmal_lane_low<mode>_intrinsic,
5545 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
5546 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
5547 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
5548 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
5549 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
5550 vfmsl_lane_high<mode>_intrinsic): New define_insns.
5552 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5554 * config/arm/arm-cpus.in (fp16fml): New feature.
5555 (ALL_SIMD): Add fp16fml.
5556 (armv8.2-a): Add fp16fml as an option.
5557 (armv8.3-a): Likewise.
5558 (armv8.4-a): Add fp16fml as part of fp16.
5559 * config/arm/arm.h (TARGET_FP16FML): Define.
5560 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
5562 * config/arm/arm-modes.def (V2HF): Define.
5563 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
5564 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
5565 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
5566 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
5567 vfmsl_low, vfmsl_high): New set of builtins.
5568 * config/arm/iterators.md (PLUSMINUS): New code iterator.
5569 (vfml_op): New code attribute.
5570 (VFMLHALVES): New int iterator.
5571 (VFML, VFMLSEL): New mode attributes.
5572 (V_reg): Define mapping for V2HF.
5573 (V_hi, V_lo): New mode attributes.
5574 (VF_constraint): Likewise.
5575 (vfml_half, vfml_half_selector): New int attributes.
5576 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
5578 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
5579 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
5581 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
5582 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
5583 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
5584 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
5586 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
5587 Document new effective target and option set.
5589 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5591 * config/arm/arm-cpus.in (armv8_4): New feature.
5592 (ARMv8_4a): New fgroup.
5593 (armv8.4-a): New arch.
5594 * config/arm/arm-tables.opt: Regenerate.
5595 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
5596 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
5597 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
5598 Add matching rules for -march=armv8.4-a and extensions.
5599 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
5601 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
5604 * config/rx/rx.md (BW): New mode attribute.
5605 (sync_lock_test_and_setsi): Add mode suffix to insn output.
5607 2018-01-11 Richard Biener <rguenther@suse.de>
5609 PR tree-optimization/83435
5610 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
5611 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
5612 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
5614 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5615 Alan Hayward <alan.hayward@arm.com>
5616 David Sherwood <david.sherwood@arm.com>
5618 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
5620 (aarch64_classify_address): Initialize it. Track polynomial offsets.
5621 (aarch64_print_address_internal): Use it to check for a zero offset.
5623 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5624 Alan Hayward <alan.hayward@arm.com>
5625 David Sherwood <david.sherwood@arm.com>
5627 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
5628 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
5629 Return a poly_int64 rather than a HOST_WIDE_INT.
5630 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
5631 rather than a HOST_WIDE_INT.
5632 * config/aarch64/aarch64.h (aarch64_frame): Protect with
5633 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
5634 hard_fp_offset, frame_size, initial_adjust, callee_offset and
5635 final_offset from HOST_WIDE_INT to poly_int64.
5636 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
5637 to_constant when getting the number of units in an Advanced SIMD
5639 (aarch64_builtin_vectorized_function): Check for a constant number
5641 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
5643 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
5644 attribute instead of GET_MODE_NUNITS.
5645 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
5646 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
5647 GET_MODE_SIZE for fixed-size registers.
5648 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
5649 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
5650 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
5651 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
5652 (aarch64_print_operand, aarch64_print_address_internal)
5653 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
5654 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
5655 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
5656 Handle polynomial GET_MODE_SIZE.
5657 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
5658 wider than SImode without modification.
5659 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
5660 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
5661 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
5662 passing and returning SVE modes.
5663 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
5664 rather than GEN_INT.
5665 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
5666 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
5667 (aarch64_allocate_and_probe_stack_space): Likewise.
5668 (aarch64_layout_frame): Cope with polynomial offsets.
5669 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
5670 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
5672 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
5673 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
5674 poly_int64 rather than a HOST_WIDE_INT.
5675 (aarch64_get_separate_components, aarch64_process_components)
5676 (aarch64_expand_prologue, aarch64_expand_epilogue)
5677 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
5678 (aarch64_anchor_offset): New function, split out from...
5679 (aarch64_legitimize_address): ...here.
5680 (aarch64_builtin_vectorization_cost): Handle polynomial
5681 TYPE_VECTOR_SUBPARTS.
5682 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
5684 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
5685 number of elements from the PARALLEL rather than the mode.
5686 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
5687 rather than GET_MODE_BITSIZE.
5688 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
5689 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
5690 (aarch64_expand_vec_perm_const_1): Handle polynomial
5691 d->perm.length () and d->perm elements.
5692 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
5693 Apply to_constant to d->perm elements.
5694 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
5695 polynomial CONST_VECTOR_NUNITS.
5696 (aarch64_move_pointer): Take amount as a poly_int64 rather
5698 (aarch64_progress_pointer): Avoid temporary variable.
5699 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
5700 the mode attribute instead of GET_MODE.
5702 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5703 Alan Hayward <alan.hayward@arm.com>
5704 David Sherwood <david.sherwood@arm.com>
5706 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
5707 x exists before using it.
5708 (aarch64_add_constant_internal): Rename to...
5709 (aarch64_add_offset_1): ...this. Replace regnum with separate
5710 src and dest rtxes. Handle the case in which they're different,
5711 including when the offset is zero. Replace scratchreg with an rtx.
5712 Use 2 additions if there is no spare register into which we can
5713 move a 16-bit constant.
5714 (aarch64_add_constant): Delete.
5715 (aarch64_add_offset): Replace reg with separate src and dest
5716 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
5717 Use aarch64_add_offset_1.
5718 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
5719 an rtx rather than an int. Take the delta as a poly_int64
5720 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
5721 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
5722 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
5723 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
5724 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
5726 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
5727 aarch64_add_constant.
5729 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5731 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
5732 Use scalar_float_mode.
5734 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5736 * config/aarch64/aarch64-simd.md
5737 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
5738 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
5739 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
5740 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
5741 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
5742 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
5743 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
5744 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
5745 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
5746 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
5748 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5751 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
5752 targ_options->x_arm_arch_string is non NULL.
5754 2018-01-11 Tamar Christina <tamar.christina@arm.com>
5756 * config/aarch64/aarch64.h
5757 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
5759 2018-01-11 Sudakshina Das <sudi.das@arm.com>
5762 * expmed.c (emit_store_flag_force): Swap if const op0
5763 and change VOIDmode to mode of op0.
5765 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5767 PR rtl-optimization/83761
5768 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
5769 than bytes to mode_for_size.
5771 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5774 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
5775 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
5778 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5781 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
5782 when in layout mode.
5783 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
5784 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
5787 2018-01-10 Michael Collison <michael.collison@arm.com>
5789 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
5790 * config/aarch64/aarch64-option-extension.def: Add
5791 AARCH64_OPT_EXTENSION of 'fp16fml'.
5792 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5793 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
5794 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
5795 * config/aarch64/constraints.md (Ui7): New constraint.
5796 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
5797 (VFMLA_SEL_W): Ditto.
5800 (VFMLA16_LOW): New int iterator.
5801 (VFMLA16_HIGH): Ditto.
5802 (UNSPEC_FMLAL): New unspec.
5803 (UNSPEC_FMLSL): Ditto.
5804 (UNSPEC_FMLAL2): Ditto.
5805 (UNSPEC_FMLSL2): Ditto.
5806 (f16mac): New code attribute.
5807 * config/aarch64/aarch64-simd-builtins.def
5808 (aarch64_fmlal_lowv2sf): Ditto.
5809 (aarch64_fmlsl_lowv2sf): Ditto.
5810 (aarch64_fmlalq_lowv4sf): Ditto.
5811 (aarch64_fmlslq_lowv4sf): Ditto.
5812 (aarch64_fmlal_highv2sf): Ditto.
5813 (aarch64_fmlsl_highv2sf): Ditto.
5814 (aarch64_fmlalq_highv4sf): Ditto.
5815 (aarch64_fmlslq_highv4sf): Ditto.
5816 (aarch64_fmlal_lane_lowv2sf): Ditto.
5817 (aarch64_fmlsl_lane_lowv2sf): Ditto.
5818 (aarch64_fmlal_laneq_lowv2sf): Ditto.
5819 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
5820 (aarch64_fmlalq_lane_lowv4sf): Ditto.
5821 (aarch64_fmlsl_lane_lowv4sf): Ditto.
5822 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
5823 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
5824 (aarch64_fmlal_lane_highv2sf): Ditto.
5825 (aarch64_fmlsl_lane_highv2sf): Ditto.
5826 (aarch64_fmlal_laneq_highv2sf): Ditto.
5827 (aarch64_fmlsl_laneq_highv2sf): Ditto.
5828 (aarch64_fmlalq_lane_highv4sf): Ditto.
5829 (aarch64_fmlsl_lane_highv4sf): Ditto.
5830 (aarch64_fmlalq_laneq_highv4sf): Ditto.
5831 (aarch64_fmlsl_laneq_highv4sf): Ditto.
5832 * config/aarch64/aarch64-simd.md:
5833 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
5834 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5835 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
5836 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5837 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
5838 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
5839 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
5840 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
5841 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
5842 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
5843 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
5844 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
5845 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
5846 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
5847 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
5848 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
5849 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
5850 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
5851 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
5852 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
5853 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
5854 (vfmlsl_low_u32): Ditto.
5855 (vfmlalq_low_u32): Ditto.
5856 (vfmlslq_low_u32): Ditto.
5857 (vfmlal_high_u32): Ditto.
5858 (vfmlsl_high_u32): Ditto.
5859 (vfmlalq_high_u32): Ditto.
5860 (vfmlslq_high_u32): Ditto.
5861 (vfmlal_lane_low_u32): Ditto.
5862 (vfmlsl_lane_low_u32): Ditto.
5863 (vfmlal_laneq_low_u32): Ditto.
5864 (vfmlsl_laneq_low_u32): Ditto.
5865 (vfmlalq_lane_low_u32): Ditto.
5866 (vfmlslq_lane_low_u32): Ditto.
5867 (vfmlalq_laneq_low_u32): Ditto.
5868 (vfmlslq_laneq_low_u32): Ditto.
5869 (vfmlal_lane_high_u32): Ditto.
5870 (vfmlsl_lane_high_u32): Ditto.
5871 (vfmlal_laneq_high_u32): Ditto.
5872 (vfmlsl_laneq_high_u32): Ditto.
5873 (vfmlalq_lane_high_u32): Ditto.
5874 (vfmlslq_lane_high_u32): Ditto.
5875 (vfmlalq_laneq_high_u32): Ditto.
5876 (vfmlslq_laneq_high_u32): Ditto.
5877 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
5878 (AARCH64_FL_FOR_ARCH8_4): New.
5879 (AARCH64_ISA_F16FML): New ISA flag.
5880 (TARGET_F16FML): New feature flag for fp16fml.
5881 (doc/invoke.texi): Document new fp16fml option.
5883 2018-01-10 Michael Collison <michael.collison@arm.com>
5885 * config/aarch64/aarch64-builtins.c:
5886 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
5887 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5888 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
5889 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
5890 (AARCH64_ISA_SHA3): New ISA flag.
5891 (TARGET_SHA3): New feature flag for sha3.
5892 * config/aarch64/iterators.md (sha512_op): New int attribute.
5893 (CRYPTO_SHA512): New int iterator.
5894 (UNSPEC_SHA512H): New unspec.
5895 (UNSPEC_SHA512H2): Ditto.
5896 (UNSPEC_SHA512SU0): Ditto.
5897 (UNSPEC_SHA512SU1): Ditto.
5898 * config/aarch64/aarch64-simd-builtins.def
5899 (aarch64_crypto_sha512hqv2di): New builtin.
5900 (aarch64_crypto_sha512h2qv2di): Ditto.
5901 (aarch64_crypto_sha512su0qv2di): Ditto.
5902 (aarch64_crypto_sha512su1qv2di): Ditto.
5903 (aarch64_eor3qv8hi): Ditto.
5904 (aarch64_rax1qv2di): Ditto.
5905 (aarch64_xarqv2di): Ditto.
5906 (aarch64_bcaxqv8hi): Ditto.
5907 * config/aarch64/aarch64-simd.md:
5908 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
5909 (aarch64_crypto_sha512su0qv2di): Ditto.
5910 (aarch64_crypto_sha512su1qv2di): Ditto.
5911 (aarch64_eor3qv8hi): Ditto.
5912 (aarch64_rax1qv2di): Ditto.
5913 (aarch64_xarqv2di): Ditto.
5914 (aarch64_bcaxqv8hi): Ditto.
5915 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
5916 (vsha512h2q_u64): Ditto.
5917 (vsha512su0q_u64): Ditto.
5918 (vsha512su1q_u64): Ditto.
5919 (veor3q_u16): Ditto.
5920 (vrax1q_u64): Ditto.
5922 (vbcaxq_u16): Ditto.
5923 * config/arm/types.md (crypto_sha512): New type attribute.
5924 (crypto_sha3): Ditto.
5925 (doc/invoke.texi): Document new sha3 option.
5927 2018-01-10 Michael Collison <michael.collison@arm.com>
5929 * config/aarch64/aarch64-builtins.c:
5930 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
5931 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5932 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
5933 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
5934 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
5935 (AARCH64_ISA_SM4): New ISA flag.
5936 (TARGET_SM4): New feature flag for sm4.
5937 * config/aarch64/aarch64-simd-builtins.def
5938 (aarch64_sm3ss1qv4si): Ditto.
5939 (aarch64_sm3tt1aq4si): Ditto.
5940 (aarch64_sm3tt1bq4si): Ditto.
5941 (aarch64_sm3tt2aq4si): Ditto.
5942 (aarch64_sm3tt2bq4si): Ditto.
5943 (aarch64_sm3partw1qv4si): Ditto.
5944 (aarch64_sm3partw2qv4si): Ditto.
5945 (aarch64_sm4eqv4si): Ditto.
5946 (aarch64_sm4ekeyqv4si): Ditto.
5947 * config/aarch64/aarch64-simd.md:
5948 (aarch64_sm3ss1qv4si): Ditto.
5949 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
5950 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
5951 (aarch64_sm4eqv4si): Ditto.
5952 (aarch64_sm4ekeyqv4si): Ditto.
5953 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
5954 (sm3part_op): Ditto.
5955 (CRYPTO_SM3TT): Ditto.
5956 (CRYPTO_SM3PART): Ditto.
5957 (UNSPEC_SM3SS1): New unspec.
5958 (UNSPEC_SM3TT1A): Ditto.
5959 (UNSPEC_SM3TT1B): Ditto.
5960 (UNSPEC_SM3TT2A): Ditto.
5961 (UNSPEC_SM3TT2B): Ditto.
5962 (UNSPEC_SM3PARTW1): Ditto.
5963 (UNSPEC_SM3PARTW2): Ditto.
5964 (UNSPEC_SM4E): Ditto.
5965 (UNSPEC_SM4EKEY): Ditto.
5966 * config/aarch64/constraints.md (Ui2): New constraint.
5967 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
5968 * config/arm/types.md (crypto_sm3): New type attribute.
5969 (crypto_sm4): Ditto.
5970 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
5971 (vsm3tt1aq_u32): Ditto.
5972 (vsm3tt1bq_u32): Ditto.
5973 (vsm3tt2aq_u32): Ditto.
5974 (vsm3tt2bq_u32): Ditto.
5975 (vsm3partw1q_u32): Ditto.
5976 (vsm3partw2q_u32): Ditto.
5977 (vsm4eq_u32): Ditto.
5978 (vsm4ekeyq_u32): Ditto.
5979 (doc/invoke.texi): Document new sm4 option.
5981 2018-01-10 Michael Collison <michael.collison@arm.com>
5983 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
5984 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
5985 (AARCH64_FL_FOR_ARCH8_4): New.
5986 (AARCH64_FL_V8_4): New flag.
5987 (doc/invoke.texi): Document new armv8.4-a option.
5989 2018-01-10 Michael Collison <michael.collison@arm.com>
5991 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5992 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
5993 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
5994 * config/aarch64/aarch64-option-extension.def: Add
5995 AARCH64_OPT_EXTENSION of 'sha2'.
5996 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
5997 (crypto): Disable sha2 and aes if crypto disabled.
5998 (crypto): Enable aes and sha2 if enabled.
5999 (simd): Disable sha2 and aes if simd disabled.
6000 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
6002 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
6003 (TARGET_SHA2): New feature flag for sha2.
6004 (TARGET_AES): New feature flag for aes.
6005 * config/aarch64/aarch64-simd.md:
6006 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
6007 conditional on TARGET_AES.
6008 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
6009 (aarch64_crypto_sha1hsi): Make pattern conditional
6011 (aarch64_crypto_sha1hv4si): Ditto.
6012 (aarch64_be_crypto_sha1hv4si): Ditto.
6013 (aarch64_crypto_sha1su1v4si): Ditto.
6014 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
6015 (aarch64_crypto_sha1su0v4si): Ditto.
6016 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
6017 (aarch64_crypto_sha256su0v4si): Ditto.
6018 (aarch64_crypto_sha256su1v4si): Ditto.
6019 (doc/invoke.texi): Document new aes and sha2 options.
6021 2018-01-10 Martin Sebor <msebor@redhat.com>
6023 PR tree-optimization/83781
6024 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
6027 2018-01-11 Martin Sebor <msebor@gmail.com>
6028 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6030 PR tree-optimization/83501
6031 PR tree-optimization/81703
6033 * tree-ssa-strlen.c (get_string_cst): Rename...
6034 (get_string_len): ...to this. Handle global constants.
6035 (handle_char_store): Adjust.
6037 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
6038 Jim Wilson <jimw@sifive.com>
6040 * config/riscv/riscv-protos.h (riscv_output_return): New.
6041 * config/riscv/riscv.c (struct machine_function): New naked_p field.
6042 (riscv_attribute_table, riscv_output_return),
6043 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
6044 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
6045 (riscv_compute_frame_info): Only compute frame->mask if not a naked
6047 (riscv_expand_prologue): Add early return for naked function.
6048 (riscv_expand_epilogue): Likewise.
6049 (riscv_function_ok_for_sibcall): Return false for naked function.
6050 (riscv_set_current_function): New.
6051 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
6052 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
6053 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
6054 * doc/extend.texi (RISC-V Function Attributes): New.
6056 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6058 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
6059 check for 128-bit long double before checking TCmode.
6060 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
6061 128-bit long doubles before checking TFmode or TCmode.
6062 (FLOAT128_IBM_P): Likewise.
6064 2018-01-10 Martin Sebor <msebor@redhat.com>
6066 PR tree-optimization/83671
6067 * builtins.c (c_strlen): Unconditionally return zero for the empty
6069 Use -Warray-bounds for warnings.
6070 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
6071 for non-constant array indices with COMPONENT_REF, arrays of
6072 arrays, and pointers to arrays.
6073 (gimple_fold_builtin_strlen): Determine and set length range for
6074 non-constant character arrays.
6076 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
6079 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
6082 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
6084 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
6086 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6089 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
6090 VECTOR_MEM_ALTIVEC_OR_VSX_P.
6091 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
6092 indexed_or_indirect_operand predicate.
6093 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
6094 (*vsx_le_perm_load_v8hi): Likewise.
6095 (*vsx_le_perm_load_v16qi): Likewise.
6096 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
6097 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
6098 (*vsx_le_perm_store_v8hi): Likewise.
6099 (*vsx_le_perm_store_v16qi): Likewise.
6100 (eight unnamed splitters): Likewise.
6102 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6104 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
6105 * config/rs6000/emmintrin.h: Likewise.
6106 * config/rs6000/mmintrin.h: Likewise.
6107 * config/rs6000/xmmintrin.h: Likewise.
6109 2018-01-10 David Malcolm <dmalcolm@redhat.com>
6112 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
6114 * tree.c (tree_nop_conversion): Return true for location wrapper
6116 (maybe_wrap_with_location): New function.
6117 (selftest::check_strip_nops): New function.
6118 (selftest::test_location_wrappers): New function.
6119 (selftest::tree_c_tests): Call it.
6120 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
6121 (maybe_wrap_with_location): New decl.
6122 (EXPR_LOCATION_WRAPPER_P): New macro.
6123 (location_wrapper_p): New inline function.
6124 (tree_strip_any_location_wrapper): New inline function.
6126 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
6129 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
6130 stack_realign_offset for the largest alignment of stack slot
6132 (ix86_find_max_used_stack_alignment): New function.
6133 (ix86_finalize_stack_frame_flags): Use it. Set
6134 max_used_stack_alignment if we don't realign stack.
6135 * config/i386/i386.h (machine_function): Add
6136 max_used_stack_alignment.
6138 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
6140 * config/arm/arm.opt (-mbranch-cost): New option.
6141 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
6144 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
6147 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
6148 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
6150 2018-01-10 Richard Biener <rguenther@suse.de>
6153 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
6154 early out so it also covers the case where we have a non-NULL
6157 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6159 PR tree-optimization/83753
6160 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
6161 for non-strided grouped accesses if the number of elements is 1.
6163 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6166 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
6167 * i386.h (TARGET_USE_GATHER): Define.
6168 * x86-tune.def (X86_TUNE_USE_GATHER): New.
6170 2018-01-10 Martin Liska <mliska@suse.cz>
6173 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
6174 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
6176 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
6177 CLEANUP_NO_PARTITIONING is not set.
6179 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6181 * doc/rtl.texi: Remove documentation of (const ...) wrappers
6182 for vectors, as a partial revert of r254296.
6183 * rtl.h (const_vec_p): Delete.
6184 (const_vec_duplicate_p): Don't test for vector CONSTs.
6185 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
6186 * expmed.c (make_tree): Likewise.
6189 * common.md (E, F): Use CONSTANT_P instead of checking for
6191 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
6192 checking for CONST_VECTOR.
6194 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6197 * predict.c (force_edge_cold): Handle in more sane way edges
6200 2018-01-09 Carl Love <cel@us.ibm.com>
6202 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
6204 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
6205 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
6206 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
6207 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
6208 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
6209 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
6210 * config/rs6000/rs6000-protos.h: Add extern defition for
6211 rs6000_generate_float2_double_code.
6212 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
6214 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
6215 (float2_v2df): Add define_expand.
6217 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
6220 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
6221 op_mode in the force_to_mode call.
6223 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6225 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
6226 instead of checking each element individually.
6227 (aarch64_evpc_uzp): Likewise.
6228 (aarch64_evpc_zip): Likewise.
6229 (aarch64_evpc_ext): Likewise.
6230 (aarch64_evpc_rev): Likewise.
6231 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
6232 instead of checking each element individually. Return true without
6234 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
6235 whether all selected elements come from the same input, instead of
6236 checking each element individually. Remove calls to gen_rtx_REG,
6237 start_sequence and end_sequence and instead assert that no rtl is
6240 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6242 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
6243 order of HIGH and CONST checks.
6245 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6247 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
6248 if the destination isn't an SSA_NAME.
6250 2018-01-09 Richard Biener <rguenther@suse.de>
6252 PR tree-optimization/83668
6253 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
6255 (canonicalize_loop_form): ... here, renamed from ...
6256 (canonicalize_loop_closed_ssa_form): ... this and amended to
6257 swap successor edges for loop exit blocks to make us use
6258 the RPO order we need for initial schedule generation.
6260 2018-01-09 Joseph Myers <joseph@codesourcery.com>
6262 PR tree-optimization/64811
6263 * match.pd: When optimizing comparisons with Inf, avoid
6264 introducing or losing exceptions from comparisons with NaN.
6266 2018-01-09 Martin Liska <mliska@suse.cz>
6269 * asan.c (shadow_mem_size): Add gcc_assert.
6271 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
6273 Don't save registers in main().
6276 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
6277 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
6278 * config/avr/avr.c (avr_set_current_function): Don't error if
6279 naked, OS_task or OS_main are specified at the same time.
6280 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
6282 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
6284 * common/config/avr/avr-common.c (avr_option_optimization_table):
6285 Switch on -mmain-is-OS_task for optimizing compilations.
6287 2018-01-09 Richard Biener <rguenther@suse.de>
6289 PR tree-optimization/83572
6290 * graphite.c: Include cfganal.h.
6291 (graphite_transform_loops): Connect infinite loops to exit
6292 and remove fake edges at the end.
6294 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6296 * ipa-inline.c (edge_badness): Revert accidental checkin.
6298 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6301 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
6302 symbols; not inline clones.
6304 2018-01-09 Jakub Jelinek <jakub@redhat.com>
6307 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
6308 hard registers. Formatting fixes.
6310 PR preprocessor/83722
6311 * gcc.c (try_generate_repro): Pass
6312 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
6313 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
6316 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
6317 Kito Cheng <kito.cheng@gmail.com>
6319 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
6320 (riscv_leaf_function_p): Delete.
6321 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
6323 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6325 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
6327 (do_ifelse): New function.
6328 (do_isel): New function.
6329 (do_sub3): New function.
6330 (do_add3): New function.
6331 (do_load_mask_compare): New function.
6332 (do_overlap_load_compare): New function.
6333 (expand_compare_loop): New function.
6334 (expand_block_compare): Call expand_compare_loop() when appropriate.
6335 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
6337 (-mblock-compare-inline-loop-limit): New option.
6339 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6342 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
6343 Reverse order of second and third operands in first alternative.
6344 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
6345 of first and second elements in UNSPEC_VPERMR vector.
6346 (altivec_expand_vec_perm_le): Likewise.
6348 2018-01-08 Jeff Law <law@redhat.com>
6350 PR rtl-optimizatin/81308
6351 * tree-switch-conversion.c (cfg_altered): New file scoped static.
6352 (process_switch): If group_case_labels makes a change, then set
6354 (pass_convert_switch::execute): If a switch is converted, then
6355 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
6357 PR rtl-optimization/81308
6358 * recog.c (split_all_insns): Conditionally cleanup the CFG after
6361 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
6363 PR target/83663 - Revert r255946
6364 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
6365 generation for cases where splatting a value is not useful.
6366 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
6367 across a vec_duplicate and a paradoxical subreg forming a vector
6368 mode to a vec_concat.
6370 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6372 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
6373 -march=armv8.3-a variants.
6374 * config/arm/t-multilib: Likewise.
6375 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
6377 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6379 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
6381 (cceq_ior_compare_complement): Give it a name so I can use it, and
6382 change boolean_or_operator predicate to boolean_operator so it can
6383 be used to generate a crand.
6384 (eqne): New code iterator.
6385 (bd/bd_neg): New code_attrs.
6386 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
6387 a single define_insn.
6388 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
6389 decrement (bdnzt/bdnzf/bdzt/bdzf).
6390 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
6391 with the new names of the branch decrement patterns, and added the
6392 names of the branch decrement conditional patterns.
6394 2018-01-08 Richard Biener <rguenther@suse.de>
6396 PR tree-optimization/83563
6397 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
6400 2018-01-08 Richard Biener <rguenther@suse.de>
6403 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
6405 2018-01-08 Richard Biener <rguenther@suse.de>
6407 PR tree-optimization/83685
6408 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
6409 references to abnormals.
6411 2018-01-08 Richard Biener <rguenther@suse.de>
6414 * dwarf2out.c (output_indirect_strings): Handle empty
6415 skeleton_debug_str_hash.
6416 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
6418 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6420 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
6421 (emit_store_direct): Likewise.
6422 (arc_trampoline_adjust_address): Likewise.
6423 (arc_asm_trampoline_template): New function.
6424 (arc_initialize_trampoline): Use asm_trampoline_template.
6425 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
6426 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
6427 * config/arc/arc.md (flush_icache): Delete pattern.
6429 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6431 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
6432 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
6435 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6438 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
6439 by not USED_FOR_TARGET.
6440 (make_pass_resolve_sw_modes): Likewise.
6442 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6444 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
6447 2018-01-08 Richard Biener <rguenther@suse.de>
6450 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
6452 2018-01-08 Richard Biener <rguenther@suse.de>
6455 * match.pd ((t * 2) / 2) -> t): Add missing :c.
6457 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
6460 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
6461 basic blocks with a small number of successors.
6462 (convert_control_dep_chain_into_preds): Improve handling of
6464 (dump_predicates): Split apart into...
6465 (dump_pred_chain): ...here...
6466 (dump_pred_info): ...and here.
6467 (can_one_predicate_be_invalidated_p): Add debugging printfs.
6468 (can_chain_union_be_invalidated_p): Improve check for invalidation
6470 (uninit_uses_cannot_happen): Avoid unnecessary if
6471 convert_control_dep_chain_into_preds yielded nothing.
6473 2018-01-06 Martin Sebor <msebor@redhat.com>
6475 PR tree-optimization/83640
6476 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
6477 subtracting negative offset from size.
6478 (builtin_access::overlap): Adjust offset bounds of the access to fall
6479 within the size of the object if possible.
6481 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
6483 PR rtl-optimization/83699
6484 * expmed.c (extract_bit_field_1): Restrict the vector usage of
6485 extract_bit_field_as_subreg to cases in which the extracted
6486 value is also a vector.
6488 * lra-constraints.c (process_alt_operands): Test for the equivalence
6489 substitutions when detecting a possible reload cycle.
6491 2018-01-06 Jakub Jelinek <jakub@redhat.com>
6494 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
6495 by default if flag_selective_schedling{,2}. Formatting fixes.
6497 PR rtl-optimization/83682
6498 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
6499 if it has non-VECTOR_MODE element mode.
6500 (vec_duplicate_p): Likewise.
6503 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
6504 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
6506 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6509 * config/i386/i386-builtin.def
6510 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
6511 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
6512 Require also OPTION_MASK_ISA_AVX512F in addition to
6513 OPTION_MASK_ISA_GFNI.
6514 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
6515 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
6516 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
6517 to OPTION_MASK_ISA_GFNI.
6518 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
6519 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
6520 OPTION_MASK_ISA_AVX512BW.
6521 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
6522 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
6523 addition to OPTION_MASK_ISA_GFNI.
6524 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
6525 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
6526 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
6527 to OPTION_MASK_ISA_GFNI.
6528 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
6529 a requirement for all ISAs rather than any of them with a few
6531 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
6533 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
6534 bitmasks to be enabled with 3 exceptions, instead of requiring any
6535 enabled ISA with lots of exceptions.
6536 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
6537 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
6538 Change avx512bw in isa attribute to avx512f.
6539 * config/i386/sgxintrin.h: Add license boilerplate.
6540 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
6541 to __AVX512F__ and __AVX512VL to __AVX512VL__.
6542 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
6543 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
6545 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
6546 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
6547 temporarily sse2 rather than sse if not enabled already.
6550 * config/i386/sse.md (VI248_VLBW): Rename to ...
6551 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
6552 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
6553 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
6554 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
6555 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
6556 mode iterator instead of VI248_VLBW.
6558 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
6560 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
6561 (record_modified): Skip clobbers; add debug output.
6562 (param_change_prob): Use sreal frequencies.
6564 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6566 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
6567 punt for user-aligned variables.
6569 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6571 * tree-chrec.c (chrec_contains_symbols): Return true for
6574 2018-01-05 Sudakshina Das <sudi.das@arm.com>
6577 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
6578 of (x|y) == x for BICS pattern.
6580 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6582 PR tree-optimization/83605
6583 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
6584 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
6587 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
6589 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
6590 * config/epiphany/rtems.h: New file.
6592 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6593 Uros Bizjak <ubizjak@gmail.com>
6596 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
6597 QIreg_operand instead of register_operand predicate.
6598 * config/i386/i386.c (ix86_rop_should_change_byte_p,
6599 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
6600 comments instead of -fmitigate[-_]rop.
6602 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6605 * cgraphunit.c (symbol_table::compile): Switch to text_section
6606 before calling assembly_start debug hook.
6607 * run-rtl-passes.c (run_rtl_passes): Likewise.
6610 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6612 * tree-vrp.c (extract_range_from_binary_expr_1): Check
6613 range_int_cst_p rather than !symbolic_range_p before calling
6614 extract_range_from_multiplicative_op_1.
6616 2018-01-04 Jeff Law <law@redhat.com>
6618 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
6619 redundant test in assertion.
6621 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6623 * doc/rtl.texi: Document machine_mode wrapper classes.
6625 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6627 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
6630 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6632 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
6633 the VEC_PERM_EXPR fold to fail.
6635 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6638 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
6639 to switched_sections.
6641 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6644 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
6647 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
6650 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
6651 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
6653 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6656 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
6657 is BLKmode and bitpos not zero or mode change is needed.
6659 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6662 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
6665 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
6668 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
6669 instead of MULT rtx. Update all corresponding splitters.
6671 (*ssub<modesuffix>): Ditto.
6673 (*cmp_sadd_di): Update split patterns.
6674 (*cmp_sadd_si): Ditto.
6675 (*cmp_sadd_sidi): Ditto.
6676 (*cmp_ssub_di): Ditto.
6677 (*cmp_ssub_si): Ditto.
6678 (*cmp_ssub_sidi): Ditto.
6679 * config/alpha/predicates.md (const23_operand): New predicate.
6680 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
6681 Look for ASHIFT, not MULT inner operand.
6682 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
6684 2018-01-04 Martin Liska <mliska@suse.cz>
6686 PR gcov-profile/83669
6687 * gcov.c (output_intermediate_file): Add version to intermediate
6689 * doc/gcov.texi: Document new field 'version' in intermediate
6690 file format. Fix location of '-k' option of gcov command.
6692 2018-01-04 Martin Liska <mliska@suse.cz>
6695 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
6697 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6699 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
6701 2018-01-03 Martin Sebor <msebor@redhat.com>
6703 PR tree-optimization/83655
6704 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
6705 checking calls with invalid arguments.
6707 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6709 * tree-vect-stmts.c (vect_get_store_rhs): New function.
6710 (vectorizable_mask_load_store): Delete.
6711 (vectorizable_call): Return false for masked loads and stores.
6712 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
6713 instead of gimple_assign_rhs1.
6714 (vectorizable_load): Handle IFN_MASK_LOAD.
6715 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
6717 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6719 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
6721 (vectorizable_mask_load_store): ...here.
6722 (vectorizable_load): ...and here.
6724 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6726 * tree-vect-stmts.c (vect_build_all_ones_mask)
6727 (vect_build_zero_merge_argument): New functions, split out from...
6728 (vectorizable_load): ...here.
6730 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6732 * tree-vect-stmts.c (vect_check_store_rhs): New function,
6734 (vectorizable_mask_load_store): ...here.
6735 (vectorizable_store): ...and here.
6737 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6739 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
6741 (vectorizable_mask_load_store): ...here.
6743 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6745 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
6746 (vect_model_store_cost): Take a vec_load_store_type instead of a
6748 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
6749 (vect_model_store_cost): Take a vec_load_store_type instead of a
6751 (vectorizable_mask_load_store): Update accordingly.
6752 (vectorizable_store): Likewise.
6753 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
6755 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6757 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
6758 IFN_MASK_LOAD calls here rather than...
6759 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
6761 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6762 Alan Hayward <alan.hayward@arm.com>
6763 David Sherwood <david.sherwood@arm.com>
6765 * expmed.c (extract_bit_field_1): For vector extracts,
6766 fall back to extract_bit_field_as_subreg if vec_extract
6769 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6770 Alan Hayward <alan.hayward@arm.com>
6771 David Sherwood <david.sherwood@arm.com>
6773 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
6774 they are variable or constant sized.
6775 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
6776 slots for constant-sized data.
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 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
6783 handling COND_EXPRs with boolean comparisons, try to find a better
6784 basis for the mask type than the boolean itself.
6786 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6788 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
6789 is calculated and how it can be overridden.
6790 * genmodes.c (max_bitsize_mode_any_mode): New variable.
6791 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
6793 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
6796 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6797 Alan Hayward <alan.hayward@arm.com>
6798 David Sherwood <david.sherwood@arm.com>
6800 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
6801 Remove the mode argument.
6802 (aarch64_simd_valid_immediate): Remove the mode and inverse
6804 * config/aarch64/iterators.md (bitsize): New iterator.
6805 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
6806 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
6807 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
6808 aarch64_simd_valid_immediate.
6809 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
6810 (aarch64_reg_or_bic_imm): Likewise.
6811 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
6812 with an insn_type enum and msl with a modifier_type enum.
6813 Replace element_width with a scalar_mode. Change the shift
6814 to unsigned int. Add constructors for scalar_float_mode and
6815 scalar_int_mode elements.
6816 (aarch64_vect_float_const_representable_p): Delete.
6817 (aarch64_can_const_movi_rtx_p)
6818 (aarch64_simd_scalar_immediate_valid_for_move)
6819 (aarch64_simd_make_constant): Update call to
6820 aarch64_simd_valid_immediate.
6821 (aarch64_advsimd_valid_immediate_hs): New function.
6822 (aarch64_advsimd_valid_immediate): Likewise.
6823 (aarch64_simd_valid_immediate): Remove mode and inverse
6824 arguments. Rewrite to use the above. Use const_vec_duplicate_p
6825 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
6826 and aarch64_float_const_representable_p on the result.
6827 (aarch64_output_simd_mov_immediate): Remove mode argument.
6828 Update call to aarch64_simd_valid_immediate and use of
6829 simd_immediate_info.
6830 (aarch64_output_scalar_simd_mov_immediate): Update call
6833 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6834 Alan Hayward <alan.hayward@arm.com>
6835 David Sherwood <david.sherwood@arm.com>
6837 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
6838 (mode_nunits): Likewise CONST_MODE_NUNITS.
6839 * machmode.def (ADJUST_NUNITS): Document.
6840 * genmodes.c (mode_data::need_nunits_adj): New field.
6841 (blank_mode): Update accordingly.
6842 (adj_nunits): New variable.
6843 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
6845 (emit_mode_size_inline): Set need_bytesize_adj for all modes
6846 listed in adj_nunits.
6847 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
6848 listed in adj_nunits. Don't emit case statements for such modes.
6849 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
6850 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
6851 nothing if adj_nunits is nonnull.
6852 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
6853 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
6854 (emit_mode_fbit): Update use of print_maybe_const_decl.
6855 (emit_move_size): Likewise. Treat the array as non-const
6857 (emit_mode_adjustments): Handle adj_nunits.
6859 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6861 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
6862 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
6863 (VECTOR_MODES): Use it.
6864 (make_vector_modes): Take the prefix as an argument.
6866 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6867 Alan Hayward <alan.hayward@arm.com>
6868 David Sherwood <david.sherwood@arm.com>
6870 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
6871 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
6872 for MODE_VECTOR_BOOL.
6873 * machmode.def (VECTOR_BOOL_MODE): Document.
6874 * genmodes.c (VECTOR_BOOL_MODE): New macro.
6875 (make_vector_bool_mode): New function.
6876 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
6878 * lto-streamer-in.c (lto_input_mode_table): Likewise.
6879 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
6881 * stor-layout.c (int_mode_for_mode): Likewise.
6882 * tree.c (build_vector_type_for_mode): Likewise.
6883 * varasm.c (output_constant_pool_2): Likewise.
6884 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
6885 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
6886 for MODE_VECTOR_BOOL.
6887 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
6888 of mode class checks.
6889 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
6890 instead of a list of mode class checks.
6891 (expand_vector_scalar_condition): Likewise.
6892 (type_for_widest_vector_mode): Handle BImode as an inner mode.
6894 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6895 Alan Hayward <alan.hayward@arm.com>
6896 David Sherwood <david.sherwood@arm.com>
6898 * machmode.h (mode_size): Change from unsigned short to
6900 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
6901 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6902 or if measurement_type is not polynomial.
6903 (fixed_size_mode::includes_p): Check for constant-sized modes.
6904 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
6905 return a poly_uint16 rather than an unsigned short.
6906 (emit_mode_size): Change the type of mode_size from unsigned short
6907 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
6908 (emit_mode_adjustments): Cope with polynomial vector sizes.
6909 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6911 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6913 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
6914 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
6915 * caller-save.c (setup_save_areas): Likewise.
6916 (replace_reg_with_saved_mem): Likewise.
6917 * calls.c (emit_library_call_value_1): Likewise.
6918 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
6919 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
6920 (gen_lowpart_for_combine): Likewise.
6921 * convert.c (convert_to_integer_1): Likewise.
6922 * cse.c (equiv_constant, cse_insn): Likewise.
6923 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
6924 (cselib_subst_to_values): Likewise.
6925 * dce.c (word_dce_process_block): Likewise.
6926 * df-problems.c (df_word_lr_mark_ref): Likewise.
6927 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
6928 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
6929 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
6930 (rtl_for_decl_location): Likewise.
6931 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
6932 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
6933 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
6934 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
6935 (expand_expr_real_1): Likewise.
6936 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
6937 (pad_below): Likewise.
6938 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6939 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
6940 * ira.c (get_subreg_tracking_sizes): Likewise.
6941 * ira-build.c (ira_create_allocno_objects): Likewise.
6942 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
6943 (ira_sort_regnos_for_alter_reg): Likewise.
6944 * ira-costs.c (record_operand_costs): Likewise.
6945 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
6946 (resolve_simple_move): Likewise.
6947 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
6948 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
6949 (lra_constraints): Likewise.
6950 (CONST_POOL_OK_P): Reject variable-sized modes.
6951 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
6952 (add_pseudo_to_slot, lra_spill): Likewise.
6953 * omp-low.c (omp_clause_aligned_alignment): Likewise.
6954 * optabs-query.c (get_best_extraction_insn): Likewise.
6955 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
6956 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
6957 (expand_mult_highpart, valid_multiword_target_p): Likewise.
6958 * recog.c (offsettable_address_addr_space_p): Likewise.
6959 * regcprop.c (maybe_mode_change): Likewise.
6960 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
6961 * regrename.c (build_def_use): Likewise.
6962 * regstat.c (dump_reg_info): Likewise.
6963 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
6964 (find_reloads, find_reloads_subreg_address): Likewise.
6965 * reload1.c (eliminate_regs_1): Likewise.
6966 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
6967 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
6968 (simplify_binary_operation_1, simplify_subreg): Likewise.
6969 * targhooks.c (default_function_arg_padding): Likewise.
6970 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
6971 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
6972 (verify_gimple_assign_ternary): Likewise.
6973 * tree-inline.c (estimate_move_cost): Likewise.
6974 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
6975 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
6976 (get_address_cost_ainc): Likewise.
6977 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
6978 (vect_supportable_dr_alignment): Likewise.
6979 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6980 (vectorizable_reduction): Likewise.
6981 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
6982 (vectorizable_operation, vectorizable_load): Likewise.
6983 * tree.c (build_same_sized_truth_vector_type): Likewise.
6984 * valtrack.c (cleanup_auto_inc_dec): Likewise.
6985 * var-tracking.c (emit_note_insn_var_location): Likewise.
6986 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
6987 (ADDR_VEC_ALIGN): Likewise.
6989 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6990 Alan Hayward <alan.hayward@arm.com>
6991 David Sherwood <david.sherwood@arm.com>
6993 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
6995 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6996 or if measurement_type is polynomial.
6997 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
6998 * combine.c (make_extraction): Likewise.
6999 * dse.c (find_shift_sequence): Likewise.
7000 * dwarf2out.c (mem_loc_descriptor): Likewise.
7001 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
7002 (extract_bit_field, extract_low_bits): Likewise.
7003 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
7004 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
7005 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
7006 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
7007 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
7008 * reload.c (find_reloads): Likewise.
7009 * reload1.c (alter_reg): Likewise.
7010 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
7011 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
7012 * tree-if-conv.c (predicate_mem_writes): Likewise.
7013 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
7014 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
7015 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
7016 * valtrack.c (dead_debug_insert_temp): Likewise.
7017 * varasm.c (mergeable_constant_section): Likewise.
7018 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
7020 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7021 Alan Hayward <alan.hayward@arm.com>
7022 David Sherwood <david.sherwood@arm.com>
7024 * expr.c (expand_assignment): Cope with polynomial mode sizes
7025 when assigning to a CONCAT.
7027 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7028 Alan Hayward <alan.hayward@arm.com>
7029 David Sherwood <david.sherwood@arm.com>
7031 * machmode.h (mode_precision): Change from unsigned short to
7033 (mode_to_precision): Return a poly_uint16 rather than an unsigned
7035 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
7036 or if measurement_type is not polynomial.
7037 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
7038 in which the mode is already known to be a scalar_int_mode.
7039 * genmodes.c (emit_mode_precision): Change the type of mode_precision
7040 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
7042 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7043 for GET_MODE_PRECISION.
7044 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7045 for GET_MODE_PRECISION.
7046 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
7048 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
7049 (expand_field_assignment, make_extraction): Likewise.
7050 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
7051 (get_last_value): Likewise.
7052 * convert.c (convert_to_integer_1): Likewise.
7053 * cse.c (cse_insn): Likewise.
7054 * expr.c (expand_expr_real_1): Likewise.
7055 * lra-constraints.c (simplify_operand_subreg): Likewise.
7056 * optabs-query.c (can_atomic_load_p): Likewise.
7057 * optabs.c (expand_atomic_load): Likewise.
7058 (expand_atomic_store): Likewise.
7059 * ree.c (combine_reaching_defs): Likewise.
7060 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
7061 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
7062 * tree.h (type_has_mode_precision_p): Likewise.
7063 * ubsan.c (instrument_si_overflow): Likewise.
7065 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7066 Alan Hayward <alan.hayward@arm.com>
7067 David Sherwood <david.sherwood@arm.com>
7069 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
7070 polynomial numbers of units.
7071 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
7072 (valid_vector_subparts_p): New function.
7073 (build_vector_type): Remove temporary shim and take the number
7074 of units as a poly_uint64 rather than an int.
7075 (build_opaque_vector_type): Take the number of units as a
7076 poly_uint64 rather than an int.
7077 * tree.c (build_vector_from_ctor): Handle polynomial
7078 TYPE_VECTOR_SUBPARTS.
7079 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
7080 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
7081 (build_vector_from_val): If the number of units is variable,
7082 use build_vec_duplicate_cst for constant operands and
7083 VEC_DUPLICATE_EXPR otherwise.
7084 (make_vector_type): Remove temporary is_constant ().
7085 (build_vector_type, build_opaque_vector_type): Take the number of
7086 units as a poly_uint64 rather than an int.
7087 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
7089 * cfgexpand.c (expand_debug_expr): Likewise.
7090 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
7091 (store_constructor, expand_expr_real_1): Likewise.
7092 (const_scalar_mask_from_tree): Likewise.
7093 * fold-const-call.c (fold_const_reduction): Likewise.
7094 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
7095 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
7096 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
7097 (fold_relational_const): Likewise.
7098 (native_interpret_vector): Likewise. Change the size from an
7099 int to an unsigned int.
7100 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
7101 TYPE_VECTOR_SUBPARTS.
7102 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
7103 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
7104 duplicating a non-constant operand into a variable-length vector.
7105 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
7106 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
7107 * ipa-icf.c (sem_variable::equals): Likewise.
7108 * match.pd: Likewise.
7109 * omp-simd-clone.c (simd_clone_subparts): Likewise.
7110 * print-tree.c (print_node): Likewise.
7111 * stor-layout.c (layout_type): Likewise.
7112 * targhooks.c (default_builtin_vectorization_cost): Likewise.
7113 * tree-cfg.c (verify_gimple_comparison): Likewise.
7114 (verify_gimple_assign_binary): Likewise.
7115 (verify_gimple_assign_ternary): Likewise.
7116 (verify_gimple_assign_single): Likewise.
7117 * tree-pretty-print.c (dump_generic_node): Likewise.
7118 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7119 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
7120 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
7121 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
7122 (vect_shift_permute_load_chain): Likewise.
7123 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
7124 (expand_vector_condition, optimize_vector_constructor): Likewise.
7125 (lower_vec_perm, get_compute_type): Likewise.
7126 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7127 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
7128 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
7129 (vect_recog_mask_conversion_pattern): Likewise.
7130 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
7131 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
7132 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7133 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
7134 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
7135 (vectorizable_shift, vectorizable_operation, vectorizable_store)
7136 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
7137 (supportable_widening_operation): Likewise.
7138 (supportable_narrowing_operation): Likewise.
7139 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
7141 * varasm.c (output_constant): Likewise.
7143 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7144 Alan Hayward <alan.hayward@arm.com>
7145 David Sherwood <david.sherwood@arm.com>
7147 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
7148 so that both the length == 3 and length != 3 cases set up their
7149 own permute vectors. Add comments explaining why we know the
7150 number of elements is constant.
7151 (vect_permute_load_chain): Likewise.
7153 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7154 Alan Hayward <alan.hayward@arm.com>
7155 David Sherwood <david.sherwood@arm.com>
7157 * machmode.h (mode_nunits): Change from unsigned char to
7159 (ONLY_FIXED_SIZE_MODES): New macro.
7160 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
7161 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
7162 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
7164 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
7165 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
7166 or if measurement_type is not polynomial.
7167 * genmodes.c (ZERO_COEFFS): New macro.
7168 (emit_mode_nunits_inline): Make mode_nunits_inline return a
7170 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
7171 Use ZERO_COEFFS when emitting initializers.
7172 * data-streamer.h (bp_pack_poly_value): New function.
7173 (bp_unpack_poly_value): Likewise.
7174 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7175 for GET_MODE_NUNITS.
7176 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7177 for GET_MODE_NUNITS.
7178 * tree.c (make_vector_type): Remove temporary shim and make
7179 the real function take the number of units as a poly_uint64
7181 (build_vector_type_for_mode): Handle polynomial nunits.
7182 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
7183 * emit-rtl.c (const_vec_series_p_1): Likewise.
7184 (gen_rtx_CONST_VECTOR): Likewise.
7185 * fold-const.c (test_vec_duplicate_folding): Likewise.
7186 * genrecog.c (validate_pattern): Likewise.
7187 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
7188 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7189 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
7190 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
7191 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
7192 * rtlanal.c (subreg_get_info): Likewise.
7193 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7194 (vect_grouped_load_supported): Likewise.
7195 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
7196 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
7197 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
7198 (simplify_const_unary_operation, simplify_binary_operation_1)
7199 (simplify_const_binary_operation, simplify_ternary_operation)
7200 (test_vector_ops_duplicate, test_vector_ops): Likewise.
7201 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
7202 instead of CONST_VECTOR_NUNITS.
7203 * varasm.c (output_constant_pool_2): Likewise.
7204 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
7205 explicit-encoded elements in the XVEC for variable-length vectors.
7207 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7209 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
7211 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7212 Alan Hayward <alan.hayward@arm.com>
7213 David Sherwood <david.sherwood@arm.com>
7215 * coretypes.h (fixed_size_mode): Declare.
7216 (fixed_size_mode_pod): New typedef.
7217 * builtins.h (target_builtins::x_apply_args_mode)
7218 (target_builtins::x_apply_result_mode): Change type to
7219 fixed_size_mode_pod.
7220 * builtins.c (apply_args_size, apply_result_size, result_vector)
7221 (expand_builtin_apply_args_1, expand_builtin_apply)
7222 (expand_builtin_return): Update accordingly.
7224 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7226 * cse.c (hash_rtx_cb): Hash only the encoded elements.
7227 * cselib.c (cselib_hash_rtx): Likewise.
7228 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
7229 CONST_VECTOR encoding.
7231 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7232 Jeff Law <law@redhat.com>
7235 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
7236 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
7237 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
7238 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
7241 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
7242 explicitly probe *sp in a noreturn function if there were any callee
7243 register saves or frame pointer is needed.
7245 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7248 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
7249 BLKmode for ternary, binary or unary expressions.
7252 * var-tracking.c (delete_vta_debug_insn): New inline function.
7253 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
7254 insns from get_insns () to NULL instead of each bb separately.
7255 Use delete_vta_debug_insn. No longer static.
7256 (vt_debug_insns_local, variable_tracking_main_1): Adjust
7257 delete_vta_debug_insns callers.
7258 * rtl.h (delete_vta_debug_insns): Declare.
7259 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
7260 instead of variable_tracking_main.
7262 2018-01-03 Martin Sebor <msebor@redhat.com>
7264 PR tree-optimization/83603
7265 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
7266 arguments past the endof the argument list in functions declared
7267 without a prototype.
7268 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
7269 Avoid checking when arguments are null.
7271 2018-01-03 Martin Sebor <msebor@redhat.com>
7274 * doc/extend.texi (attribute const): Fix a typo.
7275 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
7276 issuing -Wsuggest-attribute for void functions.
7278 2018-01-03 Martin Sebor <msebor@redhat.com>
7280 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
7281 offset_int::from instead of wide_int::to_shwi.
7282 (maybe_diag_overlap): Remove assertion.
7283 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
7284 * gimple-ssa-sprintf.c (format_directive): Same.
7285 (parse_directive): Same.
7286 (sprintf_dom_walker::compute_format_length): Same.
7287 (try_substitute_return_value): Same.
7289 2018-01-03 Jeff Law <law@redhat.com>
7292 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
7293 non-constant residual for zero at runtime and avoid probing in
7294 that case. Reorganize code for trailing problem to mirror handling
7297 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7299 PR tree-optimization/83501
7300 * tree-ssa-strlen.c (get_string_cst): New.
7301 (handle_char_store): Call get_string_cst.
7303 2018-01-03 Martin Liska <mliska@suse.cz>
7305 PR tree-optimization/83593
7306 * tree-ssa-strlen.c: Include tree-cfg.h.
7307 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
7308 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
7309 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
7311 (strlen_dom_walker::before_dom_children): Call
7312 gimple_purge_dead_eh_edges. Dump tranformation with details
7314 (strlen_dom_walker::before_dom_children): Update call by adding
7315 new argument cleanup_eh.
7316 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
7318 2018-01-03 Martin Liska <mliska@suse.cz>
7321 * cif-code.def (VARIADIC_THUNK): New enum value.
7322 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
7325 2018-01-03 Jan Beulich <jbeulich@suse.com>
7327 * sse.md (mov<mode>_internal): Tighten condition for when to use
7328 vmovdqu<ssescalarsize> for TI and OI modes.
7330 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7332 Update copyright years.
7334 2018-01-03 Martin Liska <mliska@suse.cz>
7337 * ipa-visibility.c (function_and_variable_visibility): Skip
7338 functions with noipa attribure.
7340 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7342 * gcc.c (process_command): Update copyright notice dates.
7343 * gcov-dump.c (print_version): Ditto.
7344 * gcov.c (print_version): Ditto.
7345 * gcov-tool.c (print_version): Ditto.
7346 * gengtype.c (create_file): Ditto.
7347 * doc/cpp.texi: Bump @copying's copyright year.
7348 * doc/cppinternals.texi: Ditto.
7349 * doc/gcc.texi: Ditto.
7350 * doc/gccint.texi: Ditto.
7351 * doc/gcov.texi: Ditto.
7352 * doc/install.texi: Ditto.
7353 * doc/invoke.texi: Ditto.
7355 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7357 * vector-builder.h (vector_builder::m_full_nelts): Change from
7358 unsigned int to poly_uint64.
7359 (vector_builder::full_nelts): Update prototype accordingly.
7360 (vector_builder::new_vector): Likewise.
7361 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
7362 (vector_builder::operator ==): Likewise.
7363 (vector_builder::finalize): Likewise.
7364 * int-vector-builder.h (int_vector_builder::int_vector_builder):
7365 Take the number of elements as a poly_uint64 rather than an
7367 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
7368 from unsigned int to poly_uint64.
7369 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
7370 (vec_perm_indices::new_vector): Likewise.
7371 (vec_perm_indices::length): Likewise.
7372 (vec_perm_indices::nelts_per_input): Likewise.
7373 (vec_perm_indices::input_nelts): Likewise.
7374 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
7375 number of elements per input as a poly_uint64 rather than an
7376 unsigned int. Use the original encoding for variable-length
7377 vectors, rather than clamping each individual element.
7378 For the second and subsequent elements in each pattern,
7379 clamp the step and base before clamping their sum.
7380 (vec_perm_indices::series_p): Handle polynomial element counts.
7381 (vec_perm_indices::all_in_range_p): Likewise.
7382 (vec_perm_indices_to_tree): Likewise.
7383 (vec_perm_indices_to_rtx): Likewise.
7384 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
7385 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
7386 (tree_vector_builder::new_binary_operation): Handle polynomial
7387 element counts. Return false if we need to know the number
7388 of elements at compile time.
7389 * fold-const.c (fold_vec_perm): Punt if the number of elements
7390 isn't known at compile time.
7392 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7394 * vec-perm-indices.h (vec_perm_builder): Change element type
7395 from HOST_WIDE_INT to poly_int64.
7396 (vec_perm_indices::element_type): Update accordingly.
7397 (vec_perm_indices::clamp): Handle polynomial element_types.
7398 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7399 (vec_perm_indices::all_in_range_p): Likewise.
7400 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
7402 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
7403 polynomial vec_perm_indices element types.
7404 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
7405 * fold-const.c (fold_vec_perm): Likewise.
7406 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
7407 * tree-vect-generic.c (lower_vec_perm): Likewise.
7408 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7409 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
7410 element type to HOST_WIDE_INT.
7412 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7413 Alan Hayward <alan.hayward@arm.com>
7414 David Sherwood <david.sherwood@arm.com>
7416 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
7417 rather than an int. Use plus_constant.
7418 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
7419 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
7421 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7422 Alan Hayward <alan.hayward@arm.com>
7423 David Sherwood <david.sherwood@arm.com>
7425 * calls.c (emit_call_1, expand_call): Change struct_value_size from
7426 a HOST_WIDE_INT to a poly_int64.
7428 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7429 Alan Hayward <alan.hayward@arm.com>
7430 David Sherwood <david.sherwood@arm.com>
7432 * calls.c (load_register_parameters): Cope with polynomial
7433 mode sizes. Require a constant size for BLKmode parameters
7434 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
7435 forces a parameter to be padded at the lsb end in order to
7436 fill a complete number of words, require the parameter size
7437 to be ordered wrt UNITS_PER_WORD.
7439 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7440 Alan Hayward <alan.hayward@arm.com>
7441 David Sherwood <david.sherwood@arm.com>
7443 * reload1.c (spill_stack_slot_width): Change element type
7444 from unsigned int to poly_uint64_pod.
7445 (alter_reg): Treat mode sizes as polynomial.
7447 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7448 Alan Hayward <alan.hayward@arm.com>
7449 David Sherwood <david.sherwood@arm.com>
7451 * reload.c (complex_word_subreg_p): New function.
7452 (reload_inner_reg_of_subreg, push_reload): Use it.
7454 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7455 Alan Hayward <alan.hayward@arm.com>
7456 David Sherwood <david.sherwood@arm.com>
7458 * lra-constraints.c (process_alt_operands): Reject matched
7459 operands whose sizes aren't ordered.
7460 (match_reload): Refer to this check here.
7462 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7463 Alan Hayward <alan.hayward@arm.com>
7464 David Sherwood <david.sherwood@arm.com>
7466 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
7467 that the mode size is in the set {1, 2, 4, 8, 16}.
7469 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7470 Alan Hayward <alan.hayward@arm.com>
7471 David Sherwood <david.sherwood@arm.com>
7473 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
7474 Use plus_constant instead of gen_rtx_PLUS.
7476 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7477 Alan Hayward <alan.hayward@arm.com>
7478 David Sherwood <david.sherwood@arm.com>
7480 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
7481 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
7482 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
7483 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
7484 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
7485 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
7486 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
7487 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
7488 * config/i386/i386.c (ix86_push_rounding): ...this new function.
7489 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
7491 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
7492 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
7493 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
7494 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
7495 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
7496 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
7497 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
7498 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
7499 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
7500 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
7502 * expr.c (emit_move_resolve_push): Treat the input and result
7503 of PUSH_ROUNDING as a poly_int64.
7504 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
7505 (emit_push_insn): Likewise.
7506 * lra-eliminations.c (mark_not_eliminable): Likewise.
7507 * recog.c (push_operand): Likewise.
7508 * reload1.c (elimination_effects): Likewise.
7509 * rtlanal.c (nonzero_bits1): Likewise.
7510 * calls.c (store_one_arg): Likewise. Require the padding to be
7511 known at compile time.
7513 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7514 Alan Hayward <alan.hayward@arm.com>
7515 David Sherwood <david.sherwood@arm.com>
7517 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
7518 Use plus_constant instead of gen_rtx_PLUS.
7520 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7521 Alan Hayward <alan.hayward@arm.com>
7522 David Sherwood <david.sherwood@arm.com>
7524 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
7527 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7528 Alan Hayward <alan.hayward@arm.com>
7529 David Sherwood <david.sherwood@arm.com>
7531 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
7532 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
7533 via stack temporaries. Treat the mode size as polynomial too.
7535 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7536 Alan Hayward <alan.hayward@arm.com>
7537 David Sherwood <david.sherwood@arm.com>
7539 * expr.c (expand_expr_real_2): When handling conversions involving
7540 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
7541 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
7542 as a poly_uint64 too.
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 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
7550 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7551 Alan Hayward <alan.hayward@arm.com>
7552 David Sherwood <david.sherwood@arm.com>
7554 * combine.c (can_change_dest_mode): Handle polynomial
7555 REGMODE_NATURAL_SIZE.
7556 * expmed.c (store_bit_field_1): Likewise.
7557 * expr.c (store_constructor): Likewise.
7558 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
7559 and polynomial REGMODE_NATURAL_SIZE.
7560 (gen_lowpart_common): Likewise.
7561 * reginfo.c (record_subregs_of_mode): Likewise.
7562 * rtlanal.c (read_modify_subreg_p): Likewise.
7564 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7565 Alan Hayward <alan.hayward@arm.com>
7566 David Sherwood <david.sherwood@arm.com>
7568 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
7569 numbers of elements.
7571 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7572 Alan Hayward <alan.hayward@arm.com>
7573 David Sherwood <david.sherwood@arm.com>
7575 * match.pd: Cope with polynomial numbers of vector elements.
7577 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7578 Alan Hayward <alan.hayward@arm.com>
7579 David Sherwood <david.sherwood@arm.com>
7581 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
7582 in a POINTER_PLUS_EXPR.
7584 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7585 Alan Hayward <alan.hayward@arm.com>
7586 David Sherwood <david.sherwood@arm.com>
7588 * omp-simd-clone.c (simd_clone_subparts): New function.
7589 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
7590 (ipa_simd_modify_function_body): Likewise.
7592 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7593 Alan Hayward <alan.hayward@arm.com>
7594 David Sherwood <david.sherwood@arm.com>
7596 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
7597 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
7598 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
7599 (expand_vector_condition, vector_element): Likewise.
7600 (subparts_gt): New function.
7601 (get_compute_type): Use subparts_gt.
7602 (count_type_subparts): Delete.
7603 (expand_vector_operations_1): Use subparts_gt instead of
7604 count_type_subparts.
7606 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7607 Alan Hayward <alan.hayward@arm.com>
7608 David Sherwood <david.sherwood@arm.com>
7610 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
7611 (vect_compile_time_alias): ...this new function. Do the calculation
7612 on poly_ints rather than trees.
7613 (vect_prune_runtime_alias_test_list): Update call accordingly.
7615 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7616 Alan Hayward <alan.hayward@arm.com>
7617 David Sherwood <david.sherwood@arm.com>
7619 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
7621 (vect_schedule_slp_instance): Likewise.
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-slp.c (vect_get_and_check_slp_defs): Reject
7628 constant and extern definitions for variable-length vectors.
7629 (vect_get_constant_vectors): Note that the number of units
7630 is known to be constant.
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-stmts.c (vectorizable_conversion): Treat the number
7637 of units as polynomial. Choose between WIDE and NARROW based
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-stmts.c (simd_clone_subparts): New function.
7645 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
7647 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7648 Alan Hayward <alan.hayward@arm.com>
7649 David Sherwood <david.sherwood@arm.com>
7651 * tree-vect-stmts.c (vectorizable_call): Treat the number of
7652 vectors as polynomial. Use build_index_vector for
7655 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7656 Alan Hayward <alan.hayward@arm.com>
7657 David Sherwood <david.sherwood@arm.com>
7659 * tree-vect-stmts.c (get_load_store_type): Treat the number of
7660 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
7661 for variable-length vectors.
7662 (vectorizable_mask_load_store): Treat the number of units as
7663 polynomial, asserting that it is constant if the condition has
7664 already been enforced.
7665 (vectorizable_store, vectorizable_load): Likewise.
7667 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7668 Alan Hayward <alan.hayward@arm.com>
7669 David Sherwood <david.sherwood@arm.com>
7671 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
7672 of units as polynomial. Punt if we can't tell at compile time
7673 which vector contains the final result.
7675 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7676 Alan Hayward <alan.hayward@arm.com>
7677 David Sherwood <david.sherwood@arm.com>
7679 * tree-vect-loop.c (vectorizable_induction): Treat the number
7680 of units as polynomial. Punt on SLP inductions. Use an integer
7681 VEC_SERIES_EXPR for variable-length integer reductions. Use a
7682 cast of such a series for variable-length floating-point
7685 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7686 Alan Hayward <alan.hayward@arm.com>
7687 David Sherwood <david.sherwood@arm.com>
7689 * tree.h (build_index_vector): Declare.
7690 * tree.c (build_index_vector): New function.
7691 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
7692 of units as polynomial, forcibly converting it to a constant if
7693 vectorizable_reduction has already enforced the condition.
7694 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
7695 to create a {1,2,3,...} vector.
7696 (vectorizable_reduction): Treat the number of units as polynomial.
7697 Choose vectype_in based on the largest scalar element size rather
7698 than the smallest number of units. Enforce the restrictions
7701 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7702 Alan Hayward <alan.hayward@arm.com>
7703 David Sherwood <david.sherwood@arm.com>
7705 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
7706 number of units as polynomial.
7708 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7709 Alan Hayward <alan.hayward@arm.com>
7710 David Sherwood <david.sherwood@arm.com>
7712 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
7713 * target.def (autovectorize_vector_sizes): Return the vector sizes
7714 by pointer, using vector_sizes rather than a bitmask.
7715 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
7716 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
7717 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
7719 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
7720 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
7721 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
7722 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
7723 * omp-general.c (omp_max_vf): Likewise.
7724 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7725 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
7726 * tree-vect-loop.c (vect_analyze_loop): Likewise.
7727 * tree-vect-slp.c (vect_slp_bb): Likewise.
7728 * doc/tm.texi: Regenerate.
7729 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
7731 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
7732 the vector size as a poly_uint64 rather than an unsigned int.
7733 (current_vector_size): Change from an unsigned int to a poly_uint64.
7734 (get_vectype_for_scalar_type): Update accordingly.
7735 * tree.h (build_truth_vector_type): Take the size and number of
7736 units as a poly_uint64 rather than an unsigned int.
7737 (build_vector_type): Add a temporary overload that takes
7738 the number of units as a poly_uint64 rather than an unsigned int.
7739 * tree.c (make_vector_type): Likewise.
7740 (build_truth_vector_type): Take the number of units as a poly_uint64
7741 rather than an unsigned int.
7743 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7744 Alan Hayward <alan.hayward@arm.com>
7745 David Sherwood <david.sherwood@arm.com>
7747 * target.def (get_mask_mode): Take the number of units and length
7748 as poly_uint64s rather than unsigned ints.
7749 * targhooks.h (default_get_mask_mode): Update accordingly.
7750 * targhooks.c (default_get_mask_mode): Likewise.
7751 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
7752 * doc/tm.texi: Regenerate.
7754 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7755 Alan Hayward <alan.hayward@arm.com>
7756 David Sherwood <david.sherwood@arm.com>
7758 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
7759 * omp-general.c (omp_max_vf): Likewise.
7760 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
7761 (expand_omp_simd): Handle polynomial safelen.
7762 * omp-low.c (omplow_simd_context): Add a default constructor.
7763 (omplow_simd_context::max_vf): Change from int to poly_uint64.
7764 (lower_rec_simd_input_clauses): Update accordingly.
7765 (lower_rec_input_clauses): Likewise.
7767 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7768 Alan Hayward <alan.hayward@arm.com>
7769 David Sherwood <david.sherwood@arm.com>
7771 * tree-vectorizer.h (vect_nunits_for_cost): New function.
7772 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
7773 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
7774 (vect_analyze_slp_cost): Likewise.
7775 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
7776 (vect_model_load_cost): Likewise.
7778 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7779 Alan Hayward <alan.hayward@arm.com>
7780 David Sherwood <david.sherwood@arm.com>
7782 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
7783 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
7784 from an unsigned int * to a poly_uint64_pod *.
7785 (calculate_unrolling_factor): New function.
7786 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
7788 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7789 Alan Hayward <alan.hayward@arm.com>
7790 David Sherwood <david.sherwood@arm.com>
7792 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
7793 from an unsigned int to a poly_uint64.
7794 (_loop_vec_info::slp_unrolling_factor): Likewise.
7795 (_loop_vec_info::vectorization_factor): Change from an int
7797 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
7798 (vect_get_num_vectors): New function.
7799 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
7800 (vect_get_num_copies): Use vect_get_num_vectors.
7801 (vect_analyze_data_ref_dependences): Change max_vf from an int *
7802 to an unsigned int *.
7803 (vect_analyze_data_refs): Change min_vf from an int * to a
7805 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7806 than an unsigned HOST_WIDE_INT.
7807 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
7808 (vect_analyze_data_ref_dependence): Change max_vf from an int *
7809 to an unsigned int *.
7810 (vect_analyze_data_ref_dependences): Likewise.
7811 (vect_compute_data_ref_alignment): Handle polynomial vf.
7812 (vect_enhance_data_refs_alignment): Likewise.
7813 (vect_prune_runtime_alias_test_list): Likewise.
7814 (vect_shift_permute_load_chain): Likewise.
7815 (vect_supportable_dr_alignment): Likewise.
7816 (dependence_distance_ge_vf): Take the vectorization factor as a
7817 poly_uint64 rather than an unsigned HOST_WIDE_INT.
7818 (vect_analyze_data_refs): Change min_vf from an int * to a
7820 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
7821 vfm1 as a poly_uint64 rather than an int. Make the same change
7822 for the returned bound_scalar.
7823 (vect_gen_vector_loop_niters): Handle polynomial vf.
7824 (vect_do_peeling): Likewise. Update call to
7825 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
7826 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
7828 * tree-vect-loop.c (vect_determine_vectorization_factor)
7829 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
7830 (vect_get_known_peeling_cost): Likewise.
7831 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
7832 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
7833 (vect_transform_loop): Likewise. Use the lowest possible VF when
7834 updating the upper bounds of the loop.
7835 (vect_min_worthwhile_factor): Make static. Return an unsigned int
7837 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
7838 polynomial unroll factors.
7839 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
7840 (vect_make_slp_decision): Likewise.
7841 (vect_supported_load_permutation_p): Likewise, and polynomial
7843 (vect_analyze_slp_cost): Handle polynomial vf.
7844 (vect_slp_analyze_node_operations): Likewise.
7845 (vect_slp_analyze_bb_1): Likewise.
7846 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7847 than an unsigned HOST_WIDE_INT.
7848 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
7849 (vectorizable_load): Handle polynomial vf.
7850 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
7852 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
7854 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7855 Alan Hayward <alan.hayward@arm.com>
7856 David Sherwood <david.sherwood@arm.com>
7858 * match.pd: Handle bit operations involving three constants
7859 and try to fold one pair.
7861 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7863 * tree-vect-loop-manip.c: Include gimple-fold.h.
7864 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
7865 niters_maybe_zero parameters. Handle other cases besides a step of 1.
7866 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
7867 Add a path that uses a step of VF instead of 1, but disable it
7869 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
7870 and niters_no_overflow parameters. Update calls to
7871 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
7872 Create a new SSA name if the latter choses to use a ste other
7873 than zero, and return it via niters_vector_mult_vf_var.
7874 * tree-vect-loop.c (vect_transform_loop): Update calls to
7875 vect_do_peeling, vect_gen_vector_loop_niters and
7876 slpeel_make_loop_iterate_ntimes.
7877 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
7878 (vect_gen_vector_loop_niters): Update declarations after above changes.
7880 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
7882 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
7883 128-bit round to integer instructions.
7884 (ceil<mode>2): Likewise.
7885 (btrunc<mode>2): Likewise.
7886 (round<mode>2): Likewise.
7888 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7890 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
7891 unaligned VSX load/store on P8/P9.
7892 (expand_block_clear): Allow the use of unaligned VSX
7893 load/store on P8/P9.
7895 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7897 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
7899 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
7900 swap associated with both a load and a store.
7902 2018-01-02 Andrew Waterman <andrew@sifive.com>
7904 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
7905 * config/riscv/riscv.md (clear_cache): Use it.
7907 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
7909 * web.c: Remove out-of-date comment.
7911 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7913 * expr.c (fixup_args_size_notes): Check that any existing
7914 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
7915 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
7916 (emit_single_push_insn): ...here.
7918 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7920 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
7921 (const_vector_encoded_nelts): New function.
7922 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
7923 (const_vector_int_elt, const_vector_elt): Declare.
7924 * emit-rtl.c (const_vector_int_elt_1): New function.
7925 (const_vector_elt): Likewise.
7926 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
7927 of CONST_VECTOR_ELT.
7929 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7931 * expr.c: Include rtx-vector-builder.h.
7932 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
7933 directly on the tree encoding.
7934 (const_vector_from_tree): Likewise.
7935 * optabs.c: Include rtx-vector-builder.h.
7936 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
7937 sequence of "u" values.
7938 * vec-perm-indices.c: Include rtx-vector-builder.h.
7939 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
7940 directly on the vec_perm_indices encoding.
7942 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7944 * doc/rtl.texi (const_vector): Describe new encoding scheme.
7945 * Makefile.in (OBJS): Add rtx-vector-builder.o.
7946 * rtx-vector-builder.h: New file.
7947 * rtx-vector-builder.c: Likewise.
7948 * rtl.h (rtx_def::u2): Add a const_vector field.
7949 (CONST_VECTOR_NPATTERNS): New macro.
7950 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
7951 (CONST_VECTOR_DUPLICATE_P): Likewise.
7952 (CONST_VECTOR_STEPPED_P): Likewise.
7953 (CONST_VECTOR_ENCODED_ELT): Likewise.
7954 (const_vec_duplicate_p): Check for a duplicated vector encoding.
7955 (unwrap_const_vec_duplicate): Likewise.
7956 (const_vec_series_p): Check for a non-duplicated vector encoding.
7957 Say that the function only returns true for integer vectors.
7958 * emit-rtl.c: Include rtx-vector-builder.h.
7959 (gen_const_vec_duplicate_1): Delete.
7960 (gen_const_vector): Call gen_const_vec_duplicate instead of
7961 gen_const_vec_duplicate_1.
7962 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
7963 (gen_const_vec_duplicate): Use rtx_vector_builder.
7964 (gen_const_vec_series): Likewise.
7965 (gen_rtx_CONST_VECTOR): Likewise.
7966 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
7967 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7968 Build a new vector rather than modifying a CONST_VECTOR in-place.
7969 (handle_special_swappables): Update call accordingly.
7970 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
7971 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7972 Build a new vector rather than modifying a CONST_VECTOR in-place.
7973 (handle_special_swappables): Update call accordingly.
7975 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7977 * simplify-rtx.c (simplify_const_binary_operation): Use
7978 CONST_VECTOR_ELT instead of XVECEXP.
7980 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7982 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
7983 the selector elements to be different from the data elements
7984 if the selector is a VECTOR_CST.
7985 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
7986 ssizetype for the selector.
7988 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7990 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
7991 before testing each element individually.
7992 * tree-vect-generic.c (lower_vec_perm): Likewise.
7994 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7996 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
7997 * selftest-run-tests.c (selftest::run_tests): Call it.
7998 * vector-builder.h (vector_builder::operator ==): New function.
7999 (vector_builder::operator !=): Likewise.
8000 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
8001 (vec_perm_indices::all_from_input_p): New function.
8002 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
8003 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
8004 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
8005 instead of reading the VECTOR_CST directly. Detect whether both
8006 vector inputs are the same before constructing the vec_perm_indices,
8007 and update the number of inputs argument accordingly. Use the
8008 utility functions added above. Only construct sel2 if we need to.
8010 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8012 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
8013 the broadcast of the low byte.
8014 (expand_mult_highpart): Use an explicit encoding for the permutes.
8015 * optabs-query.c (can_mult_highpart_p): Likewise.
8016 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
8017 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8018 (vectorizable_bswap): Likewise.
8019 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
8020 explicit encoding for the power-of-2 permutes.
8021 (vect_permute_store_chain): Likewise.
8022 (vect_grouped_load_supported): Likewise.
8023 (vect_permute_load_chain): Likewise.
8025 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8027 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
8028 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
8029 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
8030 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8031 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
8032 (vect_gen_perm_mask_any): Likewise.
8034 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8036 * int-vector-builder.h: New file.
8037 * vec-perm-indices.h: Include int-vector-builder.h.
8038 (vec_perm_indices): Redefine as an int_vector_builder.
8039 (auto_vec_perm_indices): Delete.
8040 (vec_perm_builder): Redefine as a stand-alone class.
8041 (vec_perm_indices::vec_perm_indices): New function.
8042 (vec_perm_indices::clamp): Likewise.
8043 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
8044 (vec_perm_indices::new_vector): New function.
8045 (vec_perm_indices::new_expanded_vector): Update for new
8046 vec_perm_indices class.
8047 (vec_perm_indices::rotate_inputs): New function.
8048 (vec_perm_indices::all_in_range_p): Operate directly on the
8049 encoded form, without computing elided elements.
8050 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
8051 encoding. Update for new vec_perm_indices class.
8052 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
8053 the given vec_perm_builder.
8054 (expand_vec_perm_var): Update vec_perm_builder constructor.
8055 (expand_mult_highpart): Use vec_perm_builder instead of
8056 auto_vec_perm_indices.
8057 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
8058 vec_perm_indices instead of auto_vec_perm_indices. Use a single
8059 or double series encoding as appropriate.
8060 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
8061 vec_perm_indices instead of auto_vec_perm_indices.
8062 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8063 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8064 (vect_permute_store_chain): Likewise.
8065 (vect_grouped_load_supported): Likewise.
8066 (vect_permute_load_chain): Likewise.
8067 (vect_shift_permute_load_chain): Likewise.
8068 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8069 (vect_transform_slp_perm_load): Likewise.
8070 (vect_schedule_slp_instance): Likewise.
8071 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8072 (vectorizable_mask_load_store): Likewise.
8073 (vectorizable_bswap): Likewise.
8074 (vectorizable_store): Likewise.
8075 (vectorizable_load): Likewise.
8076 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
8077 vec_perm_indices instead of auto_vec_perm_indices. Use
8078 tree_to_vec_perm_builder to read the vector from a tree.
8079 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
8080 vec_perm_builder instead of a vec_perm_indices.
8081 (have_whole_vector_shift): Use vec_perm_builder and
8082 vec_perm_indices instead of auto_vec_perm_indices. Leave the
8083 truncation to calc_vec_perm_mask_for_shift.
8084 (vect_create_epilog_for_reduction): Likewise.
8085 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
8086 from auto_vec_perm_indices to vec_perm_indices.
8087 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8088 instead of changing individual elements.
8089 (aarch64_vectorize_vec_perm_const): Use new_vector to install
8090 the vector in d.perm.
8091 * config/arm/arm.c (expand_vec_perm_d::perm): Change
8092 from auto_vec_perm_indices to vec_perm_indices.
8093 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8094 instead of changing individual elements.
8095 (arm_vectorize_vec_perm_const): Use new_vector to install
8096 the vector in d.perm.
8097 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
8098 Update vec_perm_builder constructor.
8099 (rs6000_expand_interleave): Likewise.
8100 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
8101 (rs6000_expand_interleave): Likewise.
8103 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8105 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
8106 to qimode could truncate the indices.
8107 * optabs.c (expand_vec_perm_var): Likewise.
8109 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8111 * Makefile.in (OBJS): Add vec-perm-indices.o.
8112 * vec-perm-indices.h: New file.
8113 * vec-perm-indices.c: Likewise.
8114 * target.h (vec_perm_indices): Replace with a forward class
8116 (auto_vec_perm_indices): Move to vec-perm-indices.h.
8117 * optabs.h: Include vec-perm-indices.h.
8118 (expand_vec_perm): Delete.
8119 (selector_fits_mode_p, expand_vec_perm_var): Declare.
8120 (expand_vec_perm_const): Declare.
8121 * target.def (vec_perm_const_ok): Replace with...
8122 (vec_perm_const): ...this new hook.
8123 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
8124 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
8125 * doc/tm.texi: Regenerate.
8126 * optabs.def (vec_perm_const): Delete.
8127 * doc/md.texi (vec_perm_const): Likewise.
8128 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
8129 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
8130 expand_vec_perm for constant permutation vectors. Assert that
8131 the mode of variable permutation vectors is the integer equivalent
8132 of the mode that is being permuted.
8133 * optabs-query.h (selector_fits_mode_p): Declare.
8134 * optabs-query.c: Include vec-perm-indices.h.
8135 (selector_fits_mode_p): New function.
8136 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
8137 is defined, instead of checking whether the vec_perm_const_optab
8138 exists. Use targetm.vectorize.vec_perm_const instead of
8139 targetm.vectorize.vec_perm_const_ok. Check whether the indices
8140 fit in the vector mode before using a variable permute.
8141 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
8142 vec_perm_indices instead of an rtx.
8143 (expand_vec_perm): Replace with...
8144 (expand_vec_perm_const): ...this new function. Take the selector
8145 as a vec_perm_indices rather than an rtx. Also take the mode of
8146 the selector. Update call to shift_amt_for_vec_perm_mask.
8147 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
8148 Use vec_perm_indices::new_expanded_vector to expand the original
8149 selector into bytes. Check whether the indices fit in the vector
8150 mode before using a variable permute.
8151 (expand_vec_perm_var): Make global.
8152 (expand_mult_highpart): Use expand_vec_perm_const.
8153 * fold-const.c: Includes vec-perm-indices.h.
8154 * tree-ssa-forwprop.c: Likewise.
8155 * tree-vect-data-refs.c: Likewise.
8156 * tree-vect-generic.c: Likewise.
8157 * tree-vect-loop.c: Likewise.
8158 * tree-vect-slp.c: Likewise.
8159 * tree-vect-stmts.c: Likewise.
8160 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
8162 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
8163 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
8164 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
8165 (aarch64_vectorize_vec_perm_const): ...this new function.
8166 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8167 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8168 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
8169 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
8170 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8171 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8172 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
8174 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
8175 check for NEON modes.
8176 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
8177 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
8178 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
8179 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
8181 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
8182 the old VEC_PERM_CONST conditions.
8183 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
8184 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
8185 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
8186 (ia64_vectorize_vec_perm_const_ok): Merge into...
8187 (ia64_vectorize_vec_perm_const): ...this new function.
8188 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
8189 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
8190 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
8191 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
8192 * config/mips/mips.c (mips_expand_vec_perm_const)
8193 (mips_vectorize_vec_perm_const_ok): Merge into...
8194 (mips_vectorize_vec_perm_const): ...this new function.
8195 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
8196 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
8197 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
8198 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
8199 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
8200 (rs6000_expand_vec_perm_const): Delete.
8201 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
8203 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8204 (altivec_expand_vec_perm_const_le): Take each operand individually.
8205 Operate on constant selectors rather than rtxes.
8206 (altivec_expand_vec_perm_const): Likewise. Update call to
8207 altivec_expand_vec_perm_const_le.
8208 (rs6000_expand_vec_perm_const): Delete.
8209 (rs6000_vectorize_vec_perm_const_ok): Delete.
8210 (rs6000_vectorize_vec_perm_const): New function.
8211 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8212 an element count and rtx array.
8213 (rs6000_expand_extract_even): Update call accordingly.
8214 (rs6000_expand_interleave): Likewise.
8215 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
8216 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
8217 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
8218 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
8219 (rs6000_expand_vec_perm_const): Delete.
8220 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8221 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8222 (altivec_expand_vec_perm_const_le): Take each operand individually.
8223 Operate on constant selectors rather than rtxes.
8224 (altivec_expand_vec_perm_const): Likewise. Update call to
8225 altivec_expand_vec_perm_const_le.
8226 (rs6000_expand_vec_perm_const): Delete.
8227 (rs6000_vectorize_vec_perm_const_ok): Delete.
8228 (rs6000_vectorize_vec_perm_const): New function. Remove stray
8229 reference to the SPE evmerge intructions.
8230 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8231 an element count and rtx array.
8232 (rs6000_expand_extract_even): Update call accordingly.
8233 (rs6000_expand_interleave): Likewise.
8234 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
8235 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
8237 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8239 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8241 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
8242 vector mode and that that mode matches the mode of the data
8244 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
8245 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
8246 directly using expand_vec_perm_1 when forcing selectors into
8248 (expand_vec_perm_var): New function, split out from expand_vec_perm.
8250 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8252 * optabs-query.h (can_vec_perm_p): Delete.
8253 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
8254 * optabs-query.c (can_vec_perm_p): Split into...
8255 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
8256 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
8257 particular selector is valid.
8258 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8259 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8260 (vect_grouped_load_supported): Likewise.
8261 (vect_shift_permute_load_chain): Likewise.
8262 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8263 (vect_transform_slp_perm_load): Likewise.
8264 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8265 (vectorizable_bswap): Likewise.
8266 (vect_gen_perm_mask_checked): Likewise.
8267 * fold-const.c (fold_ternary_loc): Likewise. Don't take
8268 implementations of variable permutation vectors into account
8269 when deciding which selector to use.
8270 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
8271 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
8272 with a false third argument.
8273 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
8274 to test whether the constant selector is valid and can_vec_perm_var_p
8275 to test whether a variable selector is valid.
8277 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8279 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
8280 * optabs-query.c (can_vec_perm_p): Likewise.
8281 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
8282 instead of vec_perm_indices.
8283 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
8284 (vect_gen_perm_mask_checked): Likewise,
8285 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
8286 (vect_gen_perm_mask_checked): Likewise,
8288 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8290 * optabs-query.h (qimode_for_vec_perm): Declare.
8291 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
8292 (qimode_for_vec_perm): ...this new function.
8293 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
8295 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8297 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
8298 does not have a conditional at the top.
8300 2018-01-02 Richard Biener <rguenther@suse.de>
8302 * ipa-inline.c (big_speedup_p): Fix expression.
8304 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8307 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
8310 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8314 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
8315 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
8316 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
8317 cond_taken_branch_cost 3->4.
8319 2018-01-01 Jakub Jelinek <jakub@redhat.com>
8321 PR tree-optimization/83581
8322 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
8323 TODO_cleanup_cfg if any changes have been made.
8326 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
8327 convert_modes if target mode has the right side, but different mode
8331 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
8332 last argument when extracting from CONCAT. If either from_real or
8333 from_imag is NULL, use expansion through memory. If result is not
8334 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
8335 the parts directly to inner mode, if even that fails, use expansion
8339 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
8340 check for bswap in mode rather than HImode and use that in expand_unop
8343 Copyright (C) 2018 Free Software Foundation, Inc.
8345 Copying and distribution of this file, with or without modification,
8346 are permitted in any medium without royalty provided the copyright
8347 notice and this notice are preserved.