1 2018-03-08 Jakub Jelinek <jakub@redhat.com>
3 PR tree-optimization/84740
4 * tree-switch-conversion.c (process_switch): Call build_constructors
5 only if info.phi_count is non-zero.
7 PR tree-optimization/84739
8 * tree-tailcall.c (find_tail_calls): Check call arguments against
9 DECL_ARGUMENTS (current_function_decl) rather than
10 DECL_ARGUMENTS (func) when checking for tail recursion.
12 2018-03-07 Jakub Jelinek <jakub@redhat.com>
14 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
15 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
16 Volker Reichelt's entry and add entries for people that perform
17 GCC fuzzy testing and report numerous bugs.
19 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
22 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
23 readonly data in sdata, if that is disabled.
24 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
25 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
26 -mreadonly-in-sdata option.
28 2018-03-07 Martin Sebor <msebor@redhat.com>
30 PR tree-optimization/84468
31 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
32 basic block when looking for nul assignment.
34 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
37 * except.h (output_function_exception_table): Adjust prototype.
38 * except.c (output_function_exception_table): Remove FNNAME parameter
39 and add SECTION parameter. Ouput one part of the table at a time.
40 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
41 the first part of the exception table and emit unwind directives.
42 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
43 (i386_pe_seh_cold_init): Likewise.
44 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
45 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
46 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
47 (ix86_output_call_insn): Emit a nop in one more case for SEH.
48 * config/i386/winnt.c: Include except.h.
49 (struct seh_frame_state): Add reg_offset, after_prologue and
50 in_cold_section fields.
51 (i386_pe_seh_end_prologue): Set seh->after_prologue.
52 (i386_pe_seh_cold_init): New function.
53 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
54 to seh->in_cold_section.
55 (seh_emit_push): Record the offset of the push.
56 (seh_emit_save): Record the offet of the save.
57 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
58 Test seh->after_prologue to disregard the epilogue.
59 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
60 (i386_pe_end_cold_function): New function.
62 2018-03-07 Jakub Jelinek <jakub@redhat.com>
65 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
66 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
69 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
71 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
72 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
75 * multiple_target.c: Include tree-inline.h and intl.h.
76 (expand_target_clones): Diagnose and fail if node->definition and
77 !tree_versionable_function_p (node->decl).
79 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
81 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
83 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
84 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
85 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
87 2018-03-06 Jakub Jelinek <jakub@redhat.com>
90 * combine.c (try_combine): Use reg_or_subregno instead of handling
91 just paradoxical SUBREGs and REGs.
93 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
95 * config/arc/arc.c (arc_finalize_pic): Remove function.
96 (arc_must_save_register): We use single base PIC register, remove
97 checks to save/restore the PIC register.
98 (arc_expand_prologue): Likewise.
99 * config/arc/arc-protos.h (arc_set_default_type_attributes):
101 (arc_verify_short): Likewise.
102 (arc_attr_type): Likewise.
103 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
104 (walk_stores): Likewise.
105 (arc_address_cost): Make it static.
106 (arc_verify_short): Likewise.
107 (branch_dest): Likewise.
108 (arc_attr_type): Likewise.
109 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
110 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
111 (arc_final_prescan_insn): Remove inserting the nops due to
112 hardware hazards. It is done in reorg step.
113 (insn_length_variant_t): Remove.
114 (insn_length_parameters_t): Likewise.
115 (arc_insn_length_parameters): Likewise.
116 (arc_get_insn_variants): Likewise.
117 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
119 2018-03-06 Jakub Jelinek <jakub@redhat.com>
122 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
125 PR tree-optimization/84687
126 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
128 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
130 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
132 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
133 Rename to ppc_speculation_barrier.
134 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
135 __builtin_ppc_speculation_barrier.
137 2018-03-05 Jakub Jelinek <jakub@redhat.com>
140 * combine.c (combine_simplify_rtx): Don't try to simplify if
141 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
144 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
146 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
147 to 32 bytes when compiling for POWER9.
149 2018-03-05 Jakub Jelinek <jakub@redhat.com>
152 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
153 regparm >= 3 with no arg reg available also for calls with
154 flag_force_indirect_call. Pass decl to ix86_function_regparm.
157 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
159 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
161 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
164 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
166 2018-03-05 Richard Biener <rguenther@suse.de>
168 PR tree-optimization/84486
169 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
170 When inserting a __builtin_assume_aligned call set the LHS
171 SSA name alignment info accordingly.
173 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
175 PR tree-optimization/84114
176 * config/aarch64/aarch64.c (aarch64_reassociation_width)
177 Avoid reassociation of FLOAT_MODE addition.
179 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
181 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
182 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
183 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
184 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
185 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
186 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
187 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
189 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
190 __builtin_ia32_wbinvd): New builtins.
191 (SPECIAL_ARGS2): New.
192 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
193 (SPECIAL_ARGS2): New.
194 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
195 (ix86_valid_target_attribute_inner_p): Ditto.
196 (ix86_init_mmx_sse_builtins): Add special_args2.
197 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
198 TARGET_WBNOINVD_P): New.
199 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
200 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
201 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
202 * config/i386/immintrin.h (_wbinvd): New intrinsic.
203 * config/i386/pconfigintrin.h: New file.
204 * config/i386/wbnoinvdintrin.h: Ditto.
205 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
206 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
208 2018-03-05 Richard Biener <rguenther@suse.de>
210 PR tree-optimization/84670
211 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
213 (BB_VISITED_WITH_VISITED_SUCCS): New define.
214 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
215 (compute_antic_aux): Only assert the number of values in ANTIC_IN
216 doesn't grow if all successors (recursively) were visited at least
219 2018-03-05 Richard Biener <rguenther@suse.de>
221 PR tree-optimization/84650
222 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
223 if executed in the loop pipeline.
225 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
227 * doc/configfiles.texi (Configuration Files): Move info about
228 conditionalizing $target-protos.h to...
229 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
230 differs from $target-protos.h.
232 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
233 Chung-Ju Wu <jasonwucj@gmail.com>
235 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
236 * config/nds32/nds32-multiple.md (setmemsi): Define.
237 * config/nds32/nds32-memory-manipulation.c
238 (nds32_gen_dup_4_byte_to_word_value): New.
239 (emit_setmem_word_loop): New.
240 (emit_setmem_byte_loop): New.
241 (nds32_expand_setmem_loop): New.
242 (nds32_expand_setmem_loop_v3m): New.
243 (nds32_expand_setmem_unroll): New.
244 (nds32_expand_setmem): New.
246 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
247 Chung-Ju Wu <jasonwucj@gmail.com>
249 * config/nds32/nds32-memory-manipulation.c
250 (nds32_emit_load_store): New.
251 (nds32_emit_post_inc_load_store): New.
252 (nds32_emit_mem_move): New.
253 (nds32_emit_mem_move_block): New.
254 (nds32_expand_movmemsi_loop_unknown_size): New.
255 (nds32_expand_movmemsi_loop_known_size): New.
256 (nds32_expand_movmemsi_loop): New.
257 (nds32_expand_movmemsi_unroll): New.
258 (nds32_expand_movmemqi): Rename ...
259 (nds32_expand_movmemsi): ... to this.
260 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
261 (movmemsi): ... to this.
262 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
263 (nds32_expand_movmemsi): ... to this.
265 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
266 Monk Chiang <sh.chiang04@gmail.com>
267 Chung-Ju Wu <jasonwucj@gmail.com>
269 * config/nds32/nds32-protos.h
270 (nds32_expand_load_multiple): New arguments.
271 (nds32_expand_store_multiple): Ditto.
272 (nds32_valid_multiple_load_store): Rename ...
273 (nds32_valid_multiple_load_store_p): ... to this.
274 * config/nds32/nds32-memory-manipulation.c
275 (nds32_expand_load_multiple): Refine implementation.
276 (nds32_expand_store_multiple): Ditto.
277 * config/nds32/nds32-multiple.md
278 (load_multiple): Update nds32_expand_load_multiple interface.
279 (store_multiple): Update nds32_expand_store_multiple interface.
280 * config/nds32/nds32-predicates.c
281 (nds32_valid_multiple_load_store): Rename ...
282 (nds32_valid_multiple_load_store_p): ... to this and refine
284 * config/nds32/predicates.md
285 (nds32_load_multiple_and_update_address_operation): New predicate.
286 (nds32_store_multiple_and_update_address_operation): New predicate.
288 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
289 Chung-Ju Wu <jasonwucj@gmail.com>
291 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
292 (combo): New attribute.
293 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
295 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
297 * config/nds32/nds32.opt: Change -mcmodel= default value.
299 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
300 Monk Chiang <sh.chiang04@gmail.com>
301 Chung-Ju Wu <jasonwucj@gmail.com>
303 * config/nds32/constants.md (unspec_element): New enum.
304 * config/nds32/constraints.md (Umw): New constraint.
305 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
306 * config/nds32/nds32-intrinsic.md: Likewise.
307 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
308 (nds32_valid_smw_lwm_base_p): New.
309 (nds32_output_smw_single_word): New.
310 (nds32_output_lmw_single_word): New.
311 (nds32_expand_unaligned_load): New.
312 (nds32_expand_unaligned_store): New.
313 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
314 (nds32_output_smw_single_word): Declare.
315 (nds32_output_lmw_single_word): Declare.
316 (nds32_expand_unaligned_load): Declare.
317 (nds32_expand_unaligned_store): Declare.
318 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
319 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
320 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
321 NDS32_BUILTIN_UASTORE_DW.
322 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
325 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
326 Kito Cheng <kito.cheng@gmail.com>
327 Chung-Ju Wu <jasonwucj@gmail.com>
329 * config/nds32/nds32-intrinsic.c
330 (nds32_expand_builtin_null_ftype_reg): Delete.
331 (nds32_expand_builtin_reg_ftype_imm): Ditto.
332 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
333 (nds32_read_argument): New.
334 (nds32_legitimize_target): Ditto.
335 (nds32_legitimize_argument): Ditto.
336 (nds32_check_constant_argument): Ditto.
337 (nds32_expand_unop_builtin): Ditto.
338 (nds32_expand_unopimm_builtin): Ditto.
339 (nds32_expand_binop_builtin): Ditto.
340 (nds32_builtin_decl_impl): Ditto.
341 (builtin_description): Ditto.
342 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
343 (nds32_init_builtins_impl): Ditto.
344 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
345 (nds32_builtin_decl): New.
346 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
347 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
349 2018-03-02 Jeff Law <law@redhat.com>
351 * reorg.c (stop_search_p): Handle DEBUG_INSN.
352 (redundant_insn, fill_simple_delay_slots): Likewise.
353 (fill_slots_from_thread): Likewise.
354 * resource.c (mark_referenced_resources): Likewise.
355 (mark_set_resources, find_dead_or_set_registers): Likewise.
357 2018-03-02 Jakub Jelinek <jakub@redhat.com>
359 * substring-locations.h (format_warning_va): Formatting fix for
361 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
363 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
364 * substring-locations.c: Include intl.h.
365 (format_warning_va): Turned into small wrapper around
366 format_warning_n_va, renamed to ...
367 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
368 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
370 (format_warning_at_substring_n): New function.
371 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
372 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
373 format_warning_at_substring with just a shorter name instead of
374 const function pointer.
375 (fmtwarn_n): New function.
376 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
377 appropriate, get rid of all the fmtstr temporaries, move conditionals
378 with G_() wrapped string literals directly into fmtwarn arguments,
379 cast dir.len to (int), formatting fixes.
381 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
383 * doc/invoke.texi: Remove "Cilk Plus" references.
385 2018-03-02 Jakub Jelinek <jakub@redhat.com>
386 Richard Biener <rguenther@suse.de>
389 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
390 for error or warning attributes if CALL_FROM_THUNK_P is set.
393 2018-03-02 Jakub Jelinek <jakub@redhat.com>
396 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
397 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
400 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
401 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
403 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
404 instead of -1U in last predictors element's probability member.
406 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
409 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
410 arguments if they are comparable.
412 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
414 PR tree-optimization/84634
415 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
416 masks and masked_loop_p with a single loop_masks, making sure it's
417 null for bb vectorization.
419 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
421 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
422 (vect_analyze_data_ref_access): Use loop->safe_len rather than
423 loop->force_vectorize to check whether there is no alias.
425 2018-03-02 Jakub Jelinek <jakub@redhat.com>
428 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
430 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
432 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
433 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
434 instead of a loop around prev_real_insn.
435 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
439 * config/i386/i386.c (ix86_print_operand): Use conditional
440 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
443 2018-03-02 Richard Biener <rguenther@suse.de>
445 PR tree-optimization/84427
446 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
447 (bitmap_set_subtract_values): Rewrite to handle multiple
450 (prune_clobbered_mems): Likewise.
451 (phi_translate): Take edge instead of pred/phiblock.
452 (phi_translate_1): Likewise.
453 (phi_translate_set): Likewise. Insert all translated
454 exprs for a value into the set, keeping possibly multiple
455 expressions per value.
456 (compute_antic_aux): Adjust for phi_translate changes.
457 When intersecting union the expressions and prune those
458 not in the final value set, keeping possibly multiple
459 expressions per value. Do not use value-insertion
460 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
461 all expressions. Add verification that the value-sets
462 only shrink during iteration.
463 (compute_partial_antic_aux): Adjust for the phi_translate changes.
464 (do_pre_regular_insertion): Likewise.
465 (do_pre_partial_partial_insertion): Likewise.
467 2018-03-02 Richard Biener <rguenther@suse.de>
470 * config/darwin.c (saved_debug_info_level): New static global.
471 (darwin_asm_lto_start): Disable debug info generation for LTO out.
472 (darwin_asm_lto_end): Restore debug info generation settings.
474 2018-03-01 Martin Liska <mliska@suse.cz>
477 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
480 2018-03-01 Richard Biener <rguenther@suse.de>
483 * dwarf2out.c (gen_variable_die): Properly handle late VLA
484 type annotation with LTO when debug was disabled at compile-time.
486 2018-03-01 Matthew Fortune <mfortune@gmail.com>
488 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
490 (mips_final_postscan_insn): Likewise.
492 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
494 PR rtl-optimization/84528
495 * alias.c (init_alias_target): Add commentary.
496 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
497 a unique base value if the frame pointer is not eliminated
498 to the stack pointer.
500 2018-03-01 Tom de Vries <tom@codesourcery.com>
502 PR rtl-optimization/83327
503 * lra-int.h (hard_regs_spilled_into): Declare.
504 * lra.c (hard_regs_spilled_into): Define.
505 (init_reg_info): Init hard_regs_spilled_into.
506 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
507 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
508 (process_bb_lives): Handle hard_regs_spilled_into.
509 (lra_create_live_ranges_1): Before doing liveness propagation, clear
510 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
512 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
514 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
515 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
516 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
517 * config/rs6000/aix72.h: New file.
519 2018-02-28 Jakub Jelinek <jakub@redhat.com>
521 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
522 instead of warning_at with conditional singular and plural messages
526 * stor-layout.c (update_alignment_for_field): For
527 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
528 && !DECL_PACKED (field), do the alignment update, just use
529 only desired_align instead of MAX (type_align, desired_align)
531 (place_field): Don't do known_align < desired_align handling
532 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
533 is non-NULL, instead do it after rli->prev_field handling and
534 only if not within a bitfield word. For DECL_PACKED (field)
535 use type_align of BITS_PER_UNIT.
537 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
539 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
540 superfluous parentheses and trailing spaces.
542 2018-02-28 Richard Biener <rguenther@suse.de>
544 PR tree-optimization/84584
545 * graphite-scop-detection.c (scop_detection::add_scop): Discard
546 SCoPs with fake exit edge.
548 2018-02-28 Martin Liska <mliska@suse.cz>
551 * timevar.c (timer::print): Fix format to properly print 100%
554 2018-02-28 Richard Biener <rguenther@suse.de>
557 * genmatch.c (capture_info::walk_match): Do not mark
558 captured expressions without operands as expr_p given
559 they act more like predicates and should be subject to
560 "lost tail" side-effect preserving.
562 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
564 PR rtl-optimization/81611
565 * auto-inc-dec.c (attempt_change): Move dead note from
566 mem_insn if it's the next use of regno
567 (find_address): Take address use of reg holding
568 non-incremented value. Add parm to limit search to the named
570 (merge_in_block): Attempt to use a mem insn that is the next
571 use of the original regno.
573 2018-02-27 Martin Sebor <msebor@redhat.com>
576 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
577 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
579 2018-02-27 Martin Sebor <msebor@redhat.com>
582 * diagnostic-core.h (warning_n, error_n, inform_n): Change
583 n argument to unsigned HOST_WIDE_INT.
584 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
585 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
586 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
587 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
589 2018-02-27 Richard Biener <rguenther@suse.de>
591 PR tree-optimization/84512
592 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
593 Do not use the estimate returned from record_stmt_cost for
594 the scalar iteration cost but sum properly using add_stmt_cost.
596 2018-02-27 Richard Biener <rguenther@suse.de>
598 PR tree-optimization/84466
599 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
600 Adjust last change to less strictly validate use operands.
602 2018-02-27 Martin Liska <mliska@suse.cz>
604 PR gcov-profile/84548
605 * gcov.c (process_file): Allow partial overlap and consider it
606 also as group functions.
607 (output_lines): Properly calculate range of lines for a group.
609 2018-02-27 Martin Liska <mliska@suse.cz>
611 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
612 'ggc' suffixes. Change first column width.
613 (timer::print): Fix formatting of the column.
615 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
617 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
618 preserve inline entry blocks for the sake of debug inline
619 entry point markers alone.
620 (remove_unused_locals): Suggest in comments a better place to
621 force the preservation of inline entry blocks that are
622 otherwise unused, but do not preserve them.
624 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
626 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
628 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
631 * config/i386/constraints.md (Bs): Replace
632 ix86_indirect_branch_register with
633 TARGET_INDIRECT_BRANCH_REGISTER.
635 * config/i386/i386.md (indirect_jump): Likewise.
636 (tablejump): Likewise.
637 (*sibcall_memory): Likewise.
638 (*sibcall_value_memory): Likewise.
639 Peepholes of indirect call and jump via memory: Likewise.
640 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
641 (*sibcall_value_GOT_32): Likewise.
642 * config/i386/predicates.md (indirect_branch_operand): Likewise.
643 (GOT_memory_operand): Likewise.
644 (call_insn_operand): Likewise.
645 (sibcall_insn_operand): Likewise.
646 (GOT32_symbol_operand): Likewise.
647 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
649 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
651 PR rtl-optimization/83496
652 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
653 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
654 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
655 redundant insn, if any.
656 (relax_delay_slots): Likewise.
657 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
659 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
661 PR tree-optimization/83965
662 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
663 that grouped statements are part of a reduction chain. Return
664 true if the statement is not marked as a reduction itself but
666 (vect_recog_dot_prod_pattern): Don't check whether the statement
667 is part of a group here.
668 (vect_recog_sad_pattern): Likewise.
669 (vect_recog_widen_sum_pattern): Likewise.
671 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
674 * final.c (rest_of_clean_state): Also look for calls inside sequences.
676 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
679 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
681 (ix86_output_indirect_function_return): New prototype.
682 (ix86_split_simple_return_pop_internal): Likewise.
683 * config/i386/i386.c (indirect_return_via_cx): New.
684 (indirect_return_via_cx_bnd): Likewise.
685 (indirect_thunk_name): Handle return va CX_REG.
686 (output_indirect_thunk_function): Create alias for
687 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
688 (ix86_output_indirect_jmp): Remove the bool argument.
689 (ix86_output_indirect_function_return): New function.
690 (ix86_split_simple_return_pop_internal): Likewise.
691 * config/i386/i386.md (*indirect_jump): Don't pass false
692 to ix86_output_indirect_jmp.
693 (*tablejump_1): Likewise.
694 (simple_return_pop_internal): Change it to define_insn_and_split.
695 Call ix86_split_simple_return_pop_internal to split it for
697 (simple_return_indirect_internal): Call
698 ix86_output_indirect_function_return instead of
699 ix86_output_indirect_jmp.
701 2018-02-26 Jakub Jelinek <jakub@redhat.com>
704 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
705 memset and value initialization afterwards.
707 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
709 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
711 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
714 * common/config/aarch64/aarch64-common.c
715 (aarch_option_optimization_table[]): Switch
716 off fomit-frame-pointer
718 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
719 Chung-Ju Wu <jasonwucj@gmail.com>
721 * config/nds32/nds32-multiple.md (load_multiple): Disallow
723 (store_multiple): Ditto.
725 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
727 * config.gcc: Add --with-cpu support for nds32 target.
728 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
729 * config/nds32/nds32.opt: Add -mcpu= option.
731 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
733 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
734 isel=yes): Warn for these deprecated options.
736 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
738 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
739 ISA_2_5_MASKS_EMBEDDED.
741 2018-02-23 Jakub Jelinek <jakub@redhat.com>
743 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
744 p->max as pointers rather than using iterative_hash_expr.
746 2018-02-23 Carl Love <cel@us.ibm.com>
748 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
749 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
751 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
752 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
753 P8V_BUILTIN_VEC_VUNSIGNED2.
755 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
758 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
759 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
761 * lra-constraints.c (curr_insn_transform): Set up
762 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
764 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
765 * lra-eliminations.c (spill_pseudos): Ditto.
766 (process_insn_for_elimination): Ditto.
767 * lra-lives.c (reg_early_clobber_p): Use the new macros.
768 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
769 LRA_NON_CLOBBERED_ALT.
771 2018-02-22 Martin Sebor <msebor@redhat.com>
773 PR tree-optimization/84480
774 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
775 to maybe_diag_stxncpy_trunc. Call it.
776 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
777 from gimple_fold_builtin_strcpy. Print inlining stack.
778 (handle_builtin_stxncpy): Print inlining stack.
779 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
781 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
784 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
785 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
786 and -fcheck-pointer-bounds are used together.
787 (indirect_thunk_prefix): New enum.
788 (indirect_thunk_need_prefix): New function.
789 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
790 "_nt" instead of "_bnd" for NOTRACK prefix.
791 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
792 (output_indirect_thunk_function): Likewise.
794 (ix86_code_end): Update output_indirect_thunk_function calls.
795 (ix86_output_indirect_branch_via_reg): Replace
796 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
797 (ix86_output_indirect_branch_via_push): Likewise.
798 (ix86_output_function_return): Likewise.
799 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
800 incompatible with -fcf-protection=branch and
801 -fcheck-pointer-bounds.
803 2018-02-22 Steve Ellcey <sellcey@cavium.com>
806 * config/aarch64/aarch64.c (aarch64_print_address_internal):
807 Change gcc_assert call to output_operand_lossage.
809 2018-02-22 Steve Ellcey <sellcey@cavium.com>
811 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
813 2018-02-22 DJ Delorie <dj@redhat.com>
814 Sebastian Perta <sebastian.perta@renesas.com>
815 Oleg Endo <olegendo@gcc.gnu.org>
817 * config/rx/rx.c (rx_rtx_costs): New function.
818 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
820 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
822 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
824 2018-02-22 Martin Liska <mliska@suse.cz>
827 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
828 Add "native" as a possible value.
830 2018-02-22 Martin Liska <mliska@suse.cz>
833 * config/i386/i386.c (ix86_option_override_internal):
834 Add "native" as a possible value for -march and -mtune.
836 2018-02-22 Jakub Jelinek <jakub@redhat.com>
839 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
840 to all type variants.
842 PR tree-optimization/84503
843 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
844 width as info->bitpos + info->bitsize - start.
845 (merged_store_group::merge_overlapping): Simplify width computation.
846 (check_no_overlap): New function.
847 (imm_store_chain_info::try_coalesce_bswap): Compute expected
848 start + width and last_order of the group, fail if check_no_overlap
850 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
851 to group if check_no_overlap fails.
853 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
855 * config/rs6000/altivec.md: Delete contraint arguments to
856 define_expand, define_split, and define_peephole2, and in
857 define_insn_and_split if always unused.
858 * config/rs6000/darwin.md: Ditto.
859 * config/rs6000/dfp.md: Ditto.
860 * config/rs6000/rs6000.md: Ditto.
861 * config/rs6000/sync.md: Ditto.
862 * config/rs6000/vector.md: Ditto.
863 * config/rs6000/vsx.md: Ditto.
865 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
867 * config/rs6000/altivec.md: Write output control strings as braced
868 blocks instead of double-quoted strings.
869 * config/rs6000/darwin.md: Ditto.
870 * config/rs6000/rs6000.md: Ditto.
871 * config/rs6000/vector.md: Ditto.
872 * config/rs6000/vsx.md: Ditto.
874 2018-02-21 Jason Merrill <jason@redhat.com>
876 PR c++/84314 - ICE with templates and fastcall attribute.
877 * attribs.c (build_type_attribute_qual_variant): Remove assert.
879 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
881 * ipa-cp.c (determine_versionability): Fix comment typos.
883 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
886 * ipa-cp.c (determine_versionability): Do not version functions caling
889 2018-02-21 Martin Liska <mliska@suse.cz>
892 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
893 Add "native" as a possible value.
894 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
895 the macro when native cpu detection is available.
897 2018-02-21 Martin Liska <mliska@suse.cz>
900 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
901 Add "native" as a possible value.
902 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
903 when native cpu detection is available.
905 2018-02-21 Jakub Jelinek <jakub@redhat.com>
906 Martin Sebor <msebor@redhat.com>
908 PR tree-optimization/84478
909 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
911 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
912 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
913 support which is conservatively correct, for 2 only stay conservative
914 for maxlen. Formatting and comment capitalization fixes. Add STRICT
915 argument to the 2 argument get_range_strlen, adjust 6 arg
916 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
918 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
919 (gimple_fold_builtin_strlen): Pass true as last argument to
922 2018-02-20 Martin Sebor <msebor@redhat.com>
925 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
926 (builtin_memref::set_base_and_offset): Same. Handle inner references.
927 (builtin_memref::builtin_memref): Factor out parts into
928 set_base_and_offset and call it.
930 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
933 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
934 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
935 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
936 search at the associated MODE_INT.
938 2018-02-20 Jeff Law <law@redhat.com>
941 PR tree-optimization/81592
943 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
944 for range data rather than using global data.
945 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
946 range data rather than using global data.
947 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
948 pass it to children as needed.
949 (struct directive::fmtresult): Similarly.
950 (struct directive::set_width): Similarly.
951 (struct directive::set_precision): Similarly.
952 (format_integer, format_directive, parse_directive): Similarly.
953 (format_none): Accept unnamed vr_values parameter.
954 (format_percent, format_floating, format_character): Similarly.
955 (format_string, format_plain): Similarly.
956 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
957 the EVRP range analyzer for range data rather than using global data.
958 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
959 gimple-ssa-evrp-analyze.h
960 (class sprintf_dom_walker): Add after_dom_children member function.
961 Add evrp_range_analyzer member.
962 (sprintf_dom_walker::before_dom_children): Call into the EVRP
963 range analyzer as needed.
964 (sprintf_dom_walker::after_dom_children): New member function.
965 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
967 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
968 (evrp_range_analyzer::pop_to_marker): Likewise.
970 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
972 PR tree-optimization/84419
973 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
974 with the required type if its current type is compatible but
977 2018-02-20 Jakub Jelinek <jakub@redhat.com>
980 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
983 2018-02-20 Martin Liska <mliska@suse.cz>
986 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
987 possible values if we don't have a hint.
989 2018-02-20 Martin Liska <mliska@suse.cz>
993 * final.c (shorten_branches): Build align_tab array with one
995 * opts.c (finish_options): Add alignment option limit check.
996 (MAX_CODE_ALIGN): Likewise.
997 (MAX_CODE_ALIGN_VALUE): Likewise.
998 * doc/invoke.texi: Document maximum allowed option value for
999 all -falign-* options.
1001 2018-02-19 Jakub Jelinek <jakub@redhat.com>
1004 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
1005 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
1006 * var-tracking.c (emit_note_insn_var_location): Remove all references
1007 to NOTE_INSN_CALL_ARG_LOCATION.
1008 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
1009 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
1010 Use copy_rtx_if_shared.
1011 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
1012 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
1013 (dwarf2out_var_location): Remove handling of
1014 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
1016 * final.c (final_scan_insn): Remove all references to
1017 NOTE_INSN_CALL_ARG_LOCATION.
1018 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
1019 before dumping final insns.
1020 * except.c (emit_note_eh_region_end): Remove all references to
1021 NOTE_INSN_CALL_ARG_LOCATION.
1022 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
1023 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
1024 * config/arc/arc.c (hwloop_optimize): Likewise.
1025 * config/arm/arm.c (create_fix_barrier): Likewise.
1026 * config/s390/s390.c (s390_chunkify_start): Likewise.
1027 * config/sh/sh.c (find_barrier): Likewise.
1028 * config/i386/i386.c (rest_of_insert_endbranch,
1029 ix86_seh_fixup_eh_fallthru): Likewise.
1030 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
1031 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
1032 * config/frv/frv.c (frv_function_prologue): Likewise.
1033 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
1035 (note_outside_basic_block_p): Remove all references to
1036 NOTE_INSN_CALL_ARG_LOCATION.
1037 * gengtype.c (adjust_field_rtx_def): Likewise.
1038 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
1040 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
1041 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
1044 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
1047 PR tree-optimization/84452
1048 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
1049 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
1052 2018-02-19 Martin Liska <mliska@suse.cz>
1055 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
1057 2018-02-19 Martin Liska <mliska@suse.cz>
1058 Richard Sandiford <richard.sandiford@linaro.org>
1060 PR tree-optimization/82491
1061 * gimple-fold.c (get_base_constructor): Make earlier bail out
1064 2018-02-19 Carl Love <cel@us.ibm.com>
1066 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
1067 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
1069 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
1070 P8V_BUILTIN_VEC_NEG.
1072 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
1074 * config/rl78/rl78.md (movdf): New define expand.
1076 2018-02-19 Martin Liska <mliska@suse.cz>
1079 * doc/invoke.texi: Fix typo.
1080 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
1082 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
1084 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
1085 handle rs6000_single_float and rs6000_double_float specially for
1088 2018-02-16 Jeff Law <law@redhat.com>
1090 * config/rx/rx.c (add_pop_cfi_notes): New function.;
1093 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1096 * ipa-inline.c (inline_small_functions): Fix a typo.
1098 2018-02-16 Nathan Sidwell <nathan@acm.org>
1100 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
1102 2018-02-16 Carl Love <cel@us.ibm.com>
1104 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
1105 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
1106 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
1107 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
1108 expansion to P8V_BUILTIN_VEC_FLOAT2.
1110 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
1112 PR rtl-optimization/70023
1113 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
1114 src_regno into account.
1116 2018-02-16 Carl Love <cel@us.ibm.com>
1118 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
1119 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
1120 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
1121 * config/rs6000/rs6000.c: Remove case statements for
1122 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
1123 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
1124 and P9V_BUILTIN_VEC_VINSERT4B.
1125 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
1126 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
1127 * config/rs6000/vsx.md:
1128 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
1131 2018-02-16 Carl Love <cel@us.ibm.com>
1133 * config/rs6000/altivec.h: Add builtin names vec_extract4b
1135 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
1137 * config/rs6000/rs6000-c.c: Add the definitions for
1138 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
1139 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
1140 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
1141 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
1142 definition for insert4b and define insn *insert3b_internal.
1143 * doc/extend.texi: Add documentation for vec_extract4b.
1145 2018-02-16 Nathan Sidwell <nathan@acm.org>
1147 * doc/extend.texi (Backwards Compatibility): Mention friend
1148 injection. Note for-scope is deprecated.
1149 * doc/invoke.texi (-ffriend-injection): Deprecate.
1151 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
1153 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
1154 that moved to I2, also allow destinations that are a paradoxical
1155 subreg (instead of a normal reg).
1157 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
1160 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
1163 2018-02-16 Richard Biener <rguenther@suse.de>
1165 PR tree-optimization/84037
1166 PR tree-optimization/84016
1168 * config/i386/i386.c (ix86_builtin_vectorization_cost):
1169 Adjust vec_construct for the fact we need additional higher latency
1170 128bit inserts for AVX256 and AVX512 vector builds.
1171 (ix86_add_stmt_cost): Scale vector construction cost for
1174 2018-02-16 Richard Biener <rguenther@suse.de>
1176 PR tree-optimization/84417
1177 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
1178 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
1179 (non_rewritable_lvalue_p): Likewise, use poly-ints.
1181 2018-02-16 Martin Liska <mliska@suse.cz>
1184 * internal-fn.def (ASAN_CHECK): Set proper flags.
1185 (ASAN_MARK): Likewise.
1187 2018-02-16 Julia Koval <julia.koval@intel.com>
1189 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
1190 from PTA_CANNONLAKE.
1192 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1195 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
1196 Use ++iter rather than iter++ for std::list iterators.
1197 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
1198 defer deleting them until all nodes in the forest are processed. Do
1199 free even leaf nodes. Change to_process into auto_vec.
1202 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
1203 * vec.h (vec_default_construct): Use memset instead of placement new
1204 if BROKEN_VALUE_INITIALIZATION is defined.
1205 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
1206 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
1209 PR rtl-optimization/83723
1210 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
1211 * lra.c (lra_substitute_pseudo): Likewise. If true, use
1212 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
1214 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
1216 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
1218 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
1220 PR rtl-optimization/81443
1221 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
1222 from inner REGs to paradoxical SUBREGs.
1224 2018-02-16 Richard Biener <rguenther@suse.de>
1226 PR tree-optimization/84399
1227 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1228 For operands we can analyze at their definition make sure we can
1229 analyze them at each use as well.
1231 2018-02-16 Richard Biener <rguenther@suse.de>
1233 PR tree-optimization/84190
1234 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
1235 volatile accesses if the decl isn't volatile.
1237 2018-02-15 Jason Merrill <jason@redhat.com>
1239 PR c++/84314 - ICE with templates and fastcall attribute.
1240 * attribs.c (build_type_attribute_qual_variant): Don't clobber
1241 TYPE_CANONICAL on an existing type.
1243 2018-02-15 Jakub Jelinek <jakub@redhat.com>
1245 PR tree-optimization/84383
1246 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
1247 dstoff nor call operand_equal_p if dstbase is NULL.
1249 PR tree-optimization/84334
1250 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
1251 also a CONSTANT_CLASS_P, punt.
1253 2018-02-14 Jim Wilson <jimw@sifive.com>
1255 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
1256 first SMALL_OPERAND check. New local min_second_step. Move assert
1257 to where locals are set. Add TARGET_RVC support.
1258 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
1260 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
1262 * doc/invoke.texi: Correct -Wformat-overflow code sample.
1264 2018-02-14 Martin Sebor <msebor@redhat.com>
1266 PR tree-optimization/83698
1267 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
1268 arrays constrain the offset range to their bounds.
1269 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
1270 (builtin_access::overlap): Avoid setting the size of overlap if it's
1272 (maybe_diag_overlap): Also consider arrays when deciding what values
1273 of offsets to include in diagnostics.
1275 2018-02-14 Martin Sebor <msebor@redhat.com>
1278 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
1279 that correspond to the kind of a declaration.
1281 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
1284 * config/pa/pa.md: Load address of PIC label using the linkage table
1285 if the label is nonlocal.
1287 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
1289 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
1290 warning message if user requests -maltivec=be.
1291 * doc/invoke.texi: Document deprecation of -maltivec=be.
1293 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
1296 * config/rs6000/rs6000-c.c: Update definitions for
1297 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
1298 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
1300 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1303 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
1304 add _get_ssp intrinsics. Remove argument from
1305 __builtin_ia32_rdssp[d|q].
1306 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
1307 * config/i386/i386-builtin.def: Remove argument from
1308 __builtin_ia32_rdssp[d|q].
1309 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
1310 ix86_expand_special_args_builtin for _rdssp[d|q].
1311 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
1312 Clear register before usage.
1313 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
1314 Add documentation for new _get_ssp and _inc_ssp intrinsics.
1316 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
1318 PR tree-optimization/84357
1319 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
1320 operand 1 of an ARRAY_REF too.
1322 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
1325 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
1326 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1328 (set_of_reg): New struct.
1329 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
1330 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
1331 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1333 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
1334 Split into bitclr, bitset, bitinvert patterns if appropriate.
1335 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
1336 use rx_fuse_in_memory_bitop.
1337 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
1338 to named insn, correct maximum insn length.
1340 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
1343 * machmode.def: Define a complex mode for PARTIAL_INT.
1344 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
1346 * doc/rtl.texi: Document CSPImode.
1347 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
1349 (msp430_hard_regno_nregs_with_padding): Likewise.
1351 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
1354 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
1356 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
1358 PR rtl-optimization/84169
1359 * combine.c (try_combine): New variable split_i2i3. Set it to true if
1360 we generated a parallel as new i3 and we split that to new i2 and i3
1361 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
1362 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
1363 those to i2, not i1. Partially rewrite this scan code.
1365 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1368 * stor-layout.c (place_field): For variable length fields, adjust
1369 offset_align afterwards not just based on the field's alignment,
1370 but also on the size.
1373 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
1374 of exps and logs in the use_exp2 case.
1376 2018-02-13 Jeff Law <law@redhat.com>
1378 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
1381 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
1384 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
1388 * common.opt (gas-loc-support, gas-locview-support): New.
1389 (ginline-points, ginternal-reset-location-views): New.
1390 * doc/invoke.texi: Document them. Use @itemx where intended.
1391 (gvariable-location-views): Adjust.
1392 * target.def (reset_location_view): New.
1393 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
1394 (TARGET_RESET_LOCATION_VIEW): New.
1395 * doc/tm.texi: Rebuilt.
1396 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
1397 (dwarf2out_default_as_locview_support): New.
1398 (output_asm_line_debug_info): Use option variables.
1399 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
1400 (output_loc_list): Likewise.
1401 (add_high_low_attributes): Check option variables.
1402 Don't output entry view attribute in strict mode.
1403 (gen_inlined_subroutine_die): Check option variables.
1404 (dwarf2out_inline_entry): Likewise.
1405 (init_sections_and_labels): Likewise.
1406 (dwarf2out_early_finish): Likewise.
1407 (maybe_reset_location_view): New, from...
1408 (dwarf2out_var_location): ... here. Call it.
1409 * debug.h (dwarf2out_default_as_loc_support): Declare.
1410 (dwarf2out_default_as_locview_support): Declare.
1411 * hooks.c (hook_int_rtx_insn_0): New.
1412 * hooks.h (hook_int_rtx_insn_0): Declare.
1413 * toplev.c (process_options): Take -gas-loc-support and
1414 -gas-locview-support from dwarf2out. Enable
1415 -gvariable-location-views by default only with locview
1416 assembler support. Enable -ginternal-reset-location-views by
1417 default only if the target defines the corresponding hook.
1418 Enable -ginline-points by default if location views are
1419 enabled; force it disabled if statement frontiers are
1421 * tree-inline.c (expand_call_inline): Check option variables.
1422 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
1424 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
1426 PR tree-optimization/84321
1427 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
1428 handling. Also check whether the anti-range contains any values
1429 that satisfy the mask; switch to a VR_RANGE if not.
1431 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
1434 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
1436 2018-02-13 Martin Jambor <mjambor@suse.cz>
1439 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
1440 of call statements, also set location of a load to a temporary.
1442 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
1444 * config/rl78/rl78.c (add_vector_labels): New function.
1445 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
1446 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
1447 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
1448 which checks that no arguments are passed.
1449 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
1450 * doc/extend.texi: Documentation for the new attribute.
1452 2018-02-13 Andreas Schwab <schwab@suse.de>
1454 * config/riscv/linux.h (CPP_SPEC): Define.
1456 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1459 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
1460 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
1461 OPTION_MASK_ISA_AES as first argument to def_builtin_const
1462 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
1463 instead of OPTION_MASK_ISA_PCLMUL as first argument to
1464 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
1465 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
1466 temporarily for AES and PCLMUL builtins.
1468 PR tree-optimization/84339
1469 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
1470 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
1474 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
1475 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
1476 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
1478 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
1480 * omp-simd-clone.h: New file.
1481 * omp-simd-clone.c: Include omp-simd-clone.h.
1482 (expand_simd_clones): No longer static.
1483 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
1484 cgraph.h and omp-simd-clone.h.
1485 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
1486 (vect_recog_widen_shift_pattern): Formatting fix.
1487 (vect_pattern_recog_1): Don't check optab for calls.
1490 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
1491 operands[2] into a REG before using gen_lowpart on it.
1493 2018-02-12 Jeff Law <law@redhat.com>
1496 * config/sh/sh.c (find_barrier): Consider a sibling call
1499 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
1500 successfully back substituting a reg.
1502 2018-02-12 Richard Biener <rguenther@suse.de>
1504 PR tree-optimization/84037
1505 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
1506 parameter, move visited init to caller.
1507 (vect_slp_analyze_operations): Separate cost from validity
1508 check, initialize visited once for all instances.
1509 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
1511 * tree-vect-stmts.c (vect_model_simple_cost): Make early
1513 (vect_model_promotion_demotion_cost): Likewise.
1514 (vectorizable_bswap): Guard cost modeling with !slp_node
1515 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
1517 (vectorizable_call): Likewise.
1518 (vectorizable_conversion): Likewise.
1519 (vectorizable_assignment): Likewise.
1520 (vectorizable_shift): Likewise.
1521 (vectorizable_operation): Likewise.
1522 (vectorizable_store): Likewise.
1523 (vectorizable_load): Likewise.
1524 (vectorizable_condition): Likewise.
1525 (vectorizable_comparison): Likewise.
1527 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
1530 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
1531 (ASAN_MARK): Fix fnspec to account for return value, change pointer
1532 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
1534 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1537 * params.def (inline-min-speedup): Increase from 8 to 15.
1538 (max-inline-insns-auto): Decrease from 40 to 30.
1539 * ipa-split.c (consider_split): Add some buffer for function to
1540 be considered inlining candidate.
1541 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
1544 2018-02-12 Richard Biener <rguenther@suse.de>
1546 PR tree-optimization/84037
1547 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
1548 matched stmts if we cannot swap the non-matched ones.
1550 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1552 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
1553 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
1554 _mm_maskz_scalef_round_ss): New intrinsics.
1555 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
1556 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
1557 __builtin_ia32_scalefss_round): Remove.
1558 (__builtin_ia32_scalefsd_mask_round,
1559 __builtin_ia32_scalefss_mask_round): New intrinsics.
1560 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
1561 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
1562 ((match_operand:VF_128 2 "<round_nimm_predicate>"
1563 "<round_constraint>")): Changed to ...
1564 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
1565 "<round_scalar_constraint>")): ... this.
1566 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
1567 %0, %1, %2<round_op3>}"): Changed to ...
1568 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
1569 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
1570 %2<round_scalar_mask_op3>}"): ... this.
1571 * config/i386/subst.md (round_scalar_nimm_predicate): New.
1573 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1575 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
1576 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
1577 (_mm_maskz_sqrt_round_ss): New intrinsics.
1578 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
1579 (__builtin_ia32_sqrtsd_mask_round)
1580 (__builtin_ia32_sqrtss_mask_round): New builtins.
1581 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
1582 (__builtin_ia32_sqrtss_round): Remove.
1583 (__builtin_ia32_sqrtsd_mask_round)
1584 (__builtin_ia32_sqrtss_mask_round): New builtins.
1585 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
1586 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
1587 ((match_operand:VF_128 1 "vector_operand"
1588 "xBm,<round_constraint>")): Changed to ...
1589 ((match_operand:VF_128 1 "vector_operand"
1590 "xBm,<round_scalar_constraint>")): ... this.
1591 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
1592 %0, %2, %<iptr>1<round_op3>}): Changed to ...
1593 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
1594 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
1595 %<iptr>1<round_scalar_mask_op3>}): ... this.
1596 ((set_attr "prefix" "<round_prefix>")): Changed to ...
1597 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
1599 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
1602 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
1603 Cast vec_cmpeq result to correct type.
1604 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
1605 Cast vec_cmpgt result to correct type.
1607 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
1609 * final.c (final_scan_insn_1): Renamed from...
1610 (final_scan_insn): ... this. New wrapper, to recover
1611 seen from the outermost call in recursive ones.
1612 * config/sparc/sparc.c (output_return): Drop seen from call.
1613 (output_sibcall): Likewise.
1614 * config/visium/visium.c (output_branch): Likewise.
1616 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
1618 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
1621 2018-02-10 Alan Modra <amodra@gmail.com>
1624 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
1625 Specify LR as an input.
1627 2018-02-10 Jakub Jelinek <jakub@redhat.com>
1630 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
1631 remove_member_access_dummy_vars): New functions.
1632 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
1633 lower_omp_1, execute_lower_omp): Use them.
1635 PR rtl-optimization/84308
1636 * shrink-wrap.c (spread_components): Release todo vector.
1638 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
1640 PR rtl-optimization/57193
1641 * ira-color.c (struct allocno_color_data): Add member
1642 conflict_allocno_hard_prefs.
1643 (update_conflict_allocno_hard_prefs): New.
1644 (bucket_allocno_compare_func): Add a preference based on
1645 conflict_allocno_hard_prefs.
1646 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
1647 (color_allocnos): Remove a dead code. Initiate
1648 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
1650 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1653 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
1654 constraint from =wa to wa. Avoid a subreg on the output operand,
1655 instead use a pseudo and subreg it in a move.
1656 (p9_xxbrd_<mode>): Changed to ...
1657 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
1658 (p9_xxbrd_v2df): New expander.
1659 (p9_xxbrw_<mode>): Changed to ...
1660 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
1661 (p9_xxbrw_v4sf): New expander.
1663 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1665 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
1667 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
1670 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
1671 multiply in 32-bit mode.
1672 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
1673 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
1676 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1678 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
1679 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
1680 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
1681 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
1683 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
1686 * dwarf2out.c (is_trivial_indirect_ref): New function.
1687 (dwarf2out_late_global_decl): Do not generate a location
1688 attribute for variables that have a non-trivial DECL_VALUE_EXPR
1689 and that are not defined in the current unit.
1691 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1693 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
1694 instead of a libcall for UNORDERED.
1696 2018-02-09 Tamar Christina <tamar.christina@arm.com>
1699 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
1700 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
1702 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1705 * config/s390/s390.c (s390_set_current_function): Invoke
1706 s390_indirect_branch_settings also if fndecl didn't change.
1708 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1710 * config/rs6000/rs6000.md (blockage): Set length to zero.
1712 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1714 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
1716 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1719 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
1720 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
1724 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
1725 PARALLEL incoming that failed vt_get_decl_and_offset check.
1728 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
1729 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
1731 (get_variable_section): For decls in named .bss* sections pass true as
1732 second argument to bss_initializer_p.
1734 2018-02-09 Marek Polacek <polacek@redhat.com>
1735 Jakub Jelinek <jakub@redhat.com>
1738 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
1739 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
1740 Sync some changes from cxx_fold_indirect_ref.
1742 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1744 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
1746 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
1747 (BLOCK_INLINE_ENTRY_LABEL): New.
1748 (dwarf2out_var_location): Disregard inline entry markers.
1749 (inline_entry_data): New struct.
1750 (inline_entry_data_hasher): New hashtable type.
1751 (inline_entry_data_hasher::hash): New.
1752 (inline_entry_data_hasher::equal): New.
1753 (inline_entry_data_table): New variable.
1754 (add_high_low_attributes): Add DW_AT_entry_pc and
1755 DW_AT_GNU_entry_view attributes if a pending entry is found
1756 in inline_entry_data_table. Add old entry_pc attribute only
1757 if debug nonbinding markers are disabled.
1758 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
1759 markers are enabled.
1760 (block_within_block_p, dwarf2out_inline_entry): New.
1761 (dwarf2out_finish): Check that no entries remained in
1762 inline_entry_data_table.
1763 * final.c (reemit_insn_block_notes): Handle inline entry notes.
1764 (final_scan_insn, notice_source_line): Likewise.
1765 (rest_of_clean_state): Skip inline entry markers.
1766 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
1768 * gimple.c (gimple_build_debug_inline_entry): New.
1769 * gimple.h (enum gimple_debug_subcode): Add
1770 GIMPLE_DEBUG_INLINE_ENTRY.
1771 (gimple_build_debug_inline_entry): Declare.
1772 (gimple_debug_inline_entry_p): New.
1773 (gimple_debug_nonbind_marker_p): Adjust.
1774 * insn-notes.def (INLINE_ENTRY): New.
1775 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
1776 inline entry marker notes.
1777 (print_insn): Likewise.
1778 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
1779 (INSN_DEBUG_MARKER_KIND): Likewise.
1780 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
1781 * tree-inline.c (expand_call_inline): Build and insert
1782 debug_inline_entry stmt.
1783 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
1784 inline entry blocks early, if nonbind markers are enabled.
1785 (dump_scope_block): Dump fragment info.
1786 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
1787 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
1788 (gimple_build_debug_inline_entry): New.
1789 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
1790 Enable/disable inline entry points too.
1791 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
1792 (DEBUG_INSN): Describe inline entry markers.
1794 * common.opt (gvariable-location-views): New.
1795 (gvariable-location-views=incompat5): New.
1796 * config.in: Rebuilt.
1797 * configure: Rebuilt.
1798 * configure.ac: Test assembler for view support.
1799 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
1800 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
1801 * dwarf2out.c (var_loc_view): New typedef.
1802 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
1803 (dwarf2out_locviews_in_attribute): New.
1804 (dwarf2out_locviews_in_loclist): New.
1805 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
1806 (enum dw_line_info_opcode): Add LI_adv_address.
1807 (struct dw_line_info_table): Add view.
1808 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
1809 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
1810 (zero_view_p): New variable.
1811 (ZERO_VIEW_P): New macro.
1812 (output_asm_line_debug_info): New.
1813 (struct var_loc_node): Add view.
1814 (add_AT_view_list, AT_loc_list): New.
1815 (add_var_loc_to_decl): Add view param. Test it against last.
1816 (new_loc_list): Add view params. Record them.
1817 (AT_loc_list_ptr): Handle loc and view lists.
1818 (view_list_to_loc_list_val_node): New.
1819 (print_dw_val): Handle dw_val_class_view_list.
1820 (size_of_die): Likewise.
1821 (value_format): Likewise.
1822 (loc_list_has_views): New.
1823 (gen_llsym): Set vl_symbol too.
1824 (maybe_gen_llsym, skip_loc_list_entry): New.
1825 (dwarf2out_maybe_output_loclist_view_pair): New.
1826 (output_loc_list): Output view list or entries too.
1827 (output_view_list_offset): New.
1828 (output_die): Handle dw_val_class_view_list.
1829 (output_dwarf_version): New.
1830 (output_compilation_unit_header): Use it.
1831 (output_skeleton_debug_sections): Likewise.
1832 (output_rnglists, output_line_info): Likewise.
1833 (output_pubnames, output_aranges): Update version comments.
1834 (output_one_line_info_table): Output view numbers in asm comments.
1835 (dw_loc_list): Determine current endview, pass it to new_loc_list.
1836 Call maybe_gen_llsym.
1837 (loc_list_from_tree_1): Adjust.
1838 (add_AT_location_description): Create view list attribute if
1839 needed, check it's absent otherwise.
1840 (convert_cfa_to_fb_loc_list): Adjust.
1841 (maybe_emit_file): Call output_asm_line_debug_info for test.
1842 (dwarf2out_var_location): Reset views as needed. Precompute
1843 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
1844 attribute. Set view.
1845 (new_line_info_table): Reset next view.
1846 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
1847 (dwarf2out_source_line): Likewise. Output view resets and labels to
1848 the assembler, or select appropriate line info opcodes.
1849 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
1850 (optimize_string_length): Catch it. Adjust.
1851 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
1852 dw_val_class_view_list, and remove it if no longer needed.
1853 (hash_loc_list): Hash view numbers.
1854 (loc_list_hasher::equal): Compare them.
1855 (optimize_location_lists): Check whether a view list symbol is
1856 needed, and whether the locview attribute is present, and
1857 whether they match. Remove the locview attribute if no longer
1859 (index_location_lists): Call skip_loc_list_entry for test.
1860 (dwarf2out_finish): Call output_asm_line_debug_info for test.
1861 Use output_dwarf_version.
1862 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
1863 (struct dw_val_node): Add val_view_list.
1864 * final.c (SEEN_NEXT_VIEW): New.
1865 (set_next_view_needed): New.
1866 (clear_next_view_needed): New.
1867 (maybe_output_next_view): New.
1868 (final_start_function): Rename to...
1869 (final_start_function_1): ... this. Take pointer to FIRST,
1870 add SEEN parameter. Emit param bindings in the initial view.
1871 (final_start_function): Reintroduce SEEN-less interface.
1872 (final): Rename to...
1873 (final_1): ... this. Take SEEN parameter. Output final pending
1874 next view at the end.
1875 (final): Reintroduce seen-less interface.
1876 (final_scan_insn): Output pending next view before switching
1877 sections or ending a block. Mark the next view as needed when
1878 outputting variable locations. Notify debug backend of section
1879 changes, and of location view changes.
1880 (rest_of_handle_final): Adjust.
1881 * toplev.c (process_options): Autodetect value for debug variable
1882 location views option. Warn on incompat5 without -gdwarf-5.
1883 * doc/invoke.texi (gvariable-location-views): New.
1884 (gvariable-location-views=incompat5): New.
1885 (gno-variable-location-views): New.
1887 2018-02-08 David Malcolm <dmalcolm@redhat.com>
1889 PR tree-optimization/84136
1890 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
1891 that the result of find_edge is non-NULL.
1893 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1896 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
1897 storing integer register in SImode. Fix cost of 256 and 512
1898 byte aligned SSE register store.
1900 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1902 * config/i386/i386.c (ix86_multiplication_cost): Fix
1903 multiplication cost for TARGET_AVX512DQ.
1905 2018-02-08 Marek Polacek <polacek@redhat.com>
1907 PR tree-optimization/84238
1908 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
1911 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1913 PR tree-optimization/84265
1914 * tree-vect-stmts.c (vectorizable_store): Don't treat
1915 VMAT_CONTIGUOUS accesses as grouped.
1916 (vectorizable_load): Likewise.
1918 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1920 PR tree-optimization/81635
1921 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
1922 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
1923 (test_round_for_mask): New functions.
1924 (wide_int_cc_tests): Call test_round_for_mask.
1925 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
1926 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
1927 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
1928 range returned by get_range_info.
1930 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1933 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
1934 * symtab.c: Include builtins.h
1935 (symtab_node::output_to_lto_symbol_table_p): Move here
1936 from lto-streamer-out.c:output_symbol_p.
1937 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
1938 (output_symbol_p): Move all logic to symtab.c
1939 (produce_symtab): Update.
1941 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1943 * config/s390/s390-opts.h (enum indirect_branch): Define.
1944 * config/s390/s390-protos.h (s390_return_addr_from_memory)
1945 (s390_indirect_branch_via_thunk)
1946 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
1947 (enum s390_indirect_branch_type): Define.
1948 * config/s390/s390.c (struct s390_frame_layout, struct
1949 machine_function): Remove.
1950 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
1951 (indirect_branch_table_label_no, indirect_branch_table_name):
1953 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
1954 (enum s390_indirect_branch_option): Define.
1955 (s390_return_addr_from_memory): New function.
1956 (s390_handle_string_attribute): New function.
1957 (s390_attribute_table): Add new attribute handler.
1958 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
1959 (s390_indirect_branch_via_thunk): New function.
1960 (s390_indirect_branch_via_inline_thunk): New function.
1961 (s390_function_ok_for_sibcall): When jumping via thunk disallow
1962 sibling call optimization for non z10 compiles.
1963 (s390_emit_call): Force indirect branch target to be a single
1964 register. Add r1 clobber for non-z10 compiles.
1965 (s390_emit_epilogue): Emit return jump via return_use expander.
1966 (s390_reorg): Handle JUMP_INSNs as execute targets.
1967 (s390_option_override_internal): Perform validity checks for the
1968 new command line options.
1969 (s390_indirect_branch_attrvalue): New function.
1970 (s390_indirect_branch_settings): New function.
1971 (s390_set_current_function): Invoke s390_indirect_branch_settings.
1972 (s390_output_indirect_thunk_function): New function.
1973 (s390_code_end): Implement target hook.
1974 (s390_case_values_threshold): Implement target hook.
1975 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
1977 * config/s390/s390.h (struct s390_frame_layout)
1978 (struct machine_function): Move here from s390.c.
1979 (TARGET_INDIRECT_BRANCH_NOBP_RET)
1980 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
1981 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
1982 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
1983 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
1984 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
1985 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
1986 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
1987 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
1988 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
1989 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
1990 (mnemonic attribute): Add values which aren't recognized
1992 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
1993 pattern for branch conversion. Fix mnemonic attribute.
1994 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
1995 indirect branch via thunk if requested.
1996 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
1997 ("*indirect_jump"): Disable for branch conversion using out of
1999 ("indirect_jump_via_thunk<mode>_z10")
2000 ("indirect_jump_via_thunk<mode>")
2001 ("indirect_jump_via_inlinethunk<mode>_z10")
2002 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
2003 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
2004 ("casesi_jump_via_inlinethunk<mode>_z10")
2005 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
2006 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
2007 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
2008 ("*indirect2_jump"): Disable for branch conversion.
2009 ("casesi_jump"): Turn into expander and expand patterns for branch
2011 ("return_use"): New expander.
2012 ("*return"): Emit return via thunk and rename it to ...
2013 ("*return<mode>"): ... this one.
2014 * config/s390/s390.opt: Add new options and and enum for the
2017 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2019 * lra-constraints.c (match_reload): Unconditionally use
2020 gen_lowpart_SUBREG, rather than selecting between that
2021 and equivalent gen_rtx_SUBREG code.
2023 2018-02-08 Richard Biener <rguenther@suse.de>
2025 PR tree-optimization/84233
2026 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
2027 changed flag instead of boguously re-using phi_inserted.
2029 2018-02-08 Martin Jambor <mjambor@suse.cz>
2031 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
2032 static local variables.
2034 2018-02-08 Richard Biener <rguenther@suse.de>
2036 PR tree-optimization/84278
2037 * tree-vect-stmts.c (vectorizable_store): When looking for
2038 smaller vector types to perform grouped strided loads/stores
2039 make sure the mode is supported by the target.
2040 (vectorizable_load): Likewise.
2042 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2044 * config/aarch64/aarch64.c (aarch64_components_for_bb):
2045 Increase LDP/STP opportunities by adding adjacent callee-saves.
2047 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2049 PR rtl-optimization/84068
2050 PR rtl-optimization/83459
2051 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
2053 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
2055 PR tree-optimization/84224
2056 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
2057 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
2060 2018-02-07 Iain Sandoe <iain@codesourcery.com>
2063 * config/rs6000/altivec.md (*restore_world): Remove LR use.
2064 * config/rs6000/predicates.md (restore_world_operation): Adjust op
2065 count, remove one USE.
2067 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
2069 * doc/install.texi (Configuration): Document the
2070 --with-long-double-format={ibm,ieee} PowerPC configuration
2074 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
2075 Convert from define_expand to be define_insn_and_split. Rework
2076 float/double/_Float128 conversions to QI/HI/SImode to work with
2077 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
2078 conversions to QI/HImode types did a store and then a load to
2079 truncate the value. For conversions to VSX registers, don't split
2080 the insn, instead emit the code directly. Use the code iterator
2081 any_fix to combine signed and unsigned conversions.
2082 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
2083 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
2084 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2085 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
2086 (fix_<mode>di2_hw): Likewise.
2087 (fixuns_<mode>di2_hw): Likewise.
2088 (fix_<mode>si2_hw): Likewise.
2089 (fixuns_<mode>si2_hw): Likewise.
2090 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
2091 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2092 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
2093 fix<uns>_trunc<SFDF:mode>si2_p8.
2094 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
2096 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
2097 (fix<uns>_<mode>_mem): Likewise.
2098 (fctiw<u>z_<mode>_mem): Likewise.
2099 (fix<uns>_<mode>_mem): Likewise.
2100 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
2101 the register allocator from doing a direct move to the GPRs to do
2102 a store, and instead use the ISA 3.0 store byte/half-word from
2103 vector register instruction. For IEEE 128-bit floating point,
2104 also optimize stores of 32-bit ints.
2105 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
2107 2018-02-07 Alan Hayward <alan.hayward@arm.com>
2109 * genextract.c (push_pathstr_operand): New function to support
2111 (walk_rtx): Call push_pathstr_operand.
2112 (print_path): Support [a-zA-Z].
2114 2018-02-07 Richard Biener <rguenther@suse.de>
2116 PR tree-optimization/84037
2117 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
2118 (cse_and_gimplify_to_preheader): Declare.
2119 (vect_get_place_in_interleaving_chain): Likewise.
2120 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
2122 (_loop_vec_info::~_loop_vec_info): Delete it.
2123 (cse_and_gimplify_to_preheader): New function.
2124 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
2125 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
2126 (vectorizable_load): Likewise. For grouped stores always base
2127 the IV on the first element.
2128 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
2129 condition before gimplifying.
2131 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2133 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
2134 *DIV_EXPR and *MOD_EXPR.
2136 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
2139 * config/i386/i386.c (ix86_option_override_internal): Mask out
2140 the CF_SET bit when checking -fcf-protection.
2142 2018-02-07 Tom de Vries <tom@codesourcery.com>
2145 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
2148 2018-02-07 Richard Biener <rguenther@suse.de>
2150 PR tree-optimization/84204
2151 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
2154 PR tree-optimization/84205
2155 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
2156 special-case isl_ast_op_zdiv_r.
2158 PR tree-optimization/84223
2159 * graphite-scop-detection.c (gather_bbs::before_dom_children):
2160 Only add conditions from within the region.
2161 (gather_bbs::after_dom_children): Adjust.
2163 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
2166 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
2167 * config/avr/avr.md: Only post-reload split REG-REG moves if
2168 either register is GENERAL_REG_P.
2170 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2172 PR tree-optimization/84235
2173 * tree-ssa-scopedtables.c
2174 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
2175 if the subtraction is performed in floating point type where NaNs are
2176 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
2177 build 1. Formatting fix.
2179 2018-02-06 Jakub Jelinek <jakub@redhat.com>
2182 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
2183 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
2184 and skip it regardless of bb boundaries. Use CALL_P macro,
2185 don't test INSN_P (insn) together with CALL_P or JUMP_P check
2186 unnecessarily, formatting fix.
2188 2018-02-06 Michael Collison <michael.collison@arm.com>
2190 * config/arm/thumb2.md:
2191 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
2192 (*thumb_mov_notscc): Ditto.
2194 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
2197 * config/rs6000/rs6000.md (su code attribute): Use "u" for
2198 unsigned_fix, not "s".
2200 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2202 * configure.ac (gcc_fn_eh_frame_ro): New function.
2203 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
2204 correct .eh_frame permissions.
2205 * configure: Regenerate.
2207 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
2209 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
2212 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2214 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2215 Display warning message for -mno-speculate-indirect-jumps.
2217 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
2219 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
2221 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
2223 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
2225 PR tree-optimization/84225
2226 * tree-eh.c (find_trapping_overflow): Only call
2227 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
2229 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2232 * config/i386/i386.c: Reimplement the check of possible options
2233 -mibt/-mshstk conbination. Change error messages.
2234 * doc/invoke.texi: Fix a typo: remove extra '='.
2236 2018-02-06 Marek Polacek <polacek@redhat.com>
2238 PR tree-optimization/84228
2239 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
2241 2018-02-06 Tamar Christina <tamar.christina@arm.com>
2244 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
2245 emitted arch directives.
2246 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
2247 __ARM_FEATURE_COPROC before changing architectures.
2249 2018-02-06 Richard Biener <rguenther@suse.de>
2251 * config/i386/i386.c (print_reg): Fix typo.
2252 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
2254 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
2256 * configure: Regenerate.
2258 2018-02-05 Martin Sebor <msebor@redhat.com>
2260 PR tree-optimization/83369
2261 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
2264 2018-02-05 Martin Liska <mliska@suse.cz>
2266 * doc/invoke.texi: Cherry-pick upstream r323995.
2268 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
2270 * ira.c (ira_init_register_move_cost): Adjust comment.
2272 2018-02-05 Martin Liska <mliska@suse.cz>
2274 PR gcov-profile/84137
2275 * doc/gcov.texi: Fix typo in documentation.
2277 2018-02-05 Martin Liska <mliska@suse.cz>
2279 PR gcov-profile/83879
2280 * doc/gcov.texi: Document necessity of --dynamic-list-data when
2281 using dlopen functionality.
2283 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
2285 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
2286 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
2287 _mm_maskz_range_ss, _mm_mask_range_round_ss,
2288 _mm_maskz_range_round_ss): New intrinsics.
2289 (__builtin_ia32_rangesd128_round)
2290 (__builtin_ia32_rangess128_round): Remove.
2291 (__builtin_ia32_rangesd128_mask_round,
2292 __builtin_ia32_rangess128_mask_round): New builtins.
2293 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
2294 __builtin_ia32_rangess128_round): Remove.
2295 (__builtin_ia32_rangesd128_mask_round,
2296 __builtin_ia32_rangess128_mask_round): New builtins.
2297 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
2298 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
2299 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
2300 "<round_saeonly_constraint>")): Changed to ...
2301 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
2302 "<round_saeonly_scalar_constraint>")): ... this.
2303 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
2304 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
2305 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
2306 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
2307 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
2309 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
2311 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
2313 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
2314 Remove all values except native, 8540 and 8548.
2316 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
2318 * config/i386/i386.c (ix86_output_function_return): Pass
2319 INVALID_REGNUM, instead of -1, as invalid register number to
2320 indirect_thunk_name and output_indirect_thunk.
2322 2018-02-02 Julia Koval <julia.koval@intel.com>
2324 * config.gcc: Add -march=icelake.
2325 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
2326 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
2327 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
2328 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
2329 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
2330 (processor_target_table): Add icelake.
2331 (ix86_option_override_internal): Handle new PTAs.
2332 (get_builtin_code_for_version): Handle icelake.
2333 (M_INTEL_COREI7_ICELAKE): New.
2334 (fold_builtin_cpu): Handle icelake.
2335 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
2336 * doc/invoke.texi: Add -march=icelake.
2338 2018-02-02 Julia Koval <julia.koval@intel.com>
2340 * config/i386/i386.c (ix86_option_override_internal): Change flags type
2341 to wide_int_bitmask.
2342 * wide-int-bitmask.h: New.
2344 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2347 * config/i386/i386.md: Replace Pmode with word_mode in
2348 builtin_setjmp_setup and builtin_longjmp to support x32.
2350 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
2354 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
2356 (rs6000_supported_cpu_names): New static variable.
2357 (linux_cpu_translation_table): Likewise.
2358 (elf_platform) <cpu>: Define new static variable and use it.
2359 Translate kernel AT_PLATFORM name to canonical name if needed.
2360 Error if platform name is unknown.
2362 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
2365 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
2367 2018-02-01 Jeff Law <law@redhat.com>
2370 * config/i386/i386.c (release_scratch_register_on_entry): Add new
2371 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
2372 the scratch if RELEASE_VIA_POP is false.
2373 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
2374 If we have to save a temporary register, decrement SIZE appropriately.
2375 Pass new arguments to release_scratch_register_on_entry.
2376 (ix86_adjust_stack_and_probe): Likewise.
2377 (ix86_emit_probe_stack_range): Pass new arguments to
2378 release_scratch_register_on_entry.
2380 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
2382 PR rtl-optimization/84157
2383 * combine.c (change_zero_ext): Use REG_P predicate in
2384 front of HARD_REGISTER_P predicate.
2386 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
2388 * config/avr/avr.c (avr_option_override): Move disabling of
2389 -fdelete-null-pointer-checks to...
2390 * common/config/avr/avr-common.c (avr_option_optimization_table):
2393 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2395 PR tree-optimization/81635
2396 * tree-data-ref.c (split_constant_offset_1): For types that
2397 wrap on overflow, try to use range info to prove that wrapping
2400 2018-02-01 Renlin Li <renlin.li@arm.com>
2403 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
2405 (aarch64_register_move_cost): Likewise.
2406 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
2408 (REG_CLASS_NAMES): Likewise.
2409 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
2410 TAILCALL_ADDR_REGS. Remove IP registers.
2411 * config/aarch64/aarch64.md (Ucs): Update register constraint.
2413 2018-02-01 Richard Biener <rguenther@suse.de>
2415 * domwalk.h (dom_walker::dom_walker): Add additional constructor
2416 for specifying RPO order and allow NULL for that.
2417 * domwalk.c (dom_walker::dom_walker): Likewise.
2418 (dom_walker::walk): Handle NULL RPO order.
2419 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
2421 (rewrite_update_dom_walker): Likewise.
2422 (mark_def_dom_walker): Likewise.
2424 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2426 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
2427 (aarch64_maybe_expand_sve_subreg_move): Declare.
2428 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
2429 * config/aarch64/predicates.md (aarch64_any_register_operand): New
2431 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
2432 that are semantically a reverse operation.
2433 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
2434 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
2435 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
2437 (aarch64_can_change_mode_class): For big-endian, forbid changes
2438 between two SVE modes if they have different element sizes.
2440 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2442 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
2443 the TImode handling for big-endian targets.
2445 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2447 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
2448 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
2450 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
2451 Remove BSWAP handing for big-endian targets and use the form of
2452 LD1RQ appropariate for the mode.
2454 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2456 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
2457 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
2460 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2463 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
2464 check for operands that need to go through aarch64_sve_reload_be.
2466 2018-02-01 Jakub Jelinek <jakub@redhat.com>
2468 PR tree-optimization/81661
2469 PR tree-optimization/84117
2470 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
2471 * tree-eh.c: Include gimplify.h.
2472 (find_trapping_overflow, replace_trapping_overflow,
2473 rewrite_to_non_trapping_overflow): New functions.
2474 * tree-vect-loop.c: Include tree-eh.h.
2475 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
2476 * tree-data-ref.c: Include tree-eh.h.
2477 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
2479 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
2481 PR rtl-optimization/84123
2482 * combine.c (change_zero_ext): Check if hard register satisfies
2483 can_change_dest_mode before calling gen_lowpart_SUBREG.
2485 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
2488 * ira.c (ira_init_register_move_cost): Remove assert.
2490 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2492 PR rtl-optimization/84071
2493 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
2494 * doc/tm.texi: Regenerate.
2496 2018-01-31 Richard Biener <rguenther@suse.de>
2498 PR tree-optimization/84132
2499 * tree-data-ref.c (analyze_miv_subscript): Properly
2500 check whether evolution_function_is_affine_multivariate_p
2501 before calling gcd_of_steps_may_divide_p.
2503 2018-01-31 Julia Koval <julia.koval@intel.com>
2506 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
2507 * config/i386/i386.md (rdpid_rex64) New.
2508 (rdpid): Make 32bit only.
2510 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
2513 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
2514 an IDENTIFIER_NODE for FUNCTION_TYPE's.
2516 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2519 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
2521 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
2523 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2525 PR rtl-optimization/84071
2526 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
2527 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
2529 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2531 * config/arc/arc.c (arc_handle_aux_attribute): New function.
2532 (arc_attribute_table): Add 'aux' attribute.
2533 (arc_in_small_data_p): Consider aux like variables.
2534 (arc_is_aux_reg_p): New function.
2535 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
2536 (arc_get_aux_arg): New function.
2537 (prepare_move_operands): Handle aux-register access.
2538 (arc_handle_aux_attribute): New function.
2539 * doc/extend.texi (ARC Variable attributes): Add subsection.
2541 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2543 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
2544 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
2545 (arc_attribute_table): Add 'uncached' attribute.
2546 (arc_print_operand): Print '.di' flag for uncached memory
2548 (arc_in_small_data_p): Do not consider for small data the uncached
2550 (arc_is_uncached_mem_p): New function.
2551 * config/arc/predicates.md (compact_store_memory_operand): Check
2552 for uncached memory accesses.
2553 (nonvol_nonimm_operand): Likewise.
2554 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
2556 2018-01-31 Jakub Jelinek <jakub@redhat.com>
2559 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
2560 falign-loops=): Add Optimization flag.
2562 2018-01-30 Jeff Law <law@redhat.com>
2565 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
2566 INT_REGISTERS_SAVED. Check it prior to calling
2567 get_scratch_register_on_entry.
2568 (ix86_adjust_stack_and_probe): Similarly.
2569 (ix86_emit_probe_stack_range): Similarly.
2570 (ix86_expand_prologue): Corresponding changes.
2572 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2575 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
2576 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
2578 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
2581 * lra-constraints.c (curr_insn_transform): Process AND in the
2584 2018-01-30 Jakub Jelinek <jakub@redhat.com>
2586 PR rtl-optimization/83986
2587 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
2588 dependence against last_pending_memory_flush in addition to
2591 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
2593 PR tree-optimization/81611
2594 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
2597 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2600 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
2603 2018-01-30 Richard Biener <rguenther@suse.de>
2604 Jakub Jelinek <jakub@redhat.com>
2606 PR tree-optimization/84111
2607 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
2608 inner loops added during recursion, as they don't have up-to-date
2611 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2614 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
2615 (can_inline_edge_by_limits_p): ... here.
2616 (can_early_inline_edge_p, check_callers,
2617 update_caller_keys, update_callee_keys, recursive_inlining,
2618 add_new_edges_to_heap, speculation_useful_p,
2619 inline_small_functions,
2620 inline_small_functions, flatten_function,
2621 inline_to_all_callers_1): Update.
2623 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2625 * profile-count.c (profile_count::combine_with_ipa_count): Handle
2628 2018-01-30 Richard Biener <rguenther@suse.de>
2630 PR tree-optimization/83008
2631 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
2632 invariant and constant vector uses in stmts when they need
2635 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2638 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
2639 * configure: Regenerate.
2641 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2643 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
2645 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
2646 Use gen_rtx_REG rather than gen_lowpart.
2648 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2650 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
2651 rather than 0 when creating partial subregs.
2653 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2655 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
2658 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
2661 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
2662 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
2663 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
2664 flags. This restores the settings used before the 2017-07-24.
2665 Turning off pre increment/decrement/modify allows IVOPTS to
2666 optimize DF/SF loops where the index is an int.
2668 2018-01-29 Richard Biener <rguenther@suse.de>
2669 Kelvin Nilsen <kelvin@gcc.gnu.org>
2672 * tree-vect-stmts.c (vectorizable_call): Don't call
2673 targetm.vectorize_builtin_md_vectorized_function if callee is
2676 2018-01-22 Carl Love <cel@us.ibm.com>
2678 * doc/extend.tex: Fix typo in second arg in
2679 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
2681 2018-01-29 Richard Biener <rguenther@suse.de>
2683 PR tree-optimization/84086
2684 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
2685 (flush_ssaname_freelist): When SSA names were released reset
2686 the SCEV hash table.
2688 2018-01-29 Richard Biener <rguenther@suse.de>
2690 PR tree-optimization/84057
2691 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
2692 removed paths when removing edges.
2694 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
2696 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
2697 -mfunction-return=@var{choice}.
2699 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
2702 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
2704 (layout::print_source_line): Likewise.
2705 (test_get_line_width_without_trailing_whitespace): Add test cases.
2707 2018-01-27 Jakub Jelinek <jakub@redhat.com>
2710 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
2713 2018-01-26 Jim Wilson <jimw@sifive.com>
2715 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
2717 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
2720 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2722 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
2723 and CMP + SUB-immediate -> SUBS.
2725 2018-01-26 Martin Sebor <msebor@redhat.com>
2727 PR tree-optimization/83896
2728 * tree-ssa-strlen.c (get_string_len): Rename...
2729 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
2730 Avoid assuming length is constant.
2731 (handle_char_store): Use HOST_WIDE_INT for string length.
2733 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
2736 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
2737 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
2739 2018-01-26 Richard Biener <rguenther@suse.de>
2741 PR rtl-optimization/84003
2742 * dse.c (record_store): Only record redundant stores when
2743 the earlier store aliases at least all accesses the later one does.
2745 2018-01-26 Jakub Jelinek <jakub@redhat.com>
2747 PR rtl-optimization/83985
2748 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
2749 REG_CFA_RESTORE insns.
2750 (delete_unmarked_insns): Don't ignore separate shrink wrapping
2751 REG_CFA_RESTORE insns here.
2754 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
2755 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
2757 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2759 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
2760 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
2761 (arc_init): Likewise.
2762 (arc_override_options): Likewise.
2763 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
2765 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
2767 * config/arc/arc.h (TARGET_DBNZ): Define.
2768 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
2769 properly set the tune attribute.
2770 (dbnz): Use TARGET_DBNZ guard.
2771 * config/arc/arc.opt (mtune): Add core3 option.
2773 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2775 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
2776 recognize new pic like addresses.
2777 (arc_delegitimize_address): Clean up.
2779 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2781 * config/arc/arc-arches.def: Option mrf16 valid for all
2783 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
2784 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
2785 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
2786 * config/arc/arc-tables.opt: Regenerate.
2787 * config/arc/arc.c (arc_conditional_register_usage): Handle
2788 reduced register file case.
2789 (arc_file_start): Set must have build attributes.
2790 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
2792 * config/arc/arc.opt (mrf16): Add new option.
2793 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
2794 * config/arc/genmultilib.awk: Handle new mrf16 option.
2795 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
2796 * config/arc/t-multilib: Regenerate.
2797 * doc/invoke.texi (ARC Options): Document mrf16 option.
2799 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2801 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
2802 * config/arc/arc.c (arc_handle_secure_attribute): New function.
2803 (arc_attribute_table): Add 'secure_call' attribute.
2804 (arc_print_operand): Print secure call operand.
2805 (arc_function_ok_for_sibcall): Don't optimize tail calls when
2807 (arc_is_secure_call_p): New function. * config/arc/arc.md
2808 (call_i): Add support for sjli instruction.
2809 (call_value_i): Likewise.
2810 * config/arc/constraints.md (Csc): New constraint.
2812 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2813 John Eric Martin <John.Martin@emmicro-us.com>
2815 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
2816 * config/arc/arc.c (_arc_jli_section): New struct.
2817 (arc_jli_section): New type.
2818 (rc_jli_sections): New static variable.
2819 (arc_handle_jli_attribute): New function.
2820 (arc_attribute_table): Add jli_always and jli_fixed attribute.
2821 (arc_file_end): New function.
2822 (TARGET_ASM_FILE_END): Define.
2823 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
2824 (arc_add_jli_section): New function.
2825 (jli_call_scan): Likewise.
2826 (arc_reorg): Call jli_call_scan.
2827 (arc_output_addsi): Remove 'S' from printing asm operand.
2828 (arc_is_jli_call_p): New function.
2829 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
2831 (movhi_insn): Likewise.
2832 (movsi_insn): Likewise.
2833 (movsi_set_cc_insn): Likewise.
2834 (loadqi_update): Likewise.
2835 (load_zeroextendqisi_update): Likewise.
2836 (load_signextendqisi_update): Likewise.
2837 (loadhi_update): Likewise.
2838 (load_zeroextendhisi_update): Likewise.
2839 (load_signextendhisi_update): Likewise.
2840 (loadsi_update): Likewise.
2841 (loadsf_update): Likewise.
2842 (movsicc_insn): Likewise.
2843 (bset_insn): Likewise.
2844 (bxor_insn): Likewise.
2845 (bclr_insn): Likewise.
2846 (bmsk_insn): Likewise.
2847 (bicsi3_insn): Likewise.
2848 (cmpsi_cc_c_insn): Likewise.
2849 (movsi_ne): Likewise.
2850 (movsi_cond_exec): Likewise.
2851 (clrsbsi2): Likewise.
2862 (call_i): Remove 'S' asm letter, add jli instruction.
2863 (call_value_i): Likewise.
2864 * config/arc/arc.op (mjli-always): New option.
2865 * config/arc/constraints.md (Cji): New constraint.
2866 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
2868 (subsf3_fpx): Likewise.
2869 (mulsf3_fpx): Likewise.
2870 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
2872 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
2874 * doc/invoke.texi (ARC): Document mjli-always option.
2876 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
2878 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
2879 avoid addition with 0 and use incw and decw where possible.
2881 2018-01-26 Richard Biener <rguenther@suse.de>
2883 PR tree-optimization/81082
2884 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
2885 association if it requires casting to unsigned.
2886 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
2887 from fold_plusminus_mult_expr to catch important cases late when
2888 range info is available.
2890 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2892 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
2893 * configure.ac (hidden_linkonce): New test.
2894 * configure: Regenerate.
2895 * config.in: Regenerate.
2897 2018-01-26 Julia Koval <julia.koval@intel.com>
2899 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
2900 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
2901 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
2902 _mm_mask_bitshuffle_epi64_mask): Fix type.
2903 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
2904 USI_FTYPE_V4DI_V4DI_USI): Remove.
2905 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
2906 __builtin_ia32_vpshufbitqmb256_mask,
2907 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
2908 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
2909 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
2911 2018-01-26 Alan Modra <amodra@gmail.com>
2914 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
2915 UNSPEC_VBPERMQ. Sort other unspecs.
2917 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
2919 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
2921 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
2924 * predict.c (drop_profile): Do not push/pop cfun; update also
2926 (handle_missing_profiles): Fix logic looking for zero profiles.
2928 2018-01-25 Jakub Jelinek <jakub@redhat.com>
2931 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
2932 on functions with #pragma omp declare simd or functions with simd
2934 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
2935 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
2936 Remove trailing \n from warning_at calls.
2938 2018-01-25 Tom de Vries <tom@codesourcery.com>
2941 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
2942 for neutered workers.
2944 2018-01-24 Joseph Myers <joseph@codesourcery.com>
2947 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
2948 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
2950 2018-01-24 Jeff Law <law@redhat.com>
2953 * i386.c (get_probe_interval): Move to earlier point.
2954 (ix86_compute_frame_layout): If -fstack-clash-protection and
2955 the frame is larger than the probe interval, then use pushes
2956 to save registers rather than reg->mem moves.
2957 (ix86_expand_prologue): Remove conditional for int_registers_saved
2960 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
2963 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
2964 min/max for never referenced object.
2966 2018-01-24 Jakub Jelinek <jakub@redhat.com>
2969 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
2971 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
2972 attributes from DECL_ATTRIBUTES (decl) without affecting
2973 DECL_ATTRIBUTES (current_function_decl).
2974 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
2975 functions with non-NULL DECL_ABSTRACT_ORIGIN.
2977 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
2979 PR tree-optimization/83979
2980 * fold-const.c (fold_comparison): Use constant_boolean_node
2981 instead of boolean_{true,false}_node.
2983 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
2985 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
2988 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2990 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
2991 Simplify the clause that sets the length attribute.
2992 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
2993 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
2994 clause that sets the length attribute.
2995 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
2997 2018-01-24 Tom de Vries <tom@codesourcery.com>
3000 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
3001 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
3002 Add strict parameter.
3003 (prevent_branch_around_nothing): Insert dummy insn between branch to
3004 label and label with no ptx insn inbetween.
3005 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
3007 2018-01-24 Tom de Vries <tom@codesourcery.com>
3010 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3011 for neutered threads in warp.
3012 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
3014 2018-01-24 Richard Biener <rguenther@suse.de>
3016 PR tree-optimization/83176
3017 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
3020 2018-01-24 Richard Biener <rguenther@suse.de>
3022 PR tree-optimization/82819
3023 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
3024 code generating pluses that are no-ops in the target precision.
3026 2018-01-24 Richard Biener <rguenther@suse.de>
3029 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
3031 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3033 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
3034 to merge probabilities.
3035 * predict.c (probably_never_executed): Also mark as cold functions
3036 with global 0 profile and guessed local profile.
3037 * profile-count.c (profile_probability::combine_with_count): New
3039 * profile-count.h (profile_probability::operator*,
3040 profile_probability::operator*=, profile_probability::operator/,
3041 profile_probability::operator/=): Reduce precision to adjusted
3042 and set value to guessed on contradictory divisions.
3043 (profile_probability::combine_with_freq): Remove.
3044 (profile_probability::combine_wiht_count): Declare.
3045 (profile_count::force_nonzero):: Set to adjusted.
3046 (profile_count::probability_in):: Set quality to adjusted.
3047 * tree-ssa-tail-merge.c (replace_block_by): Use
3050 2018-01-23 Andrew Waterman <andrew@sifive.com>
3051 Jim Wilson <jimw@sifive.com>
3053 * config/riscv/riscv.c (riscv_stack_boundary): New.
3054 (riscv_option_override): Set riscv_stack_boundary. Handle
3055 riscv_preferred_stack_boundary_arg.
3056 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
3057 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
3058 (STACK_BOUNDARY): Set to riscv_stack_boundary.
3059 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
3060 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
3061 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
3063 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
3066 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
3067 of struct ix86_frame.
3068 (ix86_expand_epilogue): Likewise. Add a local variable for
3069 the reg_save_offset field in struct ix86_frame.
3071 2018-01-23 Bin Cheng <bin.cheng@arm.com>
3073 PR tree-optimization/82604
3074 * tree-loop-distribution.c (enum partition_kind): New enum item
3075 PKIND_PARTIAL_MEMSET.
3076 (partition_builtin_p): Support above new enum item.
3077 (generate_code_for_partition): Ditto.
3078 (compute_access_range): Differentiate cases that equality can be
3079 proven at all loops, the innermost loops or no loops.
3080 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
3081 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
3082 (finalize_partitions, distribute_loop): Don't fuse partition of
3083 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
3084 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
3087 2018-01-23 Martin Liska <mliska@suse.cz>
3089 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
3090 order to ignore the predictor.
3091 (PRED_POLYMORPHIC_CALL): Likewise.
3092 (PRED_RECURSIVE_CALL): Likewise.
3094 2018-01-23 Martin Liska <mliska@suse.cz>
3096 * tree-profile.c (tree_profiling): Print function header to
3097 aware reader which function we are working on.
3098 * value-prof.c (gimple_find_values_to_profile): Do not print
3099 not interesting value histograms.
3101 2018-01-23 Martin Liska <mliska@suse.cz>
3103 * profile-count.h (enum profile_quality): Add
3104 profile_uninitialized as the first value. Do not number values
3105 as they are zero based.
3106 (profile_count::verify): Update sanity check.
3107 (profile_probability::verify): Likewise.
3109 2018-01-23 Nathan Sidwell <nathan@acm.org>
3111 * doc/invoke.texi (ffor-scope): Deprecate.
3113 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3115 PR tree-optimization/83510
3116 * domwalk.c (set_all_edges_as_executable): New function.
3117 (dom_walker::dom_walker): Convert bool param
3118 "skip_unreachable_blocks" to enum reachability. Move setup of
3119 edge flags to set_all_edges_as_executable and only do it when
3120 reachability is REACHABLE_BLOCKS.
3121 * domwalk.h (enum dom_walker::reachability): New enum.
3122 (dom_walker::dom_walker): Convert bool param
3123 "skip_unreachable_blocks" to enum reachability.
3124 (set_all_edges_as_executable): New decl.
3125 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
3126 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
3128 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
3129 but converting true to REACHABLE_BLOCKS.
3130 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
3132 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
3133 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
3134 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
3136 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
3137 if check_all_array_refs will be called.
3139 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3141 * tree.c (selftest::test_location_wrappers): Add more test
3144 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3146 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
3147 (selftest::test_bit_in_range): Likewise.
3149 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3152 * doc/sourcebuild.texi (vect_float): Say that the selector
3153 only describes the situation when -funsafe-math-optimizations is on.
3154 (vect_float_strict): Document.
3156 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3158 PR tree-optimization/83965
3159 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
3160 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
3161 instead of checking only for a reduction.
3162 (vect_recog_widen_sum_pattern): Likewise.
3164 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3166 * predict.c (probably_never_executed): Only use precise profile info.
3167 (compute_function_frequency): Skip after inlining hack since we now
3168 have quality checking.
3170 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3172 * profile-count.h (profile_probability::very_unlikely,
3173 profile_probability::unlikely, profile_probability::even): Set
3174 precision to guessed.
3176 2018-01-23 Richard Biener <rguenther@suse.de>
3178 PR tree-optimization/83963
3179 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
3180 Properly terminate dominator walk when crossing the exit edge not
3181 when visiting its source block.
3183 2018-01-23 Jakub Jelinek <jakub@redhat.com>
3186 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
3187 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
3189 2018-01-22 Jakub Jelinek <jakub@redhat.com>
3191 PR tree-optimization/83957
3192 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
3193 semicolon after for body surrounded by braces.
3195 PR tree-optimization/83081
3196 * profile-count.h (profile_probability::split): New method.
3197 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
3198 Use profile_probability::split.
3199 (do_compare_rtx_and_jump): Fix adjustment of probabilities
3200 when splitting a single conditional jump into 2.
3202 2018-01-22 David Malcolm <dmalcolm@redhat.com>
3204 PR tree-optimization/69452
3205 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
3208 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3210 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
3211 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
3212 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
3214 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3216 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
3217 * config/rl78/rl78.md (movdi): New define_expand.
3218 * config/rl78/rl78.c (rl78_split_movdi): New function.
3220 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
3223 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
3225 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
3226 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
3227 128-bit to produce an UNSPEC move to get the double word with the
3228 signbit and then a shift directly to do signbit.
3229 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
3230 implementation with a new version that just does either a direct
3231 move or a regular move. Move memory interface to separate insns.
3232 Move insns so they are next to the expander.
3233 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
3234 with signbit move. Split big and little endian case.
3235 (signbit<mode>2_dm_mem_le): Likewise.
3236 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
3237 (signbit<mode>2_dm2): Likewise.
3239 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3241 * config/rl78/rl78.md (anddi3): New define_expand.
3243 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3245 * config/rl78/rl78.md (umindi3): New define_expand.
3247 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3249 * config/rl78/rl78.md (smindi3): New define_expand.
3251 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3253 * config/rl78/rl78.md (smaxdi3): New define_expand.
3255 2018-01-22 Carl Love <cel@us.ibm.com>
3257 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
3258 LVX_V1TI): Add macro expansion.
3259 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
3260 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
3261 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
3262 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
3263 Change check to determine if the instruction is a byte reversing
3264 entry. Fix typo in comment.
3265 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
3266 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
3267 Add def_builtin calls for new builtins.
3268 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
3269 Add define_insn expansion.
3271 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3273 * config/rl78/rl78.md (umaxdi3): New define_expand.
3275 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3277 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
3278 for non-QImode registers.
3280 2018-01-22 Richard Biener <rguenther@suse.de>
3282 PR tree-optimization/83963
3283 * graphite-scop-detection.c (scop_detection::get_sese): Delay
3284 including the loop exit block.
3285 (scop_detection::merge_sese): Likewise.
3286 (scop_detection::add_scop): Do it here instead.
3288 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3290 * doc/sourcebuild.texi (arm_softfloat): Document.
3292 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
3295 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
3296 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
3297 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
3299 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3300 David Edelsohn <dje.gcc@gmail.com>
3303 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3304 Change "crset eq" to "crset 2".
3305 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3306 (*call_indirect_aix<mode>_nospec): Likewise.
3307 (*call_value_indirect_aix<mode>_nospec): Likewise.
3308 (*call_indirect_elfv2<mode>_nospec): Likewise.
3309 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
3310 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
3311 change assembly output from . to $.
3312 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3313 (indirect_jump<mode>_nospec): Change assembly output from . to $.
3314 (*tablejump<mode>_internal1_nospec): Likewise.
3316 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
3319 * config/sh/sh_optimize_sett_clrt.cc:
3320 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
3322 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
3324 PR tree-optimization/83940
3325 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
3326 offset_dt to vect_constant_def rather than vect_unknown_def_type.
3327 (vect_check_load_store_mask): Add a mask_dt_out parameter and
3328 use it to pass back the definition type.
3329 (vect_check_store_rhs): Likewise rhs_dt_out.
3330 (vect_build_gather_load_calls): Add a mask_dt argument and use
3331 it instead of a call to vect_is_simple_use.
3332 (vectorizable_store): Update calls to vect_check_load_store_mask
3333 and vect_check_store_rhs. Use the dt returned by the latter instead
3334 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
3335 instead of calls to vect_is_simple_use. Pass the scalar rather
3336 than the vector operand to vect_is_simple_use when handling
3337 second and subsequent copies of an rhs value.
3338 (vectorizable_load): Update calls to vect_check_load_store_mask
3339 and vect_build_gather_load_calls. Use the cached mask_dt and
3340 gs_info.offset_dt instead of calls to vect_is_simple_use.
3342 2018-01-20 Jakub Jelinek <jakub@redhat.com>
3345 * tree-emutls.c: Include gimplify.h.
3346 (lower_emutls_2): New function.
3347 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
3348 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
3349 it before further processing.
3352 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
3353 UINTVAL (trueop1) instead of INTVAL (op1).
3355 2018-01-19 Jakub Jelinek <jakub@redhat.com>
3359 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
3360 INCOMING_FRAME_SP_OFFSET if not defined.
3361 (scan_trace): Add ENTRY argument. If true and
3362 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
3363 emit a note to adjust the CFA offset.
3364 (create_cfi_notes): Adjust scan_trace callers.
3365 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
3366 INCOMING_FRAME_SP_OFFSET in the CIE.
3367 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
3368 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
3370 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
3371 * doc/tm.texi: Regenerated.
3373 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3375 PR rtl-optimization/83147
3376 * lra-constraints.c (remove_inheritance_pseudos): Use
3377 lra_substitute_pseudo_within_insn.
3379 2018-01-19 Tom de Vries <tom@codesourcery.com>
3380 Cesar Philippidis <cesar@codesourcery.com>
3383 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
3385 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
3388 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
3389 spaces for function labels.
3391 2018-01-19 Martin Liska <mliska@suse.cz>
3393 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
3394 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
3395 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
3396 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
3397 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
3398 (PRED_CONST_RETURN): Change from 69 to 65.
3399 (PRED_NULL_RETURN): Change from 91 to 71.
3400 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
3401 (PRED_LOOP_GUARD): Change from 66 to 73.
3403 2018-01-19 Martin Liska <mliska@suse.cz>
3405 * predict.c (predict_insn_def): Add new assert.
3406 (struct branch_predictor): Change type to signed integer.
3407 (test_prediction_value_range): Amend test to cover
3409 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
3410 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
3411 (PRED_LOOP_ITERATIONS_MAX): Likewise.
3412 (PRED_LOOP_IV_COMPARE): Likewise.
3413 * predict.h (PROB_UNINITIALIZED): Define new constant.
3415 2018-01-19 Martin Liska <mliska@suse.cz>
3417 * predict.c (dump_prediction): Add new format for
3418 analyze_brprob.py script which is enabled with -details
3420 * profile-count.h (precise_p): New function.
3422 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3424 PR tree-optimization/83922
3425 * tree-vect-loop.c (vect_verify_full_masking): Return false if
3426 there are no statements that need masking.
3427 (vect_active_double_reduction_p): New function.
3428 (vect_analyze_loop_operations): Use it when handling phis that
3429 are not in the loop header.
3431 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3433 PR tree-optimization/83914
3434 * tree-vect-loop.c (vectorizable_induction): Don't convert
3435 init_expr or apply the peeling adjustment for inductions
3436 that are nested within the vectorized loop.
3438 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3440 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
3443 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3447 * function.h (gimplify_parameters): Add gimple_seq * argument.
3448 * function.c: Include gimple.h and options.h.
3449 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
3450 for the added local temporaries if needed.
3451 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
3452 if there are any parameter cleanups, wrap whole body into a
3453 try/finally with the cleanups.
3455 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
3458 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
3459 Use GET_MODE_CLASS for scalar floating point.
3461 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3465 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
3466 Fix call of call_cgraph_insertion_hooks.
3468 2018-01-18 Martin Sebor <msebor@redhat.com>
3470 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
3472 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3475 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
3478 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
3481 * common.opt: (-ffile-prefix-map): New option.
3482 * opts.c (common_handle_option): Defer it.
3483 * opts-global.c (handle_common_deferred_options): Handle it.
3484 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
3485 * file-prefix-map.h: New file.
3486 (remap_debug_filename, add_debug_prefix_map): ...here.
3487 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
3488 * final.c (debug_prefix_map, add_debug_prefix_map
3489 remap_debug_filename): Move to...
3490 * file-prefix-map.c: New file.
3491 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
3492 generalize, get rid of alloca(), use strrchr() instead of strchr().
3493 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
3494 Implement in terms of add_prefix_map().
3495 (remap_macro_filename, remap_debug_filename): Implement in term of
3497 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
3498 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
3499 * dbxout.c: Include file-prefix-map.h.
3500 * varasm.c: Likewise.
3501 * vmsdbgout.c: Likewise.
3502 * xcoffout.c: Likewise.
3503 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
3504 * doc/cppopts.texi (-fmacro-prefix-map): Document.
3505 * doc/invoke.texi (-ffile-prefix-map): Document.
3506 (-fdebug-prefix-map): Update description.
3508 2018-01-18 Martin Liska <mliska@suse.cz>
3510 * config/i386/i386.c (indirect_thunk_name): Document that also
3512 (output_indirect_thunk): Document why both instructions
3513 (pause and lfence) are generated.
3515 2018-01-18 Richard Biener <rguenther@suse.de>
3517 PR tree-optimization/83887
3518 * graphite-scop-detection.c
3519 (scop_detection::get_nearest_dom_with_single_entry): Remove.
3520 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
3521 (scop_detection::merge_sese): Re-implement with a flood-fill
3522 algorithm that properly finds a SESE region if it exists.
3524 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3527 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
3528 pointer_diff optimizations use view_convert instead of convert.
3530 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3532 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3533 Generate different code for -mno-speculate-indirect-jumps.
3534 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3535 (*call_indirect_aix<mode>): Disable for
3536 -mno-speculate-indirect-jumps.
3537 (*call_indirect_aix<mode>_nospec): New define_insn.
3538 (*call_value_indirect_aix<mode>): Disable for
3539 -mno-speculate-indirect-jumps.
3540 (*call_value_indirect_aix<mode>_nospec): New define_insn.
3541 (*sibcall_nonlocal_sysv<mode>): Generate different code for
3542 -mno-speculate-indirect-jumps.
3543 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3545 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3547 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
3548 long double type, set the flags for noting the default long double
3549 type, even if we don't pass or return a long double type.
3551 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
3554 * ipa-inline.c (flatten_function): Do not overwrite final inlining
3557 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
3559 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
3560 support for merge[hl].
3561 (fold_mergehl_helper): New helper function.
3562 (tree-vector-builder.h): New #include for tree_vector_builder usage.
3563 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
3564 (altivec_vmrglw_direct): Add xxmrglw insn.
3566 2018-01-17 Andrew Waterman <andrew@sifive.com>
3568 * config/riscv/riscv.c (riscv_conditional_register_usage): If
3569 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
3571 2018-01-17 David Malcolm <dmalcolm@redhat.com>
3574 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
3575 call the lto_location_cache before reading the
3576 DECL_SOURCE_LOCATION of the types.
3578 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3579 Richard Sandiford <richard.sandiford@linaro.org>
3581 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
3582 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
3583 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
3584 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
3585 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
3587 * config/aarch64/constraints.md (aarch64_movti_operand):
3589 * config/aarch64/predicates.md (Uti): Add new constraint.
3591 2018-01-17 Carl Love <cel@us.ibm.com>
3593 * config/rs6000/vsx.md (define_expand xl_len_r,
3594 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
3595 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
3597 (define_expand, define_insn): Move the shift left from the
3598 define_insn to the define_expand for lxvl and stxvl instructions.
3599 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
3600 and XL_LEN_R definitions to PURE.
3602 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
3604 * config/i386/i386.c (indirect_thunk_name): Declare regno
3605 as unsigned int. Compare regno with INVALID_REGNUM.
3606 (output_indirect_thunk): Ditto.
3607 (output_indirect_thunk_function): Ditto.
3608 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
3609 in the call to output_indirect_thunk_function.
3611 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
3614 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
3615 rather than the size of inner_type to determine the stack slot size
3616 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
3618 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
3621 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
3624 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
3626 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
3627 endian Linux systems to optionally enable multilibs for selecting
3628 the long double type if the user configured an explicit type.
3629 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
3630 have no long double multilibs if not defined.
3631 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
3632 warn if the user used -mabi={ieee,ibm}longdouble and we built
3633 multilibs for long double.
3634 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
3635 appropriate multilib option.
3636 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
3638 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
3639 for building long double multilibs.
3640 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
3642 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
3644 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
3647 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
3649 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
3652 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
3655 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
3658 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3660 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
3661 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
3663 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
3665 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
3666 different rtl trees depending on TARGET_64BIT.
3667 (rs6000_gen_lvx): Likewise.
3669 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3671 * config/visium/visium.md (nop): Tweak comment.
3672 (hazard_nop): Likewise.
3674 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3676 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
3677 -mspeculate-indirect-jumps.
3678 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
3679 for -mno-speculate-indirect-jumps.
3680 (*call_indirect_elfv2<mode>_nospec): New define_insn.
3681 (*call_value_indirect_elfv2<mode>): Disable for
3682 -mno-speculate-indirect-jumps.
3683 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
3684 (indirect_jump): Emit different RTL for
3685 -mno-speculate-indirect-jumps.
3686 (*indirect_jump<mode>): Disable for
3687 -mno-speculate-indirect-jumps.
3688 (*indirect_jump<mode>_nospec): New define_insn.
3689 (tablejump): Emit different RTL for
3690 -mno-speculate-indirect-jumps.
3691 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
3692 (tablejumpsi_nospec): New define_expand.
3693 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
3694 (tablejumpdi_nospec): New define_expand.
3695 (*tablejump<mode>_internal1): Disable for
3696 -mno-speculate-indirect-jumps.
3697 (*tablejump<mode>_internal1_nospec): New define_insn.
3698 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
3701 2018-01-16 Artyom Skrobov tyomitch@gmail.com
3703 * caller-save.c (insert_save): Drop unnecessary parameter. All
3706 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3707 Richard Biener <rguenth@suse.de>
3710 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
3711 return early, inline manually is_gimple_sizepos. Make sure if we
3712 call gimplify_expr we don't end up with a gimple constant.
3713 * tree.c (variably_modified_type_p): Don't return true for
3714 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
3715 * gimplify.h (is_gimple_sizepos): Remove.
3717 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3719 PR tree-optimization/83857
3720 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
3721 vectorizable_live_operation for pure SLP statements.
3722 (vectorizable_live_operation): Handle PHIs.
3724 2018-01-16 Richard Biener <rguenther@suse.de>
3726 PR tree-optimization/83867
3727 * tree-vect-stmts.c (vect_transform_stmt): Precompute
3728 nested_in_vect_loop_p since the scalar stmt may get invalidated.
3730 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3733 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
3734 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
3735 If off is not INTEGER_CST, issue a may not be aligned warning
3736 rather than isn't aligned. Use isn%'t rather than isn't.
3737 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
3739 <case MULT_EXPR>: Improve the case when bottom and one of the
3740 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
3741 operand, in that case check if the other operand is multiple of
3742 bottom divided by the INTEGER_CST operand.
3744 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3747 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
3748 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
3749 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
3750 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
3751 * config/pa/pa.c (pa_function_arg_advance): Likewise.
3752 (pa_function_arg, pa_arg_partial_bytes): Likewise.
3753 (pa_function_arg_size): New function.
3755 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3757 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
3758 in a separate statement.
3760 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3762 PR tree-optimization/83847
3763 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
3764 group gathers and scatters.
3766 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3768 PR rtl-optimization/86620
3769 * params.def (max-sched-ready-insns): Bump minimum value to 1.
3771 PR rtl-optimization/83213
3772 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
3773 to last if both are JUMP_INSNs.
3775 PR tree-optimization/83843
3776 * gimple-ssa-store-merging.c
3777 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
3778 store_immediate_info for bswap/nop orig_stores.
3780 2018-01-15 Andrew Waterman <andrew@sifive.com>
3782 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
3784 <UDIV>: Increase cost if !TARGET_DIV.
3786 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
3788 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
3789 (define_attr "cr_logical_3op"): New.
3790 (cceq_ior_compare): Adjust.
3791 (cceq_ior_compare_complement): Adjust.
3792 (*cceq_rev_compare): Adjust.
3793 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
3794 (is_cracked_insn): Adjust.
3795 (insn_must_be_first_in_group): Adjust.
3796 * config/rs6000/40x.md: Adjust.
3797 * config/rs6000/440.md: Adjust.
3798 * config/rs6000/476.md: Adjust.
3799 * config/rs6000/601.md: Adjust.
3800 * config/rs6000/603.md: Adjust.
3801 * config/rs6000/6xx.md: Adjust.
3802 * config/rs6000/7450.md: Adjust.
3803 * config/rs6000/7xx.md: Adjust.
3804 * config/rs6000/8540.md: Adjust.
3805 * config/rs6000/cell.md: Adjust.
3806 * config/rs6000/e300c2c3.md: Adjust.
3807 * config/rs6000/e500mc.md: Adjust.
3808 * config/rs6000/e500mc64.md: Adjust.
3809 * config/rs6000/e5500.md: Adjust.
3810 * config/rs6000/e6500.md: Adjust.
3811 * config/rs6000/mpc.md: Adjust.
3812 * config/rs6000/power4.md: Adjust.
3813 * config/rs6000/power5.md: Adjust.
3814 * config/rs6000/power6.md: Adjust.
3815 * config/rs6000/power7.md: Adjust.
3816 * config/rs6000/power8.md: Adjust.
3817 * config/rs6000/power9.md: Adjust.
3818 * config/rs6000/rs64.md: Adjust.
3819 * config/rs6000/titan.md: Adjust.
3821 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3823 * config/i386/predicates.md (indirect_branch_operand): Rewrite
3824 ix86_indirect_branch_register logic.
3826 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3828 * config/i386/constraints.md (Bs): Update
3829 ix86_indirect_branch_register check. Don't check
3830 ix86_indirect_branch_register with GOT_memory_operand.
3832 * config/i386/predicates.md (GOT_memory_operand): Don't check
3833 ix86_indirect_branch_register here.
3834 (GOT32_symbol_operand): Likewise.
3836 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3838 * config/i386/predicates.md (constant_call_address_operand):
3839 Rewrite ix86_indirect_branch_register logic.
3840 (sibcall_insn_operand): Likewise.
3842 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3844 * config/i386/constraints.md (Bs): Replace
3845 ix86_indirect_branch_thunk_register with
3846 ix86_indirect_branch_register.
3848 * config/i386/i386.md (indirect_jump): Likewise.
3849 (tablejump): Likewise.
3850 (*sibcall_memory): Likewise.
3851 (*sibcall_value_memory): Likewise.
3852 Peepholes of indirect call and jump via memory: Likewise.
3853 * config/i386/i386.opt: Likewise.
3854 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3855 (GOT_memory_operand): Likewise.
3856 (call_insn_operand): Likewise.
3857 (sibcall_insn_operand): Likewise.
3858 (GOT32_symbol_operand): Likewise.
3860 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3863 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
3864 type rather than type addr's type points to.
3865 (expand_omp_atomic_mutex): Likewise.
3866 (expand_omp_atomic): Likewise.
3868 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3871 * config/i386/i386.c (output_indirect_thunk_function): Use
3872 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
3873 for __x86_return_thunk.
3875 2018-01-15 Richard Biener <rguenther@suse.de>
3878 * expmed.c (extract_bit_field_1): Fix typo.
3880 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3883 * config/arm/iterators.md (VF): New mode iterator.
3884 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
3885 Remove integer-related logic from pattern.
3886 (neon_vabd<mode>_3): Likewise.
3888 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3891 * common.opt (fstrict-overflow): No longer an alias.
3892 (fwrapv-pointer): New option.
3893 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
3894 also for pointer types based on flag_wrapv_pointer.
3895 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
3896 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
3897 opts->x_flag_wrapv got set.
3898 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
3899 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
3900 POINTER_TYPE_OVERFLOW_UNDEFINED.
3901 * match.pd: Likewise in address comparison pattern.
3902 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
3904 2018-01-15 Richard Biener <rguenther@suse.de>
3907 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
3908 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
3909 Reset type names to their identifier if their TYPE_DECL doesn't
3910 have linkage (and thus is used for ODR and devirt).
3911 (save_debug_info_for_decl): Remove.
3912 (save_debug_info_for_type): Likewise.
3913 (add_tree_to_fld_list): Adjust.
3914 * tree-pretty-print.c (dump_generic_node): Make dumping of
3915 type names more robust.
3917 2018-01-15 Richard Biener <rguenther@suse.de>
3919 * BASE-VER: Bump to 8.0.1.
3921 2018-01-14 Martin Sebor <msebor@redhat.com>
3924 * builtins.c (check_access): Avoid warning when the no-warning bit
3927 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
3929 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
3930 * ira-color (allocno_hard_regs_compare): Likewise.
3932 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
3935 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
3936 Use .pushsection/.popsection.
3938 2018-01-14 Martin Sebor <msebor@redhat.com>
3941 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
3943 2018-01-14 Jakub Jelinek <jakub@redhat.com>
3945 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
3946 entry from extra_headers.
3947 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
3948 extra_headers, make the list bitwise identical to the i?86-*-* one.
3950 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3952 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
3953 -mcmodel=large with -mindirect-branch=thunk,
3954 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
3955 -mfunction-return=thunk-extern.
3956 * doc/invoke.texi: Document -mcmodel=large is incompatible with
3957 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
3958 -mfunction-return=thunk and -mfunction-return=thunk-extern.
3960 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3962 * config/i386/i386.c (print_reg): Print the name of the full
3963 integer register without '%'.
3964 (ix86_print_operand): Handle 'V'.
3965 * doc/extend.texi: Document 'V' modifier.
3967 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3969 * config/i386/constraints.md (Bs): Disallow memory operand for
3970 -mindirect-branch-register.
3972 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3973 (GOT_memory_operand): Likewise.
3974 (call_insn_operand): Likewise.
3975 (sibcall_insn_operand): Likewise.
3976 (GOT32_symbol_operand): Likewise.
3977 * config/i386/i386.md (indirect_jump): Call convert_memory_address
3978 for -mindirect-branch-register.
3979 (tablejump): Likewise.
3980 (*sibcall_memory): Likewise.
3981 (*sibcall_value_memory): Likewise.
3982 Disallow peepholes of indirect call and jump via memory for
3983 -mindirect-branch-register.
3984 (*call_pop): Replace m with Bw.
3985 (*call_value_pop): Likewise.
3986 (*sibcall_pop_memory): Replace m with Bs.
3987 * config/i386/i386.opt (mindirect-branch-register): New option.
3988 * doc/invoke.texi: Document -mindirect-branch-register option.
3990 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3992 * config/i386/i386-protos.h (ix86_output_function_return): New.
3993 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
3994 set function_return_type.
3995 (indirect_thunk_name): Add ret_p to indicate thunk for function
3997 (output_indirect_thunk_function): Pass false to
3998 indirect_thunk_name.
3999 (ix86_output_indirect_branch_via_reg): Likewise.
4000 (ix86_output_indirect_branch_via_push): Likewise.
4001 (output_indirect_thunk_function): Create alias for function
4002 return thunk if regno < 0.
4003 (ix86_output_function_return): New function.
4004 (ix86_handle_fndecl_attribute): Handle function_return.
4005 (ix86_attribute_table): Add function_return.
4006 * config/i386/i386.h (machine_function): Add
4007 function_return_type.
4008 * config/i386/i386.md (simple_return_internal): Use
4009 ix86_output_function_return.
4010 (simple_return_internal_long): Likewise.
4011 * config/i386/i386.opt (mfunction-return=): New option.
4012 (indirect_branch): Mention -mfunction-return=.
4013 * doc/extend.texi: Document function_return function attribute.
4014 * doc/invoke.texi: Document -mfunction-return= option.
4016 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4018 * config/i386/i386-opts.h (indirect_branch): New.
4019 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
4020 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
4021 with local indirect jump when converting indirect call and jump.
4022 (ix86_set_indirect_branch_type): New.
4023 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
4024 (indirectlabelno): New.
4025 (indirect_thunk_needed): Likewise.
4026 (indirect_thunk_bnd_needed): Likewise.
4027 (indirect_thunks_used): Likewise.
4028 (indirect_thunks_bnd_used): Likewise.
4029 (INDIRECT_LABEL): Likewise.
4030 (indirect_thunk_name): Likewise.
4031 (output_indirect_thunk): Likewise.
4032 (output_indirect_thunk_function): Likewise.
4033 (ix86_output_indirect_branch_via_reg): Likewise.
4034 (ix86_output_indirect_branch_via_push): Likewise.
4035 (ix86_output_indirect_branch): Likewise.
4036 (ix86_output_indirect_jmp): Likewise.
4037 (ix86_code_end): Call output_indirect_thunk_function if needed.
4038 (ix86_output_call_insn): Call ix86_output_indirect_branch if
4040 (ix86_handle_fndecl_attribute): Handle indirect_branch.
4041 (ix86_attribute_table): Add indirect_branch.
4042 * config/i386/i386.h (machine_function): Add indirect_branch_type
4043 and has_local_indirect_jump.
4044 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
4046 (tablejump): Likewise.
4047 (*indirect_jump): Use ix86_output_indirect_jmp.
4048 (*tablejump_1): Likewise.
4049 (simple_return_indirect_internal): Likewise.
4050 * config/i386/i386.opt (mindirect-branch=): New option.
4051 (indirect_branch): New.
4054 (thunk-inline): Likewise.
4055 (thunk-extern): Likewise.
4056 * doc/extend.texi: Document indirect_branch function attribute.
4057 * doc/invoke.texi: Document -mindirect-branch= option.
4059 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
4062 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
4064 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
4066 * ipa-inline.c (want_inline_small_function_p): Return false if
4067 inlining has already failed with CIF_FINAL_ERROR.
4068 (update_caller_keys): Call want_inline_small_function_p before
4070 (update_callee_keys): Likewise.
4072 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4074 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
4076 (rs6000_quadword_masked_address_p): Likewise.
4077 (quad_aligned_load_p): Likewise.
4078 (quad_aligned_store_p): Likewise.
4079 (const_load_sequence_p): Add comment to describe the outer-most loop.
4080 (mimic_memory_attributes_and_flags): New function.
4081 (rs6000_gen_stvx): Likewise.
4082 (replace_swapped_aligned_store): Likewise.
4083 (rs6000_gen_lvx): Likewise.
4084 (replace_swapped_aligned_load): Likewise.
4085 (replace_swapped_load_constant): Capitalize argument name in
4086 comment describing this function.
4087 (rs6000_analyze_swaps): Add a third pass to search for vector loads
4088 and stores that access quad-word aligned addresses and replace
4089 with stvx or lvx instructions when appropriate.
4090 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
4091 New function prototype.
4092 (rs6000_quadword_masked_address_p): Likewise.
4093 (rs6000_gen_lvx): Likewise.
4094 (rs6000_gen_stvx): Likewise.
4095 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
4096 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
4097 when memory address is aligned.
4098 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
4099 this split to select lvx instruction when memory address is aligned.
4100 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
4101 instruction when memory address is aligned.
4102 (*vsx_le_perm_load_v16qi): Likewise.
4103 (four unnamed splitters): Modify to select the stvx instruction
4104 when memory is aligned.
4106 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
4108 * predict.c (determine_unlikely_bbs): Handle correctly BBs
4109 which appears in the queue multiple times.
4111 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4112 Alan Hayward <alan.hayward@arm.com>
4113 David Sherwood <david.sherwood@arm.com>
4115 * tree-vectorizer.h (vec_lower_bound): New structure.
4116 (_loop_vec_info): Add check_nonzero and lower_bounds.
4117 (LOOP_VINFO_CHECK_NONZERO): New macro.
4118 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
4119 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
4120 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
4121 fields. Make seg_len the distance travelled, not including the
4123 (dr_direction_indicator): Declare.
4124 (dr_zero_step_indicator): Likewise.
4125 (dr_known_forward_stride_p): Likewise.
4126 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
4128 (runtime_alias_check_p): Allow runtime alias checks with
4130 (operator ==): Compare access_size and align.
4131 (prune_runtime_alias_test_list): Rework for new distinction between
4132 the access_size and seg_len.
4133 (create_intersect_range_checks_index): Likewise. Cope with polynomial
4135 (get_segment_min_max): New function.
4136 (create_intersect_range_checks): Use it.
4137 (dr_step_indicator): New function.
4138 (dr_direction_indicator): Likewise.
4139 (dr_zero_step_indicator): Likewise.
4140 (dr_known_forward_stride_p): Likewise.
4141 * tree-loop-distribution.c (data_ref_segment_size): Return
4142 DR_STEP * (niters - 1).
4143 (compute_alias_check_pairs): Update call to the dr_with_seg_len
4145 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
4146 (vect_preserves_scalar_order_p): New function, split out from...
4147 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
4148 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
4149 (vect_vfa_access_size): New function.
4150 (vect_vfa_align): Likewise.
4151 (vect_compile_time_alias): Take access_size_a and access_b arguments.
4152 (dump_lower_bound): New function.
4153 (vect_check_lower_bound): Likewise.
4154 (vect_small_gap_p): Likewise.
4155 (vectorizable_with_step_bound_p): Likewise.
4156 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
4157 depencies if the vectorization factor is 1. Convert the checks
4158 for nonzero steps into checks on the bounds of DR_STEP. Try using
4159 a bunds check for variable steps if the minimum required step is
4160 relatively small. Update calls to the dr_with_seg_len
4161 constructor and to vect_compile_time_alias.
4162 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
4164 (vect_loop_versioning): Call it.
4165 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
4167 (vect_estimate_min_profitable_iters): Account for any bounds checks.
4169 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4170 Alan Hayward <alan.hayward@arm.com>
4171 David Sherwood <david.sherwood@arm.com>
4173 * doc/sourcebuild.texi (vect_scatter_store): Document.
4174 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
4176 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
4178 * genopinit.c (main): Add supports_vec_scatter_store and
4179 supports_vec_scatter_store_cached to target_optabs.
4180 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
4181 IFN_MASK_SCATTER_STORE.
4182 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
4184 * internal-fn.h (internal_store_fn_p): Declare.
4185 (internal_fn_stored_value_index): Likewise.
4186 * internal-fn.c (scatter_store_direct): New macro.
4187 (expand_scatter_store_optab_fn): New function.
4188 (direct_scatter_store_optab_supported_p): New macro.
4189 (internal_store_fn_p): New function.
4190 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
4191 IFN_MASK_SCATTER_STORE.
4192 (internal_fn_mask_index): Likewise.
4193 (internal_fn_stored_value_index): New function.
4194 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
4196 * optabs-query.h (supports_vec_scatter_store_p): Declare.
4197 * optabs-query.c (supports_vec_scatter_store_p): New function.
4198 * tree-vectorizer.h (vect_get_store_rhs): Declare.
4199 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
4200 true for scatter stores.
4201 (vect_gather_scatter_fn_p): Handle scatter stores too.
4202 (vect_check_gather_scatter): Consider using scatter stores if
4203 supports_vec_scatter_store_p.
4204 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
4206 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4207 internal_fn_stored_value_index.
4208 (check_load_store_masking): Handle scatter stores too.
4209 (vect_get_store_rhs): Make public.
4210 (vectorizable_call): Use internal_store_fn_p.
4211 (vectorizable_store): Handle scatter store internal functions.
4212 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
4213 when deciding whether the end of the group has been reached.
4214 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
4215 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
4216 (mask_scatter_store<mode>): New insns.
4218 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4219 Alan Hayward <alan.hayward@arm.com>
4220 David Sherwood <david.sherwood@arm.com>
4222 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
4223 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
4224 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
4226 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
4227 Use vect_truncate_gather_scatter_offset if we can't treat the
4228 operation as a normal gather load or scatter store.
4229 (get_group_load_store_type): Take the gather_scatter_info
4230 as argument. Try using a gather load or scatter store for
4231 single-element groups.
4232 (get_load_store_type): Update calls to get_group_load_store_type
4233 and vect_use_strided_gather_scatters_p.
4235 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4236 Alan Hayward <alan.hayward@arm.com>
4237 David Sherwood <david.sherwood@arm.com>
4239 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
4240 optional tree argument.
4241 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
4243 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
4244 but continue to use the current value as a fallback.
4245 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
4246 to compare the updates.
4247 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
4248 (get_load_store_type): Use it when handling a strided access.
4249 (vect_get_strided_load_store_ops): New function.
4250 (vect_get_data_ptr_increment): Likewise.
4251 (vectorizable_load): Handle strided gather loads. Always pass
4252 a step to vect_create_data_ref_ptr and bump_vector_ptr.
4254 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4255 Alan Hayward <alan.hayward@arm.com>
4256 David Sherwood <david.sherwood@arm.com>
4258 * doc/md.texi (gather_load@var{m}): Document.
4259 (mask_gather_load@var{m}): Likewise.
4260 * genopinit.c (main): Add supports_vec_gather_load and
4261 supports_vec_gather_load_cached to target_optabs.
4262 * optabs-tree.c (init_tree_optimization_optabs): Use
4263 ggc_cleared_alloc to allocate target_optabs.
4264 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
4265 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
4267 * internal-fn.h (internal_load_fn_p): Declare.
4268 (internal_gather_scatter_fn_p): Likewise.
4269 (internal_fn_mask_index): Likewise.
4270 (internal_gather_scatter_fn_supported_p): Likewise.
4271 * internal-fn.c (gather_load_direct): New macro.
4272 (expand_gather_load_optab_fn): New function.
4273 (direct_gather_load_optab_supported_p): New macro.
4274 (direct_internal_fn_optab): New function.
4275 (internal_load_fn_p): Likewise.
4276 (internal_gather_scatter_fn_p): Likewise.
4277 (internal_fn_mask_index): Likewise.
4278 (internal_gather_scatter_fn_supported_p): Likewise.
4279 * optabs-query.c (supports_at_least_one_mode_p): New function.
4280 (supports_vec_gather_load_p): Likewise.
4281 * optabs-query.h (supports_vec_gather_load_p): Declare.
4282 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
4283 and memory_type field.
4284 (NUM_PATTERNS): Bump to 15.
4285 * tree-vect-data-refs.c: Include internal-fn.h.
4286 (vect_gather_scatter_fn_p): New function.
4287 (vect_describe_gather_scatter_call): Likewise.
4288 (vect_check_gather_scatter): Try using internal functions for
4289 gather loads. Recognize existing calls to a gather load function.
4290 (vect_analyze_data_refs): Consider using gather loads if
4291 supports_vec_gather_load_p.
4292 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
4293 (vect_get_gather_scatter_offset_type): Likewise.
4294 (vect_convert_mask_for_vectype): Likewise.
4295 (vect_add_conversion_to_patterm): Likewise.
4296 (vect_try_gather_scatter_pattern): Likewise.
4297 (vect_recog_gather_scatter_pattern): New pattern recognizer.
4298 (vect_vect_recog_func_ptrs): Add it.
4299 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4300 internal_fn_mask_index and internal_gather_scatter_fn_p.
4301 (check_load_store_masking): Take the gather_scatter_info as an
4302 argument and handle gather loads.
4303 (vect_get_gather_scatter_ops): New function.
4304 (vectorizable_call): Check internal_load_fn_p.
4305 (vectorizable_load): Likewise. Handle gather load internal
4307 (vectorizable_store): Update call to check_load_store_masking.
4308 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
4309 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
4310 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
4311 (aarch64_gather_scale_operand_d): New predicates.
4312 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
4313 (mask_gather_load<mode>): New insns.
4315 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4316 Alan Hayward <alan.hayward@arm.com>
4317 David Sherwood <david.sherwood@arm.com>
4319 * optabs.def (fold_left_plus_optab): New optab.
4320 * doc/md.texi (fold_left_plus_@var{m}): Document.
4321 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
4322 * internal-fn.c (fold_left_direct): Define.
4323 (expand_fold_left_optab_fn): Likewise.
4324 (direct_fold_left_optab_supported_p): Likewise.
4325 * fold-const-call.c (fold_const_fold_left): New function.
4326 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
4327 * tree-parloops.c (valid_reduction_p): New function.
4328 (gather_scalar_reductions): Use it.
4329 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
4330 (vect_finish_replace_stmt): Declare.
4331 * tree-vect-loop.c (fold_left_reduction_fn): New function.
4332 (needs_fold_left_reduction_p): New function, split out from...
4333 (vect_is_simple_reduction): ...here. Accept reductions that
4334 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
4335 (vect_force_simple_reduction): Also store the reduction type in
4336 the assignment's STMT_VINFO_REDUC_TYPE.
4337 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
4338 (merge_with_identity): New function.
4339 (vect_expand_fold_left): Likewise.
4340 (vectorize_fold_left_reduction): Likewise.
4341 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
4342 scalar phi in place for it. Check for target support and reject
4343 cases that would reassociate the operation. Defer the transform
4344 phase to vectorize_fold_left_reduction.
4345 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
4346 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
4347 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
4349 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4351 * tree-if-conv.c (predicate_mem_writes): Remove redundant
4352 call to ifc_temp_var.
4354 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4355 Alan Hayward <alan.hayward@arm.com>
4356 David Sherwood <david.sherwood@arm.com>
4358 * target.def (legitimize_address_displacement): Take the original
4359 offset as a poly_int.
4360 * targhooks.h (default_legitimize_address_displacement): Update
4362 * targhooks.c (default_legitimize_address_displacement): Likewise.
4363 * doc/tm.texi: Regenerate.
4364 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
4365 as an argument, moving assert of ad->disp == ad->disp_term to...
4366 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
4367 Try calling targetm.legitimize_address_displacement before expanding
4368 the address rather than afterwards, and adjust for the new interface.
4369 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
4370 Match the new hook interface. Handle SVE addresses.
4371 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
4374 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4376 * Makefile.in (OBJS): Add early-remat.o.
4377 * target.def (select_early_remat_modes): New hook.
4378 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
4379 * doc/tm.texi: Regenerate.
4380 * targhooks.h (default_select_early_remat_modes): Declare.
4381 * targhooks.c (default_select_early_remat_modes): New function.
4382 * timevar.def (TV_EARLY_REMAT): New timevar.
4383 * passes.def (pass_early_remat): New pass.
4384 * tree-pass.h (make_pass_early_remat): Declare.
4385 * early-remat.c: New file.
4386 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
4388 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
4390 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4391 Alan Hayward <alan.hayward@arm.com>
4392 David Sherwood <david.sherwood@arm.com>
4394 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
4395 vfm1 with a bound_epilog parameter.
4396 (vect_do_peeling): Update calls accordingly, and move the prologue
4397 call earlier in the function. Treat the base bound_epilog as 0 for
4398 fully-masked loops and retain vf - 1 for other loops. Add 1 to
4399 this base when peeling for gaps.
4400 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
4401 with fully-masked loops.
4402 (vect_estimate_min_profitable_iters): Handle the single peeled
4403 iteration in that case.
4405 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4406 Alan Hayward <alan.hayward@arm.com>
4407 David Sherwood <david.sherwood@arm.com>
4409 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
4410 single-element interleaving even if the size is not a power of 2.
4411 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
4412 accesses for single-element interleaving if the group size is
4415 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4416 Alan Hayward <alan.hayward@arm.com>
4417 David Sherwood <david.sherwood@arm.com>
4419 * doc/md.texi (fold_extract_last_@var{m}): Document.
4420 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
4421 * optabs.def (fold_extract_last_optab): New optab.
4422 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
4423 * internal-fn.c (fold_extract_direct): New macro.
4424 (expand_fold_extract_optab_fn): Likewise.
4425 (direct_fold_extract_optab_supported_p): Likewise.
4426 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
4427 * tree-vect-loop.c (vect_model_reduction_cost): Handle
4428 EXTRACT_LAST_REDUCTION.
4429 (get_initial_def_for_reduction): Do not create an initial vector
4430 for EXTRACT_LAST_REDUCTION reductions.
4431 (vectorizable_reduction): Leave the scalar phi in place for
4432 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
4433 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
4434 epilogue code for EXTRACT_LAST_REDUCTION and defer the
4435 transform phase to vectorizable_condition.
4436 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
4438 (vect_finish_stmt_generation): ...here.
4439 (vect_finish_replace_stmt): New function.
4440 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
4441 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
4443 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
4445 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4446 Alan Hayward <alan.hayward@arm.com>
4447 David Sherwood <david.sherwood@arm.com>
4449 * doc/md.texi (extract_last_@var{m}): Document.
4450 * optabs.def (extract_last_optab): New optab.
4451 * internal-fn.def (EXTRACT_LAST): New internal function.
4452 * internal-fn.c (cond_unary_direct): New macro.
4453 (expand_cond_unary_optab_fn): Likewise.
4454 (direct_cond_unary_optab_supported_p): Likewise.
4455 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
4456 loops using EXTRACT_LAST.
4457 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
4458 (extract_last_<mode>): ...this optab.
4459 (vec_extract<mode><Vel>): Update accordingly.
4461 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4462 Alan Hayward <alan.hayward@arm.com>
4463 David Sherwood <david.sherwood@arm.com>
4465 * target.def (empty_mask_is_expensive): New hook.
4466 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
4467 * doc/tm.texi: Regenerate.
4468 * targhooks.h (default_empty_mask_is_expensive): Declare.
4469 * targhooks.c (default_empty_mask_is_expensive): New function.
4470 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
4471 if the target says that empty masks are expensive.
4472 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
4474 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
4476 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4477 Alan Hayward <alan.hayward@arm.com>
4478 David Sherwood <david.sherwood@arm.com>
4480 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
4481 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
4482 (vect_use_loop_mask_for_alignment_p): New function.
4483 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
4484 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
4485 niters_skip argument. Make sure that the first niters_skip elements
4486 of the first iteration are inactive.
4487 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
4488 Update call to vect_set_loop_masks_directly.
4489 (get_misalign_in_elems): New function, split out from...
4490 (vect_gen_prolog_loop_niters): ...here.
4491 (vect_update_init_of_dr): Take a code argument that specifies whether
4492 the adjustment should be added or subtracted.
4493 (vect_update_init_of_drs): Likewise.
4494 (vect_prepare_for_masked_peels): New function.
4495 (vect_do_peeling): Skip prologue peeling if we're using a mask
4496 instead. Update call to vect_update_inits_of_drs.
4497 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4499 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
4500 alignment. Do not include the number of peeled iterations in
4501 the minimum threshold in that case.
4502 (vectorizable_induction): Adjust the start value down by
4503 LOOP_VINFO_MASK_SKIP_NITERS iterations.
4504 (vect_transform_loop): Call vect_prepare_for_masked_peels.
4505 Take the number of skipped iterations into account when calculating
4507 * tree-vect-stmts.c (vect_gen_while_not): New function.
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 * doc/sourcebuild.texi (vect_fully_masked): Document.
4514 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
4516 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
4518 (vect_analyze_loop_2): ...here. Don't check the vectorization
4519 factor against the number of loop iterations if the loop is
4522 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4523 Alan Hayward <alan.hayward@arm.com>
4524 David Sherwood <david.sherwood@arm.com>
4526 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
4527 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
4528 (dump_groups): Update accordingly.
4529 (iv_use::mem_type): New member variable.
4530 (address_p): New function.
4531 (record_use): Add a mem_type argument and initialize the new
4533 (record_group_use): Add a mem_type argument. Use address_p.
4534 Remove obsolete null checks of base_object. Update call to record_use.
4535 (find_interesting_uses_op): Update call to record_group_use.
4536 (find_interesting_uses_cond): Likewise.
4537 (find_interesting_uses_address): Likewise.
4538 (get_mem_type_for_internal_fn): New function.
4539 (find_address_like_use): Likewise.
4540 (find_interesting_uses_stmt): Try find_address_like_use before
4541 calling find_interesting_uses_op.
4542 (addr_offset_valid_p): Use the iv mem_type field as the type
4543 of the addressed memory.
4544 (add_autoinc_candidates): Likewise.
4545 (get_address_cost): Likewise.
4546 (split_small_address_groups_p): Use address_p.
4547 (split_address_groups): Likewise.
4548 (add_iv_candidate_for_use): Likewise.
4549 (autoinc_possible_for_pair): Likewise.
4550 (rewrite_groups): Likewise.
4551 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
4552 (determine_group_iv_cost): Update after split of USE_ADDRESS.
4553 (get_alias_ptr_type_for_ptr_address): New function.
4554 (rewrite_use_address): Rewrite address uses in calls that were
4555 identified by find_address_like_use.
4557 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4558 Alan Hayward <alan.hayward@arm.com>
4559 David Sherwood <david.sherwood@arm.com>
4561 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
4563 * gimple-expr.h (is_gimple_addressable: Likewise.
4564 * gimple-expr.c (is_gimple_address): Likewise.
4565 * internal-fn.c (expand_call_mem_ref): New function.
4566 (expand_mask_load_optab_fn): Use it.
4567 (expand_mask_store_optab_fn): Likewise.
4569 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4570 Alan Hayward <alan.hayward@arm.com>
4571 David Sherwood <david.sherwood@arm.com>
4573 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
4574 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
4575 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
4576 (cond_umax@var{mode}): Document.
4577 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
4578 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
4579 (cond_umin_optab, cond_umax_optab): New optabs.
4580 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
4581 (COND_IOR, COND_XOR): New internal functions.
4582 * internal-fn.h (get_conditional_internal_fn): Declare.
4583 * internal-fn.c (cond_binary_direct): New macro.
4584 (expand_cond_binary_optab_fn): Likewise.
4585 (direct_cond_binary_optab_supported_p): Likewise.
4586 (get_conditional_internal_fn): New function.
4587 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
4588 Cope with reduction statements that are vectorized as calls rather
4590 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
4591 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
4592 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
4593 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4594 (UNSPEC_COND_EOR): New unspecs.
4595 (optab): Add mappings for them.
4596 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
4597 (sve_int_op, sve_fp_op): New int attributes.
4599 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4600 Alan Hayward <alan.hayward@arm.com>
4601 David Sherwood <david.sherwood@arm.com>
4603 * optabs.def (while_ult_optab): New optab.
4604 * doc/md.texi (while_ult@var{m}@var{n}): Document.
4605 * internal-fn.def (WHILE_ULT): New internal function.
4606 * internal-fn.h (direct_internal_fn_supported_p): New override
4607 that takes two types as argument.
4608 * internal-fn.c (while_direct): New macro.
4609 (expand_while_optab_fn): New function.
4610 (convert_optab_supported_p): Likewise.
4611 (direct_while_optab_supported_p): New macro.
4612 * wide-int.h (wi::udiv_ceil): New function.
4613 * tree-vectorizer.h (rgroup_masks): New structure.
4614 (vec_loop_masks): New typedef.
4615 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
4617 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
4618 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
4619 (vect_max_vf): New function.
4620 (slpeel_make_loop_iterate_ntimes): Delete.
4621 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
4622 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
4623 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
4624 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
4625 internal-fn.h, stor-layout.h and optabs-query.h.
4626 (vect_set_loop_mask): New function.
4627 (add_preheader_seq): Likewise.
4628 (add_header_seq): Likewise.
4629 (interleave_supported_p): Likewise.
4630 (vect_maybe_permute_loop_masks): Likewise.
4631 (vect_set_loop_masks_directly): Likewise.
4632 (vect_set_loop_condition_masked): Likewise.
4633 (vect_set_loop_condition_unmasked): New function, split out from
4634 slpeel_make_loop_iterate_ntimes.
4635 (slpeel_make_loop_iterate_ntimes): Rename to..
4636 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
4637 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
4638 (vect_do_peeling): Update call accordingly.
4639 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
4641 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4642 mask_compare_type, can_fully_mask_p and fully_masked_p.
4643 (release_vec_loop_masks): New function.
4644 (_loop_vec_info): Use it to free the loop masks.
4645 (can_produce_all_loop_masks_p): New function.
4646 (vect_get_max_nscalars_per_iter): Likewise.
4647 (vect_verify_full_masking): Likewise.
4648 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
4649 retries, and free the mask rgroups before retrying. Check loop-wide
4650 reasons for disallowing fully-masked loops. Make the final decision
4651 about whether use a fully-masked loop or not.
4652 (vect_estimate_min_profitable_iters): Do not assume that peeling
4653 for the number of iterations will be needed for fully-masked loops.
4654 (vectorizable_reduction): Disable fully-masked loops.
4655 (vectorizable_live_operation): Likewise.
4656 (vect_halve_mask_nunits): New function.
4657 (vect_double_mask_nunits): Likewise.
4658 (vect_record_loop_mask): Likewise.
4659 (vect_get_loop_mask): Likewise.
4660 (vect_transform_loop): Handle the case in which the final loop
4661 iteration might handle a partial vector. Call vect_set_loop_condition
4662 instead of slpeel_make_loop_iterate_ntimes.
4663 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
4664 (check_load_store_masking): New function.
4665 (prepare_load_store_mask): Likewise.
4666 (vectorizable_store): Handle fully-masked loops.
4667 (vectorizable_load): Likewise.
4668 (supportable_widening_operation): Use vect_halve_mask_nunits for
4670 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
4671 (vect_gen_while): New function.
4672 * config/aarch64/aarch64.md (umax<mode>3): New expander.
4673 (aarch64_uqdec<mode>): New insn.
4675 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4676 Alan Hayward <alan.hayward@arm.com>
4677 David Sherwood <david.sherwood@arm.com>
4679 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
4680 (reduc_xor_scal_optab): New optabs.
4681 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
4682 (reduc_xor_scal_@var{m}): Document.
4683 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
4684 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
4686 * fold-const-call.c (fold_const_call): Handle them.
4687 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
4688 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
4689 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
4690 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
4691 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
4692 (UNSPEC_XORV): New unspecs.
4693 (optab): Add entries for them.
4694 (BITWISEV): New int iterator.
4695 (bit_reduc_op): New int attributes.
4697 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4698 Alan Hayward <alan.hayward@arm.com>
4699 David Sherwood <david.sherwood@arm.com>
4701 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
4702 * internal-fn.def (VEC_SHL_INSERT): New internal function.
4703 * optabs.def (vec_shl_insert_optab): New optab.
4704 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
4705 (duplicate_and_interleave): Likewise.
4706 * tree-vect-loop.c: Include internal-fn.h.
4707 (neutral_op_for_slp_reduction): New function, split out from
4708 get_initial_defs_for_reduction.
4709 (get_initial_def_for_reduction): Handle option 2 for variable-length
4710 vectors by loading the neutral value into a vector and then shifting
4711 the initial value into element 0.
4712 (get_initial_defs_for_reduction): Replace the code argument with
4713 the neutral value calculated by neutral_op_for_slp_reduction.
4714 Use gimple_build_vector for constant-length vectors.
4715 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
4716 but the first group_size elements have a neutral value.
4717 Use duplicate_and_interleave otherwise.
4718 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
4719 Update call to get_initial_defs_for_reduction. Handle SLP
4720 reductions for variable-length vectors by creating one vector
4721 result for each scalar result, with the elements associated
4722 with other scalar results stubbed out with the neutral value.
4723 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
4724 Require IFN_VEC_SHL_INSERT for double reductions on
4725 variable-length vectors, or SLP reductions that have
4726 a neutral value. Require can_duplicate_and_interleave_p
4727 support for variable-length unchained SLP reductions if there
4728 is no neutral value, such as for MIN/MAX reductions. Also require
4729 the number of vector elements to be a multiple of the number of
4730 SLP statements when doing variable-length unchained SLP reductions.
4731 Update call to vect_create_epilog_for_reduction.
4732 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
4733 and remove initial values.
4734 (duplicate_and_interleave): Make public.
4735 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
4736 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
4738 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4739 Alan Hayward <alan.hayward@arm.com>
4740 David Sherwood <david.sherwood@arm.com>
4742 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
4743 (can_duplicate_and_interleave_p): New function.
4744 (vect_get_and_check_slp_defs): Take the vector of statements
4745 rather than just the current one. Remove excess parentheses.
4746 Restriction rejectinon of vect_constant_def and vect_external_def
4747 for variable-length vectors to boolean types, or types for which
4748 can_duplicate_and_interleave_p is false.
4749 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
4750 (duplicate_and_interleave): New function.
4751 (vect_get_constant_vectors): Use gimple_build_vector for
4752 constant-length vectors and suitable variable-length constant
4753 vectors. Use duplicate_and_interleave for other variable-length
4754 vectors. Don't defer the update when inserting new statements.
4756 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4757 Alan Hayward <alan.hayward@arm.com>
4758 David Sherwood <david.sherwood@arm.com>
4760 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
4761 min_profitable_iters doesn't go negative.
4763 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4764 Alan Hayward <alan.hayward@arm.com>
4765 David Sherwood <david.sherwood@arm.com>
4767 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
4768 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
4769 * optabs.def (vec_mask_load_lanes_optab): New optab.
4770 (vec_mask_store_lanes_optab): Likewise.
4771 * internal-fn.def (MASK_LOAD_LANES): New internal function.
4772 (MASK_STORE_LANES): Likewise.
4773 * internal-fn.c (mask_load_lanes_direct): New macro.
4774 (mask_store_lanes_direct): Likewise.
4775 (expand_mask_load_optab_fn): Handle masked operations.
4776 (expand_mask_load_lanes_optab_fn): New macro.
4777 (expand_mask_store_optab_fn): Handle masked operations.
4778 (expand_mask_store_lanes_optab_fn): New macro.
4779 (direct_mask_load_lanes_optab_supported_p): Likewise.
4780 (direct_mask_store_lanes_optab_supported_p): Likewise.
4781 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
4783 (vect_load_lanes_supported): Likewise.
4784 * tree-vect-data-refs.c (strip_conversion): New function.
4785 (can_group_stmts_p): Likewise.
4786 (vect_analyze_data_ref_accesses): Use it instead of checking
4787 for a pair of assignments.
4788 (vect_store_lanes_supported): Take a masked_p parameter.
4789 (vect_load_lanes_supported): Likewise.
4790 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
4791 vect_store_lanes_supported and vect_load_lanes_supported.
4792 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4793 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
4794 parameter. Don't allow gaps for masked accesses.
4795 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
4796 and vect_load_lanes_supported.
4797 (get_load_store_type): Take a masked_p parameter and update
4798 call to get_group_load_store_type.
4799 (vectorizable_store): Update call to get_load_store_type.
4800 Handle IFN_MASK_STORE_LANES.
4801 (vectorizable_load): Update call to get_load_store_type.
4802 Handle IFN_MASK_LOAD_LANES.
4804 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4805 Alan Hayward <alan.hayward@arm.com>
4806 David Sherwood <david.sherwood@arm.com>
4808 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
4810 * config/aarch64/aarch64-protos.h
4811 (aarch64_sve_struct_memory_operand_p): Declare.
4812 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
4813 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
4814 (VPRED, vpred): Handle SVE structure modes.
4815 * config/aarch64/constraints.md (Utx): New constraint.
4816 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
4817 (aarch64_sve_struct_nonimmediate_operand): New predicates.
4818 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
4819 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
4820 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
4821 structure modes. Split into pieces after RA.
4822 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
4823 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
4825 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
4826 SVE structure modes.
4827 (aarch64_classify_address): Likewise.
4828 (sizetochar): Move earlier in file.
4829 (aarch64_print_operand): Handle SVE register lists.
4830 (aarch64_array_mode): New function.
4831 (aarch64_sve_struct_memory_operand_p): Likewise.
4832 (TARGET_ARRAY_MODE): Redefine.
4834 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4835 Alan Hayward <alan.hayward@arm.com>
4836 David Sherwood <david.sherwood@arm.com>
4838 * target.def (array_mode): New target hook.
4839 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
4840 * doc/tm.texi: Regenerate.
4841 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
4842 * hooks.c (hook_optmode_mode_uhwi_none): New function.
4843 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
4845 * stor-layout.c (mode_for_array): Likewise. Support polynomial
4848 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4849 Alan Hayward <alan.hayward@arm.com>
4850 David Sherwood <david.sherwood@arm.com>
4852 * fold-const.c (fold_binary_loc): Check the argument types
4853 rather than the result type when testing for a vector operation.
4855 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4857 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
4858 * doc/tm.texi: Regenerate.
4860 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4861 Alan Hayward <alan.hayward@arm.com>
4862 David Sherwood <david.sherwood@arm.com>
4864 * doc/invoke.texi (-msve-vector-bits=): Document new option.
4865 (sve): Document new AArch64 extension.
4866 * doc/md.texi (w): Extend the description of the AArch64
4867 constraint to include SVE vectors.
4868 (Upl, Upa): Document new AArch64 predicate constraints.
4869 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
4871 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
4872 (msve-vector-bits=): New option.
4873 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
4874 SVE when these are disabled.
4875 (sve): New extension.
4876 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
4877 modes. Adjust their number of units based on aarch64_sve_vg.
4878 (MAX_BITSIZE_MODE_ANY_MODE): Define.
4879 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
4880 aarch64_addr_query_type.
4881 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
4882 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
4883 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
4884 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
4885 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
4886 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
4887 (aarch64_simd_imm_zero_p): Delete.
4888 (aarch64_check_zero_based_sve_index_immediate): Declare.
4889 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4890 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4891 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4892 (aarch64_sve_float_mul_immediate_p): Likewise.
4893 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4895 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
4896 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
4897 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
4898 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
4899 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
4900 (aarch64_regmode_natural_size): Likewise.
4901 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
4902 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
4904 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
4905 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
4906 for VG and the SVE predicate registers.
4907 (V_ALIASES): Add a "z"-prefixed alias.
4908 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
4909 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
4910 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
4911 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
4912 (REG_CLASS_NAMES): Add entries for them.
4913 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
4914 and the predicate registers.
4915 (aarch64_sve_vg): Declare.
4916 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
4917 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
4918 (REGMODE_NATURAL_SIZE): Define.
4919 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
4921 * config/aarch64/aarch64.c: Include cfgrtl.h.
4922 (simd_immediate_info): Add a constructor for series vectors,
4923 and an associated step field.
4924 (aarch64_sve_vg): New variable.
4925 (aarch64_dbx_register_number): Handle VG and the predicate registers.
4926 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
4927 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
4928 (VEC_ANY_DATA, VEC_STRUCT): New constants.
4929 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
4930 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
4931 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
4932 (aarch64_get_mask_mode): New functions.
4933 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
4934 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
4935 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
4936 predicate modes and predicate registers. Explicitly restrict
4937 GPRs to modes of 16 bytes or smaller. Only allow FP registers
4938 to store a vector mode if it is recognized by
4939 aarch64_classify_vector_mode.
4940 (aarch64_regmode_natural_size): New function.
4941 (aarch64_hard_regno_caller_save_mode): Return the original mode
4943 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
4944 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
4945 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
4946 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
4948 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
4949 does not overlap dest if the function is frame-related. Handle
4951 (aarch64_split_add_offset): New function.
4952 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
4953 them aarch64_add_offset.
4954 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
4955 and update call to aarch64_sub_sp.
4956 (aarch64_add_cfa_expression): New function.
4957 (aarch64_expand_prologue): Pass extra temporary registers to the
4958 functions above. Handle the case in which we need to emit new
4959 DW_CFA_expressions for registers that were originally saved
4960 relative to the stack pointer, but now have to be expressed
4961 relative to the frame pointer.
4962 (aarch64_output_mi_thunk): Pass extra temporary registers to the
4964 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
4965 IP0 and IP1 values for SVE frames.
4966 (aarch64_expand_vec_series): New function.
4967 (aarch64_expand_sve_widened_duplicate): Likewise.
4968 (aarch64_expand_sve_const_vector): Likewise.
4969 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
4970 Handle SVE constants. Use emit_move_insn to move a force_const_mem
4971 into the register, rather than emitting a SET directly.
4972 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
4973 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
4974 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
4975 (offset_9bit_signed_scaled_p): New functions.
4976 (aarch64_replicate_bitmask_imm): New function.
4977 (aarch64_bitmask_imm): Use it.
4978 (aarch64_cannot_force_const_mem): Reject expressions involving
4979 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
4980 (aarch64_classify_index): Handle SVE indices, by requiring
4981 a plain register index with a scale that matches the element size.
4982 (aarch64_classify_address): Handle SVE addresses. Assert that
4983 the mode of the address is VOIDmode or an integer mode.
4984 Update call to aarch64_classify_symbol.
4985 (aarch64_classify_symbolic_expression): Update call to
4986 aarch64_classify_symbol.
4987 (aarch64_const_vec_all_in_range_p): New function.
4988 (aarch64_print_vector_float_operand): Likewise.
4989 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
4990 "vN" for FP registers with SVE modes. Handle (const ...) vectors
4991 and the FP immediates 1.0 and 0.5.
4992 (aarch64_print_address_internal): Handle SVE addresses.
4993 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
4994 (aarch64_regno_regclass): Handle predicate registers.
4995 (aarch64_secondary_reload): Handle big-endian reloads of SVE
4997 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
4998 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
4999 (aarch64_convert_sve_vector_bits): New function.
5000 (aarch64_override_options): Use it to handle -msve-vector-bits=.
5001 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5003 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
5004 Handle SVE vector and predicate modes. Accept VL-based constants
5005 that need only one temporary register, and VL offsets that require
5006 no temporary registers.
5007 (aarch64_conditional_register_usage): Mark the predicate registers
5008 as fixed if SVE isn't available.
5009 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
5010 Return true for SVE vector and predicate modes.
5011 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
5012 rather than an unsigned int. Handle SVE modes.
5013 (aarch64_preferred_simd_mode): Update call accordingly. Handle
5015 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
5017 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5018 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5019 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5020 (aarch64_sve_float_mul_immediate_p): New functions.
5021 (aarch64_sve_valid_immediate): New function.
5022 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
5023 Explicitly reject structure modes. Check for INDEX constants.
5024 Handle PTRUE and PFALSE constants.
5025 (aarch64_check_zero_based_sve_index_immediate): New function.
5026 (aarch64_simd_imm_zero_p): Delete.
5027 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
5028 vector modes. Accept constants in the range of CNT[BHWD].
5029 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
5030 ask for an Advanced SIMD mode.
5031 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
5032 (aarch64_simd_vector_alignment): Handle SVE predicates.
5033 (aarch64_vectorize_preferred_vector_alignment): New function.
5034 (aarch64_simd_vector_alignment_reachable): Use it instead of
5036 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
5037 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
5039 (MAX_VECT_LEN): Delete.
5040 (expand_vec_perm_d): Add a vec_flags field.
5041 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
5042 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
5043 (aarch64_evpc_ext): Don't apply a big-endian lane correction
5045 (aarch64_evpc_rev): Rename to...
5046 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
5047 (aarch64_evpc_rev_global): New function.
5048 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
5049 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
5051 (aarch64_evpc_sve_tbl): New function.
5052 (aarch64_expand_vec_perm_const_1): Update after rename of
5053 aarch64_evpc_rev. Handle SVE permutes too, trying
5054 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
5055 than aarch64_evpc_tbl.
5056 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
5057 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
5058 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
5059 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
5060 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
5061 (aarch64_expand_sve_vcond): New functions.
5062 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
5063 of aarch64_vector_mode_p.
5064 (aarch64_dwarf_poly_indeterminate_value): New function.
5065 (aarch64_compute_pressure_classes): Likewise.
5066 (aarch64_can_change_mode_class): Likewise.
5067 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
5068 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
5069 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
5070 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
5071 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
5072 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
5073 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
5074 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
5076 (Dn, Dl, Dr): Accept const as well as const_vector.
5077 (Dz): Likewise. Compare against CONST0_RTX.
5078 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
5079 of "vector" where appropriate.
5080 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
5081 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
5082 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
5083 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
5084 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
5085 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
5086 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
5087 (v_int_equiv): Extend to SVE modes.
5088 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
5090 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
5091 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
5092 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
5093 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
5094 (SVE_COND_FP_CMP): New int iterators.
5095 (perm_hilo): Handle the new unpack unspecs.
5096 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
5098 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
5099 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
5100 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
5101 (aarch64_equality_operator, aarch64_constant_vector_operand)
5102 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
5103 (aarch64_sve_nonimmediate_operand): Likewise.
5104 (aarch64_sve_general_operand): Likewise.
5105 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
5106 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
5107 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
5108 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
5109 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
5110 (aarch64_sve_float_arith_immediate): Likewise.
5111 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
5112 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
5113 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
5114 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
5115 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
5116 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
5117 (aarch64_sve_float_arith_operand): Likewise.
5118 (aarch64_sve_float_arith_with_sub_operand): Likewise.
5119 (aarch64_sve_float_mul_operand): Likewise.
5120 (aarch64_sve_vec_perm_operand): Likewise.
5121 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
5122 (aarch64_mov_operand): Accept const_poly_int and const_vector.
5123 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
5124 as well as const_vector.
5125 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
5126 in file. Use CONST0_RTX and CONSTM1_RTX.
5127 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
5128 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
5129 Use aarch64_simd_imm_zero.
5130 * config/aarch64/aarch64-sve.md: New file.
5131 * config/aarch64/aarch64.md: Include it.
5132 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
5133 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
5134 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
5135 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
5136 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
5137 (sve): New attribute.
5138 (enabled): Disable instructions with the sve attribute unless
5140 (movqi, movhi): Pass CONST_POLY_INT operaneds through
5141 aarch64_expand_mov_immediate.
5142 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
5143 CNT[BHSD] immediates.
5144 (movti): Split CONST_POLY_INT moves into two halves.
5145 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
5146 Split additions that need a temporary here if the destination
5147 is the stack pointer.
5148 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
5149 (*add<mode>3_poly_1): New instruction.
5150 (set_clobber_cc): New expander.
5152 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5154 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
5155 parameter and use it instead of GET_MODE_SIZE (innermode). Use
5156 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
5157 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
5158 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
5159 Change innermode from fixed_mode_size to machine_mode.
5160 (simplify_subreg): Update call accordingly. Handle a constant-sized
5161 subreg of a variable-length CONST_VECTOR.
5163 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5164 Alan Hayward <alan.hayward@arm.com>
5165 David Sherwood <david.sherwood@arm.com>
5167 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
5168 (add_offset_to_base): New function, split out from...
5169 (create_mem_ref): ...here. When handling a scale other than 1,
5170 check first whether the address is valid without the offset.
5171 Add it into the base if so, leaving the index and scale as-is.
5173 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5176 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
5177 fold_for_warn before checking if arg2 is INTEGER_CST.
5179 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
5181 * config/rs6000/predicates.md (load_multiple_operation): Delete.
5182 (store_multiple_operation): Delete.
5183 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
5184 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
5185 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
5186 guarded by TARGET_STRING.
5187 (rs6000_output_load_multiple): Delete.
5188 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
5189 OPTION_MASK_STRING / TARGET_STRING handling.
5190 (print_operand) <'N', 'O'>: Add comment that these are unused now.
5191 (const rs6000_opt_masks) <"string">: Change mask to 0.
5192 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
5193 (MASK_STRING): Delete.
5194 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
5196 (load_multiple): Delete.
5203 (store_multiple): Delete.
5210 (movmemsi_8reg): Delete.
5211 (corresponding unnamed define_insn): Delete.
5212 (movmemsi_6reg): Delete.
5213 (corresponding unnamed define_insn): Delete.
5214 (movmemsi_4reg): Delete.
5215 (corresponding unnamed define_insn): Delete.
5216 (movmemsi_2reg): Delete.
5217 (corresponding unnamed define_insn): Delete.
5218 (movmemsi_1reg): Delete.
5219 (corresponding unnamed define_insn): Delete.
5220 * config/rs6000/rs6000.opt (mno-string): New.
5221 (mstring): Replace by deprecation warning stub.
5222 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
5224 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5226 * regrename.c (regrename_do_replace): If replacing the same
5227 reg multiple times, try to reuse last created gen_raw_REG.
5230 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
5231 main to workaround a bug in GDB.
5233 2018-01-12 Tom de Vries <tom@codesourcery.com>
5236 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
5238 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
5240 PR rtl-optimization/80481
5241 * ira-color.c (get_cap_member): New function.
5242 (allocnos_conflict_by_live_ranges_p): Use it.
5243 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
5244 (setup_slot_coalesced_allocno_live_ranges): Ditto.
5246 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
5249 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
5250 (*saddl_se_1): Ditto.
5252 (*ssubl_se_1): Ditto.
5254 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5256 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
5257 rather than wi::to_widest for DR_INITs.
5258 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
5259 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
5260 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
5262 (vect_analyze_group_access_1): Note that here.
5264 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5266 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
5267 polynomial type sizes.
5269 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5271 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
5272 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
5273 (gimple_add_tmp_var): Likewise.
5275 2018-01-12 Martin Liska <mliska@suse.cz>
5277 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
5278 (gimple_alloc_sizes): Likewise.
5279 (dump_gimple_statistics): Use PRIu64 in printf format.
5280 * gimple.h: Change uint64_t to int.
5282 2018-01-12 Martin Liska <mliska@suse.cz>
5284 * tree-core.h: Use uint64_t instead of int.
5285 * tree.c (tree_node_counts): Likewise.
5286 (tree_node_sizes): Likewise.
5287 (dump_tree_statistics): Use PRIu64 in printf format.
5289 2018-01-12 Martin Liska <mliska@suse.cz>
5291 * Makefile.in: As qsort_chk is implemented in vec.c, add
5292 vec.o to linkage of gencfn-macros.
5293 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
5294 passing the info to record_node_allocation_statistics.
5295 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
5297 * ggc-common.c (struct ggc_usage): Add operator== and use
5298 it in operator< and compare function.
5299 * mem-stats.h (struct mem_usage): Likewise.
5300 * vec.c (struct vec_usage): Remove operator< and compare
5301 function. Can be simply inherited.
5303 2018-01-12 Martin Jambor <mjambor@suse.cz>
5306 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
5307 * tree-ssa-math-opts.c: Include domwalk.h.
5308 (convert_mult_to_fma_1): New function.
5309 (fma_transformation_info): New type.
5310 (fma_deferring_state): Likewise.
5311 (cancel_fma_deferring): New function.
5312 (result_of_phi): Likewise.
5313 (last_fma_candidate_feeds_initial_phi): Likewise.
5314 (convert_mult_to_fma): Added deferring logic, split actual
5315 transformation to convert_mult_to_fma_1.
5316 (math_opts_dom_walker): New type.
5317 (math_opts_dom_walker::after_dom_children): New method, body moved
5318 here from pass_optimize_widening_mul::execute, added deferring logic
5320 (pass_optimize_widening_mul::execute): Moved most of code to
5321 math_opts_dom_walker::after_dom_children.
5322 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
5323 * config/i386/i386.c (ix86_option_override_internal): Added
5324 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
5326 2018-01-12 Richard Biener <rguenther@suse.de>
5329 * dwarf2out.c (gen_variable_die): Do not reset old_die for
5330 inline instance vars.
5332 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
5335 * config/rx/rx.c (rx_is_restricted_memory_address):
5338 2018-01-12 Richard Biener <rguenther@suse.de>
5340 PR tree-optimization/80846
5341 * target.def (split_reduction): New target hook.
5342 * targhooks.c (default_split_reduction): New function.
5343 * targhooks.h (default_split_reduction): Declare.
5344 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
5345 target requests first reduce vectors by combining low and high
5347 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
5348 (get_vectype_for_scalar_type_and_size): Export.
5349 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
5350 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
5351 * doc/tm.texi: Regenerate.
5352 * config/i386/i386.c (ix86_split_reduction): Implement
5353 TARGET_VECTORIZE_SPLIT_REDUCTION.
5355 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5358 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
5359 in PIC mode except for TARGET_VXWORKS_RTP.
5360 * config/sparc/sparc.c: Include cfgrtl.h.
5361 (TARGET_INIT_PIC_REG): Define.
5362 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
5363 (sparc_pic_register_p): New predicate.
5364 (sparc_legitimate_address_p): Use it.
5365 (sparc_legitimize_pic_address): Likewise.
5366 (sparc_delegitimize_address): Likewise.
5367 (sparc_mode_dependent_address_p): Likewise.
5368 (gen_load_pcrel_sym): Remove 4th parameter.
5369 (load_got_register): Adjust call to above. Remove obsolete stuff.
5370 (sparc_expand_prologue): Do not call load_got_register here.
5371 (sparc_flat_expand_prologue): Likewise.
5372 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
5373 (sparc_use_pseudo_pic_reg): New function.
5374 (sparc_init_pic_reg): Likewise.
5375 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5376 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
5378 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
5380 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
5381 Add item for branch_cost.
5383 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5385 PR rtl-optimization/83565
5386 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
5387 not extend the result to a larger mode for rotate operations.
5388 (num_sign_bit_copies1): Likewise.
5390 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5393 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
5395 Use values-Xc.o for -pedantic.
5396 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
5398 2018-01-12 Martin Liska <mliska@suse.cz>
5401 * ipa-devirt.c (final_warning_record::grow_type_warnings):
5403 (possible_polymorphic_call_targets): Use it.
5404 (ipa_devirt): Likewise.
5406 2018-01-12 Martin Liska <mliska@suse.cz>
5408 * profile-count.h (enum profile_quality): Use 0 as invalid
5409 enum value of profile_quality.
5411 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
5413 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
5414 -mext-string options.
5416 2018-01-12 Richard Biener <rguenther@suse.de>
5418 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
5419 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
5420 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
5422 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
5424 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5426 * configure.ac (--with-long-double-format): Add support for the
5427 configuration option to change the default long double format on
5429 * config.gcc (powerpc*-linux*-*): Likewise.
5430 * configure: Regenerate.
5431 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
5432 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
5433 used without modification.
5435 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5437 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
5438 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
5439 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
5440 MISC_BUILTIN_SPEC_BARRIER.
5441 (rs6000_init_builtins): Likewise.
5442 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
5444 (speculation_barrier): New define_insn.
5445 * doc/extend.texi: Document __builtin_speculation_barrier.
5447 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5450 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
5451 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
5452 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
5454 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
5455 integral modes instead of "ss" and "sd".
5456 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
5457 vectors with 32-bit and 64-bit elements.
5458 (vecdupssescalarmodesuffix): New mode attribute.
5459 (vec_dup<mode>): Use it.
5461 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
5464 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
5465 frame if argument is passed on stack.
5467 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5470 * ree.c (combine_reaching_defs): Optimize also
5471 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
5472 reg2=any_extend(exp); reg1=reg2;, formatting fix.
5474 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5477 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
5479 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5482 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
5483 after they are computed.
5485 2018-01-11 Bin Cheng <bin.cheng@arm.com>
5487 PR tree-optimization/83695
5488 * gimple-loop-linterchange.cc
5489 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
5490 reset cached scev information after interchange.
5491 (pass_linterchange::execute): Remove call to scev_reset_htab.
5493 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5495 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
5496 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
5497 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
5498 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
5499 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
5500 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
5501 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
5502 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
5503 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
5504 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
5505 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
5506 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
5507 (V_lane_reg): Likewise.
5508 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
5510 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
5511 (vfmal_lane_low<mode>_intrinsic,
5512 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
5513 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
5514 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
5515 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
5516 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
5517 vfmsl_lane_high<mode>_intrinsic): New define_insns.
5519 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5521 * config/arm/arm-cpus.in (fp16fml): New feature.
5522 (ALL_SIMD): Add fp16fml.
5523 (armv8.2-a): Add fp16fml as an option.
5524 (armv8.3-a): Likewise.
5525 (armv8.4-a): Add fp16fml as part of fp16.
5526 * config/arm/arm.h (TARGET_FP16FML): Define.
5527 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
5529 * config/arm/arm-modes.def (V2HF): Define.
5530 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
5531 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
5532 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
5533 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
5534 vfmsl_low, vfmsl_high): New set of builtins.
5535 * config/arm/iterators.md (PLUSMINUS): New code iterator.
5536 (vfml_op): New code attribute.
5537 (VFMLHALVES): New int iterator.
5538 (VFML, VFMLSEL): New mode attributes.
5539 (V_reg): Define mapping for V2HF.
5540 (V_hi, V_lo): New mode attributes.
5541 (VF_constraint): Likewise.
5542 (vfml_half, vfml_half_selector): New int attributes.
5543 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
5545 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
5546 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
5548 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
5549 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
5550 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
5551 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
5553 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
5554 Document new effective target and option set.
5556 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5558 * config/arm/arm-cpus.in (armv8_4): New feature.
5559 (ARMv8_4a): New fgroup.
5560 (armv8.4-a): New arch.
5561 * config/arm/arm-tables.opt: Regenerate.
5562 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
5563 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
5564 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
5565 Add matching rules for -march=armv8.4-a and extensions.
5566 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
5568 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
5571 * config/rx/rx.md (BW): New mode attribute.
5572 (sync_lock_test_and_setsi): Add mode suffix to insn output.
5574 2018-01-11 Richard Biener <rguenther@suse.de>
5576 PR tree-optimization/83435
5577 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
5578 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
5579 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
5581 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5582 Alan Hayward <alan.hayward@arm.com>
5583 David Sherwood <david.sherwood@arm.com>
5585 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
5587 (aarch64_classify_address): Initialize it. Track polynomial offsets.
5588 (aarch64_print_address_internal): Use it to check for a zero offset.
5590 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5591 Alan Hayward <alan.hayward@arm.com>
5592 David Sherwood <david.sherwood@arm.com>
5594 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
5595 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
5596 Return a poly_int64 rather than a HOST_WIDE_INT.
5597 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
5598 rather than a HOST_WIDE_INT.
5599 * config/aarch64/aarch64.h (aarch64_frame): Protect with
5600 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
5601 hard_fp_offset, frame_size, initial_adjust, callee_offset and
5602 final_offset from HOST_WIDE_INT to poly_int64.
5603 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
5604 to_constant when getting the number of units in an Advanced SIMD
5606 (aarch64_builtin_vectorized_function): Check for a constant number
5608 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
5610 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
5611 attribute instead of GET_MODE_NUNITS.
5612 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
5613 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
5614 GET_MODE_SIZE for fixed-size registers.
5615 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
5616 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
5617 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
5618 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
5619 (aarch64_print_operand, aarch64_print_address_internal)
5620 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
5621 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
5622 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
5623 Handle polynomial GET_MODE_SIZE.
5624 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
5625 wider than SImode without modification.
5626 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
5627 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
5628 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
5629 passing and returning SVE modes.
5630 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
5631 rather than GEN_INT.
5632 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
5633 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
5634 (aarch64_allocate_and_probe_stack_space): Likewise.
5635 (aarch64_layout_frame): Cope with polynomial offsets.
5636 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
5637 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
5639 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
5640 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
5641 poly_int64 rather than a HOST_WIDE_INT.
5642 (aarch64_get_separate_components, aarch64_process_components)
5643 (aarch64_expand_prologue, aarch64_expand_epilogue)
5644 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
5645 (aarch64_anchor_offset): New function, split out from...
5646 (aarch64_legitimize_address): ...here.
5647 (aarch64_builtin_vectorization_cost): Handle polynomial
5648 TYPE_VECTOR_SUBPARTS.
5649 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
5651 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
5652 number of elements from the PARALLEL rather than the mode.
5653 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
5654 rather than GET_MODE_BITSIZE.
5655 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
5656 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
5657 (aarch64_expand_vec_perm_const_1): Handle polynomial
5658 d->perm.length () and d->perm elements.
5659 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
5660 Apply to_constant to d->perm elements.
5661 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
5662 polynomial CONST_VECTOR_NUNITS.
5663 (aarch64_move_pointer): Take amount as a poly_int64 rather
5665 (aarch64_progress_pointer): Avoid temporary variable.
5666 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
5667 the mode attribute instead of GET_MODE.
5669 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5670 Alan Hayward <alan.hayward@arm.com>
5671 David Sherwood <david.sherwood@arm.com>
5673 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
5674 x exists before using it.
5675 (aarch64_add_constant_internal): Rename to...
5676 (aarch64_add_offset_1): ...this. Replace regnum with separate
5677 src and dest rtxes. Handle the case in which they're different,
5678 including when the offset is zero. Replace scratchreg with an rtx.
5679 Use 2 additions if there is no spare register into which we can
5680 move a 16-bit constant.
5681 (aarch64_add_constant): Delete.
5682 (aarch64_add_offset): Replace reg with separate src and dest
5683 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
5684 Use aarch64_add_offset_1.
5685 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
5686 an rtx rather than an int. Take the delta as a poly_int64
5687 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
5688 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
5689 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
5690 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
5691 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
5693 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
5694 aarch64_add_constant.
5696 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5698 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
5699 Use scalar_float_mode.
5701 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5703 * config/aarch64/aarch64-simd.md
5704 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
5705 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
5706 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
5707 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
5708 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
5709 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
5710 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
5711 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
5712 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
5713 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
5715 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5718 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
5719 targ_options->x_arm_arch_string is non NULL.
5721 2018-01-11 Tamar Christina <tamar.christina@arm.com>
5723 * config/aarch64/aarch64.h
5724 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
5726 2018-01-11 Sudakshina Das <sudi.das@arm.com>
5729 * expmed.c (emit_store_flag_force): Swap if const op0
5730 and change VOIDmode to mode of op0.
5732 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5734 PR rtl-optimization/83761
5735 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
5736 than bytes to mode_for_size.
5738 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5741 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
5742 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
5745 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5748 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
5749 when in layout mode.
5750 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
5751 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
5754 2018-01-10 Michael Collison <michael.collison@arm.com>
5756 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
5757 * config/aarch64/aarch64-option-extension.def: Add
5758 AARCH64_OPT_EXTENSION of 'fp16fml'.
5759 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5760 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
5761 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
5762 * config/aarch64/constraints.md (Ui7): New constraint.
5763 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
5764 (VFMLA_SEL_W): Ditto.
5767 (VFMLA16_LOW): New int iterator.
5768 (VFMLA16_HIGH): Ditto.
5769 (UNSPEC_FMLAL): New unspec.
5770 (UNSPEC_FMLSL): Ditto.
5771 (UNSPEC_FMLAL2): Ditto.
5772 (UNSPEC_FMLSL2): Ditto.
5773 (f16mac): New code attribute.
5774 * config/aarch64/aarch64-simd-builtins.def
5775 (aarch64_fmlal_lowv2sf): Ditto.
5776 (aarch64_fmlsl_lowv2sf): Ditto.
5777 (aarch64_fmlalq_lowv4sf): Ditto.
5778 (aarch64_fmlslq_lowv4sf): Ditto.
5779 (aarch64_fmlal_highv2sf): Ditto.
5780 (aarch64_fmlsl_highv2sf): Ditto.
5781 (aarch64_fmlalq_highv4sf): Ditto.
5782 (aarch64_fmlslq_highv4sf): Ditto.
5783 (aarch64_fmlal_lane_lowv2sf): Ditto.
5784 (aarch64_fmlsl_lane_lowv2sf): Ditto.
5785 (aarch64_fmlal_laneq_lowv2sf): Ditto.
5786 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
5787 (aarch64_fmlalq_lane_lowv4sf): Ditto.
5788 (aarch64_fmlsl_lane_lowv4sf): Ditto.
5789 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
5790 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
5791 (aarch64_fmlal_lane_highv2sf): Ditto.
5792 (aarch64_fmlsl_lane_highv2sf): Ditto.
5793 (aarch64_fmlal_laneq_highv2sf): Ditto.
5794 (aarch64_fmlsl_laneq_highv2sf): Ditto.
5795 (aarch64_fmlalq_lane_highv4sf): Ditto.
5796 (aarch64_fmlsl_lane_highv4sf): Ditto.
5797 (aarch64_fmlalq_laneq_highv4sf): Ditto.
5798 (aarch64_fmlsl_laneq_highv4sf): Ditto.
5799 * config/aarch64/aarch64-simd.md:
5800 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
5801 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5802 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
5803 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5804 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
5805 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
5806 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
5807 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
5808 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
5809 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
5810 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
5811 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
5812 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
5813 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
5814 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
5815 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
5816 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
5817 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
5818 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
5819 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
5820 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
5821 (vfmlsl_low_u32): Ditto.
5822 (vfmlalq_low_u32): Ditto.
5823 (vfmlslq_low_u32): Ditto.
5824 (vfmlal_high_u32): Ditto.
5825 (vfmlsl_high_u32): Ditto.
5826 (vfmlalq_high_u32): Ditto.
5827 (vfmlslq_high_u32): Ditto.
5828 (vfmlal_lane_low_u32): Ditto.
5829 (vfmlsl_lane_low_u32): Ditto.
5830 (vfmlal_laneq_low_u32): Ditto.
5831 (vfmlsl_laneq_low_u32): Ditto.
5832 (vfmlalq_lane_low_u32): Ditto.
5833 (vfmlslq_lane_low_u32): Ditto.
5834 (vfmlalq_laneq_low_u32): Ditto.
5835 (vfmlslq_laneq_low_u32): Ditto.
5836 (vfmlal_lane_high_u32): Ditto.
5837 (vfmlsl_lane_high_u32): Ditto.
5838 (vfmlal_laneq_high_u32): Ditto.
5839 (vfmlsl_laneq_high_u32): Ditto.
5840 (vfmlalq_lane_high_u32): Ditto.
5841 (vfmlslq_lane_high_u32): Ditto.
5842 (vfmlalq_laneq_high_u32): Ditto.
5843 (vfmlslq_laneq_high_u32): Ditto.
5844 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
5845 (AARCH64_FL_FOR_ARCH8_4): New.
5846 (AARCH64_ISA_F16FML): New ISA flag.
5847 (TARGET_F16FML): New feature flag for fp16fml.
5848 (doc/invoke.texi): Document new fp16fml option.
5850 2018-01-10 Michael Collison <michael.collison@arm.com>
5852 * config/aarch64/aarch64-builtins.c:
5853 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
5854 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5855 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
5856 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
5857 (AARCH64_ISA_SHA3): New ISA flag.
5858 (TARGET_SHA3): New feature flag for sha3.
5859 * config/aarch64/iterators.md (sha512_op): New int attribute.
5860 (CRYPTO_SHA512): New int iterator.
5861 (UNSPEC_SHA512H): New unspec.
5862 (UNSPEC_SHA512H2): Ditto.
5863 (UNSPEC_SHA512SU0): Ditto.
5864 (UNSPEC_SHA512SU1): Ditto.
5865 * config/aarch64/aarch64-simd-builtins.def
5866 (aarch64_crypto_sha512hqv2di): New builtin.
5867 (aarch64_crypto_sha512h2qv2di): Ditto.
5868 (aarch64_crypto_sha512su0qv2di): Ditto.
5869 (aarch64_crypto_sha512su1qv2di): Ditto.
5870 (aarch64_eor3qv8hi): Ditto.
5871 (aarch64_rax1qv2di): Ditto.
5872 (aarch64_xarqv2di): Ditto.
5873 (aarch64_bcaxqv8hi): Ditto.
5874 * config/aarch64/aarch64-simd.md:
5875 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
5876 (aarch64_crypto_sha512su0qv2di): Ditto.
5877 (aarch64_crypto_sha512su1qv2di): Ditto.
5878 (aarch64_eor3qv8hi): Ditto.
5879 (aarch64_rax1qv2di): Ditto.
5880 (aarch64_xarqv2di): Ditto.
5881 (aarch64_bcaxqv8hi): Ditto.
5882 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
5883 (vsha512h2q_u64): Ditto.
5884 (vsha512su0q_u64): Ditto.
5885 (vsha512su1q_u64): Ditto.
5886 (veor3q_u16): Ditto.
5887 (vrax1q_u64): Ditto.
5889 (vbcaxq_u16): Ditto.
5890 * config/arm/types.md (crypto_sha512): New type attribute.
5891 (crypto_sha3): Ditto.
5892 (doc/invoke.texi): Document new sha3 option.
5894 2018-01-10 Michael Collison <michael.collison@arm.com>
5896 * config/aarch64/aarch64-builtins.c:
5897 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
5898 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5899 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
5900 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
5901 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
5902 (AARCH64_ISA_SM4): New ISA flag.
5903 (TARGET_SM4): New feature flag for sm4.
5904 * config/aarch64/aarch64-simd-builtins.def
5905 (aarch64_sm3ss1qv4si): Ditto.
5906 (aarch64_sm3tt1aq4si): Ditto.
5907 (aarch64_sm3tt1bq4si): Ditto.
5908 (aarch64_sm3tt2aq4si): Ditto.
5909 (aarch64_sm3tt2bq4si): Ditto.
5910 (aarch64_sm3partw1qv4si): Ditto.
5911 (aarch64_sm3partw2qv4si): Ditto.
5912 (aarch64_sm4eqv4si): Ditto.
5913 (aarch64_sm4ekeyqv4si): Ditto.
5914 * config/aarch64/aarch64-simd.md:
5915 (aarch64_sm3ss1qv4si): Ditto.
5916 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
5917 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
5918 (aarch64_sm4eqv4si): Ditto.
5919 (aarch64_sm4ekeyqv4si): Ditto.
5920 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
5921 (sm3part_op): Ditto.
5922 (CRYPTO_SM3TT): Ditto.
5923 (CRYPTO_SM3PART): Ditto.
5924 (UNSPEC_SM3SS1): New unspec.
5925 (UNSPEC_SM3TT1A): Ditto.
5926 (UNSPEC_SM3TT1B): Ditto.
5927 (UNSPEC_SM3TT2A): Ditto.
5928 (UNSPEC_SM3TT2B): Ditto.
5929 (UNSPEC_SM3PARTW1): Ditto.
5930 (UNSPEC_SM3PARTW2): Ditto.
5931 (UNSPEC_SM4E): Ditto.
5932 (UNSPEC_SM4EKEY): Ditto.
5933 * config/aarch64/constraints.md (Ui2): New constraint.
5934 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
5935 * config/arm/types.md (crypto_sm3): New type attribute.
5936 (crypto_sm4): Ditto.
5937 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
5938 (vsm3tt1aq_u32): Ditto.
5939 (vsm3tt1bq_u32): Ditto.
5940 (vsm3tt2aq_u32): Ditto.
5941 (vsm3tt2bq_u32): Ditto.
5942 (vsm3partw1q_u32): Ditto.
5943 (vsm3partw2q_u32): Ditto.
5944 (vsm4eq_u32): Ditto.
5945 (vsm4ekeyq_u32): Ditto.
5946 (doc/invoke.texi): Document new sm4 option.
5948 2018-01-10 Michael Collison <michael.collison@arm.com>
5950 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
5951 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
5952 (AARCH64_FL_FOR_ARCH8_4): New.
5953 (AARCH64_FL_V8_4): New flag.
5954 (doc/invoke.texi): Document new armv8.4-a option.
5956 2018-01-10 Michael Collison <michael.collison@arm.com>
5958 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5959 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
5960 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
5961 * config/aarch64/aarch64-option-extension.def: Add
5962 AARCH64_OPT_EXTENSION of 'sha2'.
5963 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
5964 (crypto): Disable sha2 and aes if crypto disabled.
5965 (crypto): Enable aes and sha2 if enabled.
5966 (simd): Disable sha2 and aes if simd disabled.
5967 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
5969 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
5970 (TARGET_SHA2): New feature flag for sha2.
5971 (TARGET_AES): New feature flag for aes.
5972 * config/aarch64/aarch64-simd.md:
5973 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
5974 conditional on TARGET_AES.
5975 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
5976 (aarch64_crypto_sha1hsi): Make pattern conditional
5978 (aarch64_crypto_sha1hv4si): Ditto.
5979 (aarch64_be_crypto_sha1hv4si): Ditto.
5980 (aarch64_crypto_sha1su1v4si): Ditto.
5981 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
5982 (aarch64_crypto_sha1su0v4si): Ditto.
5983 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
5984 (aarch64_crypto_sha256su0v4si): Ditto.
5985 (aarch64_crypto_sha256su1v4si): Ditto.
5986 (doc/invoke.texi): Document new aes and sha2 options.
5988 2018-01-10 Martin Sebor <msebor@redhat.com>
5990 PR tree-optimization/83781
5991 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
5994 2018-01-11 Martin Sebor <msebor@gmail.com>
5995 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5997 PR tree-optimization/83501
5998 PR tree-optimization/81703
6000 * tree-ssa-strlen.c (get_string_cst): Rename...
6001 (get_string_len): ...to this. Handle global constants.
6002 (handle_char_store): Adjust.
6004 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
6005 Jim Wilson <jimw@sifive.com>
6007 * config/riscv/riscv-protos.h (riscv_output_return): New.
6008 * config/riscv/riscv.c (struct machine_function): New naked_p field.
6009 (riscv_attribute_table, riscv_output_return),
6010 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
6011 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
6012 (riscv_compute_frame_info): Only compute frame->mask if not a naked
6014 (riscv_expand_prologue): Add early return for naked function.
6015 (riscv_expand_epilogue): Likewise.
6016 (riscv_function_ok_for_sibcall): Return false for naked function.
6017 (riscv_set_current_function): New.
6018 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
6019 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
6020 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
6021 * doc/extend.texi (RISC-V Function Attributes): New.
6023 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6025 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
6026 check for 128-bit long double before checking TCmode.
6027 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
6028 128-bit long doubles before checking TFmode or TCmode.
6029 (FLOAT128_IBM_P): Likewise.
6031 2018-01-10 Martin Sebor <msebor@redhat.com>
6033 PR tree-optimization/83671
6034 * builtins.c (c_strlen): Unconditionally return zero for the empty
6036 Use -Warray-bounds for warnings.
6037 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
6038 for non-constant array indices with COMPONENT_REF, arrays of
6039 arrays, and pointers to arrays.
6040 (gimple_fold_builtin_strlen): Determine and set length range for
6041 non-constant character arrays.
6043 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
6046 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
6049 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
6051 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
6053 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6056 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
6057 VECTOR_MEM_ALTIVEC_OR_VSX_P.
6058 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
6059 indexed_or_indirect_operand predicate.
6060 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
6061 (*vsx_le_perm_load_v8hi): Likewise.
6062 (*vsx_le_perm_load_v16qi): Likewise.
6063 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
6064 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
6065 (*vsx_le_perm_store_v8hi): Likewise.
6066 (*vsx_le_perm_store_v16qi): Likewise.
6067 (eight unnamed splitters): Likewise.
6069 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6071 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
6072 * config/rs6000/emmintrin.h: Likewise.
6073 * config/rs6000/mmintrin.h: Likewise.
6074 * config/rs6000/xmmintrin.h: Likewise.
6076 2018-01-10 David Malcolm <dmalcolm@redhat.com>
6079 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
6081 * tree.c (tree_nop_conversion): Return true for location wrapper
6083 (maybe_wrap_with_location): New function.
6084 (selftest::check_strip_nops): New function.
6085 (selftest::test_location_wrappers): New function.
6086 (selftest::tree_c_tests): Call it.
6087 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
6088 (maybe_wrap_with_location): New decl.
6089 (EXPR_LOCATION_WRAPPER_P): New macro.
6090 (location_wrapper_p): New inline function.
6091 (tree_strip_any_location_wrapper): New inline function.
6093 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
6096 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
6097 stack_realign_offset for the largest alignment of stack slot
6099 (ix86_find_max_used_stack_alignment): New function.
6100 (ix86_finalize_stack_frame_flags): Use it. Set
6101 max_used_stack_alignment if we don't realign stack.
6102 * config/i386/i386.h (machine_function): Add
6103 max_used_stack_alignment.
6105 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
6107 * config/arm/arm.opt (-mbranch-cost): New option.
6108 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
6111 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
6114 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
6115 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
6117 2018-01-10 Richard Biener <rguenther@suse.de>
6120 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
6121 early out so it also covers the case where we have a non-NULL
6124 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6126 PR tree-optimization/83753
6127 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
6128 for non-strided grouped accesses if the number of elements is 1.
6130 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6133 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
6134 * i386.h (TARGET_USE_GATHER): Define.
6135 * x86-tune.def (X86_TUNE_USE_GATHER): New.
6137 2018-01-10 Martin Liska <mliska@suse.cz>
6140 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
6141 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
6143 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
6144 CLEANUP_NO_PARTITIONING is not set.
6146 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6148 * doc/rtl.texi: Remove documentation of (const ...) wrappers
6149 for vectors, as a partial revert of r254296.
6150 * rtl.h (const_vec_p): Delete.
6151 (const_vec_duplicate_p): Don't test for vector CONSTs.
6152 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
6153 * expmed.c (make_tree): Likewise.
6156 * common.md (E, F): Use CONSTANT_P instead of checking for
6158 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
6159 checking for CONST_VECTOR.
6161 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6164 * predict.c (force_edge_cold): Handle in more sane way edges
6167 2018-01-09 Carl Love <cel@us.ibm.com>
6169 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
6171 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
6172 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
6173 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
6174 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
6175 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
6176 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
6177 * config/rs6000/rs6000-protos.h: Add extern defition for
6178 rs6000_generate_float2_double_code.
6179 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
6181 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
6182 (float2_v2df): Add define_expand.
6184 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
6187 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
6188 op_mode in the force_to_mode call.
6190 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6192 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
6193 instead of checking each element individually.
6194 (aarch64_evpc_uzp): Likewise.
6195 (aarch64_evpc_zip): Likewise.
6196 (aarch64_evpc_ext): Likewise.
6197 (aarch64_evpc_rev): Likewise.
6198 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
6199 instead of checking each element individually. Return true without
6201 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
6202 whether all selected elements come from the same input, instead of
6203 checking each element individually. Remove calls to gen_rtx_REG,
6204 start_sequence and end_sequence and instead assert that no rtl is
6207 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6209 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
6210 order of HIGH and CONST checks.
6212 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6214 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
6215 if the destination isn't an SSA_NAME.
6217 2018-01-09 Richard Biener <rguenther@suse.de>
6219 PR tree-optimization/83668
6220 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
6222 (canonicalize_loop_form): ... here, renamed from ...
6223 (canonicalize_loop_closed_ssa_form): ... this and amended to
6224 swap successor edges for loop exit blocks to make us use
6225 the RPO order we need for initial schedule generation.
6227 2018-01-09 Joseph Myers <joseph@codesourcery.com>
6229 PR tree-optimization/64811
6230 * match.pd: When optimizing comparisons with Inf, avoid
6231 introducing or losing exceptions from comparisons with NaN.
6233 2018-01-09 Martin Liska <mliska@suse.cz>
6236 * asan.c (shadow_mem_size): Add gcc_assert.
6238 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
6240 Don't save registers in main().
6243 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
6244 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
6245 * config/avr/avr.c (avr_set_current_function): Don't error if
6246 naked, OS_task or OS_main are specified at the same time.
6247 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
6249 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
6251 * common/config/avr/avr-common.c (avr_option_optimization_table):
6252 Switch on -mmain-is-OS_task for optimizing compilations.
6254 2018-01-09 Richard Biener <rguenther@suse.de>
6256 PR tree-optimization/83572
6257 * graphite.c: Include cfganal.h.
6258 (graphite_transform_loops): Connect infinite loops to exit
6259 and remove fake edges at the end.
6261 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6263 * ipa-inline.c (edge_badness): Revert accidental checkin.
6265 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6268 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
6269 symbols; not inline clones.
6271 2018-01-09 Jakub Jelinek <jakub@redhat.com>
6274 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
6275 hard registers. Formatting fixes.
6277 PR preprocessor/83722
6278 * gcc.c (try_generate_repro): Pass
6279 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
6280 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
6283 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
6284 Kito Cheng <kito.cheng@gmail.com>
6286 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
6287 (riscv_leaf_function_p): Delete.
6288 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
6290 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6292 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
6294 (do_ifelse): New function.
6295 (do_isel): New function.
6296 (do_sub3): New function.
6297 (do_add3): New function.
6298 (do_load_mask_compare): New function.
6299 (do_overlap_load_compare): New function.
6300 (expand_compare_loop): New function.
6301 (expand_block_compare): Call expand_compare_loop() when appropriate.
6302 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
6304 (-mblock-compare-inline-loop-limit): New option.
6306 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6309 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
6310 Reverse order of second and third operands in first alternative.
6311 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
6312 of first and second elements in UNSPEC_VPERMR vector.
6313 (altivec_expand_vec_perm_le): Likewise.
6315 2018-01-08 Jeff Law <law@redhat.com>
6317 PR rtl-optimizatin/81308
6318 * tree-switch-conversion.c (cfg_altered): New file scoped static.
6319 (process_switch): If group_case_labels makes a change, then set
6321 (pass_convert_switch::execute): If a switch is converted, then
6322 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
6324 PR rtl-optimization/81308
6325 * recog.c (split_all_insns): Conditionally cleanup the CFG after
6328 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
6330 PR target/83663 - Revert r255946
6331 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
6332 generation for cases where splatting a value is not useful.
6333 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
6334 across a vec_duplicate and a paradoxical subreg forming a vector
6335 mode to a vec_concat.
6337 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6339 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
6340 -march=armv8.3-a variants.
6341 * config/arm/t-multilib: Likewise.
6342 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
6344 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6346 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
6348 (cceq_ior_compare_complement): Give it a name so I can use it, and
6349 change boolean_or_operator predicate to boolean_operator so it can
6350 be used to generate a crand.
6351 (eqne): New code iterator.
6352 (bd/bd_neg): New code_attrs.
6353 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
6354 a single define_insn.
6355 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
6356 decrement (bdnzt/bdnzf/bdzt/bdzf).
6357 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
6358 with the new names of the branch decrement patterns, and added the
6359 names of the branch decrement conditional patterns.
6361 2018-01-08 Richard Biener <rguenther@suse.de>
6363 PR tree-optimization/83563
6364 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
6367 2018-01-08 Richard Biener <rguenther@suse.de>
6370 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
6372 2018-01-08 Richard Biener <rguenther@suse.de>
6374 PR tree-optimization/83685
6375 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
6376 references to abnormals.
6378 2018-01-08 Richard Biener <rguenther@suse.de>
6381 * dwarf2out.c (output_indirect_strings): Handle empty
6382 skeleton_debug_str_hash.
6383 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
6385 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6387 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
6388 (emit_store_direct): Likewise.
6389 (arc_trampoline_adjust_address): Likewise.
6390 (arc_asm_trampoline_template): New function.
6391 (arc_initialize_trampoline): Use asm_trampoline_template.
6392 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
6393 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
6394 * config/arc/arc.md (flush_icache): Delete pattern.
6396 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6398 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
6399 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
6402 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6405 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
6406 by not USED_FOR_TARGET.
6407 (make_pass_resolve_sw_modes): Likewise.
6409 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6411 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
6414 2018-01-08 Richard Biener <rguenther@suse.de>
6417 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
6419 2018-01-08 Richard Biener <rguenther@suse.de>
6422 * match.pd ((t * 2) / 2) -> t): Add missing :c.
6424 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
6427 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
6428 basic blocks with a small number of successors.
6429 (convert_control_dep_chain_into_preds): Improve handling of
6431 (dump_predicates): Split apart into...
6432 (dump_pred_chain): ...here...
6433 (dump_pred_info): ...and here.
6434 (can_one_predicate_be_invalidated_p): Add debugging printfs.
6435 (can_chain_union_be_invalidated_p): Improve check for invalidation
6437 (uninit_uses_cannot_happen): Avoid unnecessary if
6438 convert_control_dep_chain_into_preds yielded nothing.
6440 2018-01-06 Martin Sebor <msebor@redhat.com>
6442 PR tree-optimization/83640
6443 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
6444 subtracting negative offset from size.
6445 (builtin_access::overlap): Adjust offset bounds of the access to fall
6446 within the size of the object if possible.
6448 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
6450 PR rtl-optimization/83699
6451 * expmed.c (extract_bit_field_1): Restrict the vector usage of
6452 extract_bit_field_as_subreg to cases in which the extracted
6453 value is also a vector.
6455 * lra-constraints.c (process_alt_operands): Test for the equivalence
6456 substitutions when detecting a possible reload cycle.
6458 2018-01-06 Jakub Jelinek <jakub@redhat.com>
6461 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
6462 by default if flag_selective_schedling{,2}. Formatting fixes.
6464 PR rtl-optimization/83682
6465 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
6466 if it has non-VECTOR_MODE element mode.
6467 (vec_duplicate_p): Likewise.
6470 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
6471 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
6473 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6476 * config/i386/i386-builtin.def
6477 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
6478 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
6479 Require also OPTION_MASK_ISA_AVX512F in addition to
6480 OPTION_MASK_ISA_GFNI.
6481 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
6482 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
6483 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
6484 to OPTION_MASK_ISA_GFNI.
6485 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
6486 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
6487 OPTION_MASK_ISA_AVX512BW.
6488 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
6489 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
6490 addition to OPTION_MASK_ISA_GFNI.
6491 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
6492 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
6493 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
6494 to OPTION_MASK_ISA_GFNI.
6495 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
6496 a requirement for all ISAs rather than any of them with a few
6498 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
6500 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
6501 bitmasks to be enabled with 3 exceptions, instead of requiring any
6502 enabled ISA with lots of exceptions.
6503 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
6504 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
6505 Change avx512bw in isa attribute to avx512f.
6506 * config/i386/sgxintrin.h: Add license boilerplate.
6507 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
6508 to __AVX512F__ and __AVX512VL to __AVX512VL__.
6509 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
6510 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
6512 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
6513 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
6514 temporarily sse2 rather than sse if not enabled already.
6517 * config/i386/sse.md (VI248_VLBW): Rename to ...
6518 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
6519 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
6520 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
6521 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
6522 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
6523 mode iterator instead of VI248_VLBW.
6525 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
6527 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
6528 (record_modified): Skip clobbers; add debug output.
6529 (param_change_prob): Use sreal frequencies.
6531 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6533 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
6534 punt for user-aligned variables.
6536 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6538 * tree-chrec.c (chrec_contains_symbols): Return true for
6541 2018-01-05 Sudakshina Das <sudi.das@arm.com>
6544 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
6545 of (x|y) == x for BICS pattern.
6547 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6549 PR tree-optimization/83605
6550 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
6551 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
6554 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
6556 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
6557 * config/epiphany/rtems.h: New file.
6559 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6560 Uros Bizjak <ubizjak@gmail.com>
6563 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
6564 QIreg_operand instead of register_operand predicate.
6565 * config/i386/i386.c (ix86_rop_should_change_byte_p,
6566 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
6567 comments instead of -fmitigate[-_]rop.
6569 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6572 * cgraphunit.c (symbol_table::compile): Switch to text_section
6573 before calling assembly_start debug hook.
6574 * run-rtl-passes.c (run_rtl_passes): Likewise.
6577 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6579 * tree-vrp.c (extract_range_from_binary_expr_1): Check
6580 range_int_cst_p rather than !symbolic_range_p before calling
6581 extract_range_from_multiplicative_op_1.
6583 2018-01-04 Jeff Law <law@redhat.com>
6585 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
6586 redundant test in assertion.
6588 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6590 * doc/rtl.texi: Document machine_mode wrapper classes.
6592 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6594 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
6597 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6599 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
6600 the VEC_PERM_EXPR fold to fail.
6602 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6605 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
6606 to switched_sections.
6608 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6611 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
6614 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
6617 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
6618 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
6620 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6623 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
6624 is BLKmode and bitpos not zero or mode change is needed.
6626 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6629 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
6632 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
6635 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
6636 instead of MULT rtx. Update all corresponding splitters.
6638 (*ssub<modesuffix>): Ditto.
6640 (*cmp_sadd_di): Update split patterns.
6641 (*cmp_sadd_si): Ditto.
6642 (*cmp_sadd_sidi): Ditto.
6643 (*cmp_ssub_di): Ditto.
6644 (*cmp_ssub_si): Ditto.
6645 (*cmp_ssub_sidi): Ditto.
6646 * config/alpha/predicates.md (const23_operand): New predicate.
6647 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
6648 Look for ASHIFT, not MULT inner operand.
6649 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
6651 2018-01-04 Martin Liska <mliska@suse.cz>
6653 PR gcov-profile/83669
6654 * gcov.c (output_intermediate_file): Add version to intermediate
6656 * doc/gcov.texi: Document new field 'version' in intermediate
6657 file format. Fix location of '-k' option of gcov command.
6659 2018-01-04 Martin Liska <mliska@suse.cz>
6662 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
6664 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6666 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
6668 2018-01-03 Martin Sebor <msebor@redhat.com>
6670 PR tree-optimization/83655
6671 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
6672 checking calls with invalid arguments.
6674 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6676 * tree-vect-stmts.c (vect_get_store_rhs): New function.
6677 (vectorizable_mask_load_store): Delete.
6678 (vectorizable_call): Return false for masked loads and stores.
6679 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
6680 instead of gimple_assign_rhs1.
6681 (vectorizable_load): Handle IFN_MASK_LOAD.
6682 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
6684 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6686 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
6688 (vectorizable_mask_load_store): ...here.
6689 (vectorizable_load): ...and here.
6691 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6693 * tree-vect-stmts.c (vect_build_all_ones_mask)
6694 (vect_build_zero_merge_argument): New functions, split out from...
6695 (vectorizable_load): ...here.
6697 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6699 * tree-vect-stmts.c (vect_check_store_rhs): New function,
6701 (vectorizable_mask_load_store): ...here.
6702 (vectorizable_store): ...and here.
6704 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6706 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
6708 (vectorizable_mask_load_store): ...here.
6710 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6712 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
6713 (vect_model_store_cost): Take a vec_load_store_type instead of a
6715 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
6716 (vect_model_store_cost): Take a vec_load_store_type instead of a
6718 (vectorizable_mask_load_store): Update accordingly.
6719 (vectorizable_store): Likewise.
6720 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
6722 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6724 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
6725 IFN_MASK_LOAD calls here rather than...
6726 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
6728 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6729 Alan Hayward <alan.hayward@arm.com>
6730 David Sherwood <david.sherwood@arm.com>
6732 * expmed.c (extract_bit_field_1): For vector extracts,
6733 fall back to extract_bit_field_as_subreg if vec_extract
6736 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6737 Alan Hayward <alan.hayward@arm.com>
6738 David Sherwood <david.sherwood@arm.com>
6740 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
6741 they are variable or constant sized.
6742 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
6743 slots for constant-sized data.
6745 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6746 Alan Hayward <alan.hayward@arm.com>
6747 David Sherwood <david.sherwood@arm.com>
6749 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
6750 handling COND_EXPRs with boolean comparisons, try to find a better
6751 basis for the mask type than the boolean itself.
6753 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6755 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
6756 is calculated and how it can be overridden.
6757 * genmodes.c (max_bitsize_mode_any_mode): New variable.
6758 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
6760 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
6763 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6764 Alan Hayward <alan.hayward@arm.com>
6765 David Sherwood <david.sherwood@arm.com>
6767 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
6768 Remove the mode argument.
6769 (aarch64_simd_valid_immediate): Remove the mode and inverse
6771 * config/aarch64/iterators.md (bitsize): New iterator.
6772 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
6773 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
6774 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
6775 aarch64_simd_valid_immediate.
6776 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
6777 (aarch64_reg_or_bic_imm): Likewise.
6778 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
6779 with an insn_type enum and msl with a modifier_type enum.
6780 Replace element_width with a scalar_mode. Change the shift
6781 to unsigned int. Add constructors for scalar_float_mode and
6782 scalar_int_mode elements.
6783 (aarch64_vect_float_const_representable_p): Delete.
6784 (aarch64_can_const_movi_rtx_p)
6785 (aarch64_simd_scalar_immediate_valid_for_move)
6786 (aarch64_simd_make_constant): Update call to
6787 aarch64_simd_valid_immediate.
6788 (aarch64_advsimd_valid_immediate_hs): New function.
6789 (aarch64_advsimd_valid_immediate): Likewise.
6790 (aarch64_simd_valid_immediate): Remove mode and inverse
6791 arguments. Rewrite to use the above. Use const_vec_duplicate_p
6792 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
6793 and aarch64_float_const_representable_p on the result.
6794 (aarch64_output_simd_mov_immediate): Remove mode argument.
6795 Update call to aarch64_simd_valid_immediate and use of
6796 simd_immediate_info.
6797 (aarch64_output_scalar_simd_mov_immediate): Update call
6800 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6801 Alan Hayward <alan.hayward@arm.com>
6802 David Sherwood <david.sherwood@arm.com>
6804 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
6805 (mode_nunits): Likewise CONST_MODE_NUNITS.
6806 * machmode.def (ADJUST_NUNITS): Document.
6807 * genmodes.c (mode_data::need_nunits_adj): New field.
6808 (blank_mode): Update accordingly.
6809 (adj_nunits): New variable.
6810 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
6812 (emit_mode_size_inline): Set need_bytesize_adj for all modes
6813 listed in adj_nunits.
6814 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
6815 listed in adj_nunits. Don't emit case statements for such modes.
6816 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
6817 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
6818 nothing if adj_nunits is nonnull.
6819 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
6820 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
6821 (emit_mode_fbit): Update use of print_maybe_const_decl.
6822 (emit_move_size): Likewise. Treat the array as non-const
6824 (emit_mode_adjustments): Handle adj_nunits.
6826 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6828 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
6829 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
6830 (VECTOR_MODES): Use it.
6831 (make_vector_modes): Take the prefix as an argument.
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 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
6838 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
6839 for MODE_VECTOR_BOOL.
6840 * machmode.def (VECTOR_BOOL_MODE): Document.
6841 * genmodes.c (VECTOR_BOOL_MODE): New macro.
6842 (make_vector_bool_mode): New function.
6843 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
6845 * lto-streamer-in.c (lto_input_mode_table): Likewise.
6846 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
6848 * stor-layout.c (int_mode_for_mode): Likewise.
6849 * tree.c (build_vector_type_for_mode): Likewise.
6850 * varasm.c (output_constant_pool_2): Likewise.
6851 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
6852 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
6853 for MODE_VECTOR_BOOL.
6854 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
6855 of mode class checks.
6856 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
6857 instead of a list of mode class checks.
6858 (expand_vector_scalar_condition): Likewise.
6859 (type_for_widest_vector_mode): Handle BImode as an inner mode.
6861 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6862 Alan Hayward <alan.hayward@arm.com>
6863 David Sherwood <david.sherwood@arm.com>
6865 * machmode.h (mode_size): Change from unsigned short to
6867 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
6868 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6869 or if measurement_type is not polynomial.
6870 (fixed_size_mode::includes_p): Check for constant-sized modes.
6871 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
6872 return a poly_uint16 rather than an unsigned short.
6873 (emit_mode_size): Change the type of mode_size from unsigned short
6874 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
6875 (emit_mode_adjustments): Cope with polynomial vector sizes.
6876 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6878 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6880 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
6881 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
6882 * caller-save.c (setup_save_areas): Likewise.
6883 (replace_reg_with_saved_mem): Likewise.
6884 * calls.c (emit_library_call_value_1): Likewise.
6885 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
6886 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
6887 (gen_lowpart_for_combine): Likewise.
6888 * convert.c (convert_to_integer_1): Likewise.
6889 * cse.c (equiv_constant, cse_insn): Likewise.
6890 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
6891 (cselib_subst_to_values): Likewise.
6892 * dce.c (word_dce_process_block): Likewise.
6893 * df-problems.c (df_word_lr_mark_ref): Likewise.
6894 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
6895 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
6896 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
6897 (rtl_for_decl_location): Likewise.
6898 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
6899 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
6900 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
6901 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
6902 (expand_expr_real_1): Likewise.
6903 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
6904 (pad_below): Likewise.
6905 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6906 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
6907 * ira.c (get_subreg_tracking_sizes): Likewise.
6908 * ira-build.c (ira_create_allocno_objects): Likewise.
6909 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
6910 (ira_sort_regnos_for_alter_reg): Likewise.
6911 * ira-costs.c (record_operand_costs): Likewise.
6912 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
6913 (resolve_simple_move): Likewise.
6914 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
6915 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
6916 (lra_constraints): Likewise.
6917 (CONST_POOL_OK_P): Reject variable-sized modes.
6918 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
6919 (add_pseudo_to_slot, lra_spill): Likewise.
6920 * omp-low.c (omp_clause_aligned_alignment): Likewise.
6921 * optabs-query.c (get_best_extraction_insn): Likewise.
6922 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
6923 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
6924 (expand_mult_highpart, valid_multiword_target_p): Likewise.
6925 * recog.c (offsettable_address_addr_space_p): Likewise.
6926 * regcprop.c (maybe_mode_change): Likewise.
6927 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
6928 * regrename.c (build_def_use): Likewise.
6929 * regstat.c (dump_reg_info): Likewise.
6930 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
6931 (find_reloads, find_reloads_subreg_address): Likewise.
6932 * reload1.c (eliminate_regs_1): Likewise.
6933 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
6934 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
6935 (simplify_binary_operation_1, simplify_subreg): Likewise.
6936 * targhooks.c (default_function_arg_padding): Likewise.
6937 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
6938 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
6939 (verify_gimple_assign_ternary): Likewise.
6940 * tree-inline.c (estimate_move_cost): Likewise.
6941 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
6942 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
6943 (get_address_cost_ainc): Likewise.
6944 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
6945 (vect_supportable_dr_alignment): Likewise.
6946 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6947 (vectorizable_reduction): Likewise.
6948 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
6949 (vectorizable_operation, vectorizable_load): Likewise.
6950 * tree.c (build_same_sized_truth_vector_type): Likewise.
6951 * valtrack.c (cleanup_auto_inc_dec): Likewise.
6952 * var-tracking.c (emit_note_insn_var_location): Likewise.
6953 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
6954 (ADDR_VEC_ALIGN): Likewise.
6956 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6957 Alan Hayward <alan.hayward@arm.com>
6958 David Sherwood <david.sherwood@arm.com>
6960 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
6962 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6963 or if measurement_type is polynomial.
6964 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
6965 * combine.c (make_extraction): Likewise.
6966 * dse.c (find_shift_sequence): Likewise.
6967 * dwarf2out.c (mem_loc_descriptor): Likewise.
6968 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
6969 (extract_bit_field, extract_low_bits): Likewise.
6970 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
6971 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
6972 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
6973 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
6974 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6975 * reload.c (find_reloads): Likewise.
6976 * reload1.c (alter_reg): Likewise.
6977 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
6978 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
6979 * tree-if-conv.c (predicate_mem_writes): Likewise.
6980 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
6981 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
6982 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
6983 * valtrack.c (dead_debug_insert_temp): Likewise.
6984 * varasm.c (mergeable_constant_section): Likewise.
6985 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
6987 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6988 Alan Hayward <alan.hayward@arm.com>
6989 David Sherwood <david.sherwood@arm.com>
6991 * expr.c (expand_assignment): Cope with polynomial mode sizes
6992 when assigning to a CONCAT.
6994 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6995 Alan Hayward <alan.hayward@arm.com>
6996 David Sherwood <david.sherwood@arm.com>
6998 * machmode.h (mode_precision): Change from unsigned short to
7000 (mode_to_precision): Return a poly_uint16 rather than an unsigned
7002 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
7003 or if measurement_type is not polynomial.
7004 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
7005 in which the mode is already known to be a scalar_int_mode.
7006 * genmodes.c (emit_mode_precision): Change the type of mode_precision
7007 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
7009 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7010 for GET_MODE_PRECISION.
7011 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7012 for GET_MODE_PRECISION.
7013 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
7015 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
7016 (expand_field_assignment, make_extraction): Likewise.
7017 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
7018 (get_last_value): Likewise.
7019 * convert.c (convert_to_integer_1): Likewise.
7020 * cse.c (cse_insn): Likewise.
7021 * expr.c (expand_expr_real_1): Likewise.
7022 * lra-constraints.c (simplify_operand_subreg): Likewise.
7023 * optabs-query.c (can_atomic_load_p): Likewise.
7024 * optabs.c (expand_atomic_load): Likewise.
7025 (expand_atomic_store): Likewise.
7026 * ree.c (combine_reaching_defs): Likewise.
7027 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
7028 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
7029 * tree.h (type_has_mode_precision_p): Likewise.
7030 * ubsan.c (instrument_si_overflow): Likewise.
7032 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7033 Alan Hayward <alan.hayward@arm.com>
7034 David Sherwood <david.sherwood@arm.com>
7036 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
7037 polynomial numbers of units.
7038 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
7039 (valid_vector_subparts_p): New function.
7040 (build_vector_type): Remove temporary shim and take the number
7041 of units as a poly_uint64 rather than an int.
7042 (build_opaque_vector_type): Take the number of units as a
7043 poly_uint64 rather than an int.
7044 * tree.c (build_vector_from_ctor): Handle polynomial
7045 TYPE_VECTOR_SUBPARTS.
7046 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
7047 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
7048 (build_vector_from_val): If the number of units is variable,
7049 use build_vec_duplicate_cst for constant operands and
7050 VEC_DUPLICATE_EXPR otherwise.
7051 (make_vector_type): Remove temporary is_constant ().
7052 (build_vector_type, build_opaque_vector_type): Take the number of
7053 units as a poly_uint64 rather than an int.
7054 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
7056 * cfgexpand.c (expand_debug_expr): Likewise.
7057 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
7058 (store_constructor, expand_expr_real_1): Likewise.
7059 (const_scalar_mask_from_tree): Likewise.
7060 * fold-const-call.c (fold_const_reduction): Likewise.
7061 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
7062 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
7063 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
7064 (fold_relational_const): Likewise.
7065 (native_interpret_vector): Likewise. Change the size from an
7066 int to an unsigned int.
7067 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
7068 TYPE_VECTOR_SUBPARTS.
7069 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
7070 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
7071 duplicating a non-constant operand into a variable-length vector.
7072 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
7073 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
7074 * ipa-icf.c (sem_variable::equals): Likewise.
7075 * match.pd: Likewise.
7076 * omp-simd-clone.c (simd_clone_subparts): Likewise.
7077 * print-tree.c (print_node): Likewise.
7078 * stor-layout.c (layout_type): Likewise.
7079 * targhooks.c (default_builtin_vectorization_cost): Likewise.
7080 * tree-cfg.c (verify_gimple_comparison): Likewise.
7081 (verify_gimple_assign_binary): Likewise.
7082 (verify_gimple_assign_ternary): Likewise.
7083 (verify_gimple_assign_single): Likewise.
7084 * tree-pretty-print.c (dump_generic_node): Likewise.
7085 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7086 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
7087 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
7088 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
7089 (vect_shift_permute_load_chain): Likewise.
7090 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
7091 (expand_vector_condition, optimize_vector_constructor): Likewise.
7092 (lower_vec_perm, get_compute_type): Likewise.
7093 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7094 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
7095 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
7096 (vect_recog_mask_conversion_pattern): Likewise.
7097 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
7098 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
7099 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7100 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
7101 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
7102 (vectorizable_shift, vectorizable_operation, vectorizable_store)
7103 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
7104 (supportable_widening_operation): Likewise.
7105 (supportable_narrowing_operation): Likewise.
7106 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
7108 * varasm.c (output_constant): Likewise.
7110 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7111 Alan Hayward <alan.hayward@arm.com>
7112 David Sherwood <david.sherwood@arm.com>
7114 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
7115 so that both the length == 3 and length != 3 cases set up their
7116 own permute vectors. Add comments explaining why we know the
7117 number of elements is constant.
7118 (vect_permute_load_chain): Likewise.
7120 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7121 Alan Hayward <alan.hayward@arm.com>
7122 David Sherwood <david.sherwood@arm.com>
7124 * machmode.h (mode_nunits): Change from unsigned char to
7126 (ONLY_FIXED_SIZE_MODES): New macro.
7127 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
7128 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
7129 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
7131 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
7132 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
7133 or if measurement_type is not polynomial.
7134 * genmodes.c (ZERO_COEFFS): New macro.
7135 (emit_mode_nunits_inline): Make mode_nunits_inline return a
7137 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
7138 Use ZERO_COEFFS when emitting initializers.
7139 * data-streamer.h (bp_pack_poly_value): New function.
7140 (bp_unpack_poly_value): Likewise.
7141 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7142 for GET_MODE_NUNITS.
7143 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7144 for GET_MODE_NUNITS.
7145 * tree.c (make_vector_type): Remove temporary shim and make
7146 the real function take the number of units as a poly_uint64
7148 (build_vector_type_for_mode): Handle polynomial nunits.
7149 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
7150 * emit-rtl.c (const_vec_series_p_1): Likewise.
7151 (gen_rtx_CONST_VECTOR): Likewise.
7152 * fold-const.c (test_vec_duplicate_folding): Likewise.
7153 * genrecog.c (validate_pattern): Likewise.
7154 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
7155 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7156 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
7157 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
7158 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
7159 * rtlanal.c (subreg_get_info): Likewise.
7160 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7161 (vect_grouped_load_supported): Likewise.
7162 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
7163 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
7164 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
7165 (simplify_const_unary_operation, simplify_binary_operation_1)
7166 (simplify_const_binary_operation, simplify_ternary_operation)
7167 (test_vector_ops_duplicate, test_vector_ops): Likewise.
7168 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
7169 instead of CONST_VECTOR_NUNITS.
7170 * varasm.c (output_constant_pool_2): Likewise.
7171 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
7172 explicit-encoded elements in the XVEC for variable-length vectors.
7174 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7176 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
7178 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7179 Alan Hayward <alan.hayward@arm.com>
7180 David Sherwood <david.sherwood@arm.com>
7182 * coretypes.h (fixed_size_mode): Declare.
7183 (fixed_size_mode_pod): New typedef.
7184 * builtins.h (target_builtins::x_apply_args_mode)
7185 (target_builtins::x_apply_result_mode): Change type to
7186 fixed_size_mode_pod.
7187 * builtins.c (apply_args_size, apply_result_size, result_vector)
7188 (expand_builtin_apply_args_1, expand_builtin_apply)
7189 (expand_builtin_return): Update accordingly.
7191 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7193 * cse.c (hash_rtx_cb): Hash only the encoded elements.
7194 * cselib.c (cselib_hash_rtx): Likewise.
7195 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
7196 CONST_VECTOR encoding.
7198 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7199 Jeff Law <law@redhat.com>
7202 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
7203 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
7204 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
7205 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
7208 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
7209 explicitly probe *sp in a noreturn function if there were any callee
7210 register saves or frame pointer is needed.
7212 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7215 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
7216 BLKmode for ternary, binary or unary expressions.
7219 * var-tracking.c (delete_vta_debug_insn): New inline function.
7220 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
7221 insns from get_insns () to NULL instead of each bb separately.
7222 Use delete_vta_debug_insn. No longer static.
7223 (vt_debug_insns_local, variable_tracking_main_1): Adjust
7224 delete_vta_debug_insns callers.
7225 * rtl.h (delete_vta_debug_insns): Declare.
7226 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
7227 instead of variable_tracking_main.
7229 2018-01-03 Martin Sebor <msebor@redhat.com>
7231 PR tree-optimization/83603
7232 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
7233 arguments past the endof the argument list in functions declared
7234 without a prototype.
7235 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
7236 Avoid checking when arguments are null.
7238 2018-01-03 Martin Sebor <msebor@redhat.com>
7241 * doc/extend.texi (attribute const): Fix a typo.
7242 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
7243 issuing -Wsuggest-attribute for void functions.
7245 2018-01-03 Martin Sebor <msebor@redhat.com>
7247 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
7248 offset_int::from instead of wide_int::to_shwi.
7249 (maybe_diag_overlap): Remove assertion.
7250 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
7251 * gimple-ssa-sprintf.c (format_directive): Same.
7252 (parse_directive): Same.
7253 (sprintf_dom_walker::compute_format_length): Same.
7254 (try_substitute_return_value): Same.
7256 2018-01-03 Jeff Law <law@redhat.com>
7259 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
7260 non-constant residual for zero at runtime and avoid probing in
7261 that case. Reorganize code for trailing problem to mirror handling
7264 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7266 PR tree-optimization/83501
7267 * tree-ssa-strlen.c (get_string_cst): New.
7268 (handle_char_store): Call get_string_cst.
7270 2018-01-03 Martin Liska <mliska@suse.cz>
7272 PR tree-optimization/83593
7273 * tree-ssa-strlen.c: Include tree-cfg.h.
7274 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
7275 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
7276 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
7278 (strlen_dom_walker::before_dom_children): Call
7279 gimple_purge_dead_eh_edges. Dump tranformation with details
7281 (strlen_dom_walker::before_dom_children): Update call by adding
7282 new argument cleanup_eh.
7283 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
7285 2018-01-03 Martin Liska <mliska@suse.cz>
7288 * cif-code.def (VARIADIC_THUNK): New enum value.
7289 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
7292 2018-01-03 Jan Beulich <jbeulich@suse.com>
7294 * sse.md (mov<mode>_internal): Tighten condition for when to use
7295 vmovdqu<ssescalarsize> for TI and OI modes.
7297 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7299 Update copyright years.
7301 2018-01-03 Martin Liska <mliska@suse.cz>
7304 * ipa-visibility.c (function_and_variable_visibility): Skip
7305 functions with noipa attribure.
7307 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7309 * gcc.c (process_command): Update copyright notice dates.
7310 * gcov-dump.c (print_version): Ditto.
7311 * gcov.c (print_version): Ditto.
7312 * gcov-tool.c (print_version): Ditto.
7313 * gengtype.c (create_file): Ditto.
7314 * doc/cpp.texi: Bump @copying's copyright year.
7315 * doc/cppinternals.texi: Ditto.
7316 * doc/gcc.texi: Ditto.
7317 * doc/gccint.texi: Ditto.
7318 * doc/gcov.texi: Ditto.
7319 * doc/install.texi: Ditto.
7320 * doc/invoke.texi: Ditto.
7322 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7324 * vector-builder.h (vector_builder::m_full_nelts): Change from
7325 unsigned int to poly_uint64.
7326 (vector_builder::full_nelts): Update prototype accordingly.
7327 (vector_builder::new_vector): Likewise.
7328 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
7329 (vector_builder::operator ==): Likewise.
7330 (vector_builder::finalize): Likewise.
7331 * int-vector-builder.h (int_vector_builder::int_vector_builder):
7332 Take the number of elements as a poly_uint64 rather than an
7334 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
7335 from unsigned int to poly_uint64.
7336 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
7337 (vec_perm_indices::new_vector): Likewise.
7338 (vec_perm_indices::length): Likewise.
7339 (vec_perm_indices::nelts_per_input): Likewise.
7340 (vec_perm_indices::input_nelts): Likewise.
7341 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
7342 number of elements per input as a poly_uint64 rather than an
7343 unsigned int. Use the original encoding for variable-length
7344 vectors, rather than clamping each individual element.
7345 For the second and subsequent elements in each pattern,
7346 clamp the step and base before clamping their sum.
7347 (vec_perm_indices::series_p): Handle polynomial element counts.
7348 (vec_perm_indices::all_in_range_p): Likewise.
7349 (vec_perm_indices_to_tree): Likewise.
7350 (vec_perm_indices_to_rtx): Likewise.
7351 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
7352 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
7353 (tree_vector_builder::new_binary_operation): Handle polynomial
7354 element counts. Return false if we need to know the number
7355 of elements at compile time.
7356 * fold-const.c (fold_vec_perm): Punt if the number of elements
7357 isn't known at compile time.
7359 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7361 * vec-perm-indices.h (vec_perm_builder): Change element type
7362 from HOST_WIDE_INT to poly_int64.
7363 (vec_perm_indices::element_type): Update accordingly.
7364 (vec_perm_indices::clamp): Handle polynomial element_types.
7365 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7366 (vec_perm_indices::all_in_range_p): Likewise.
7367 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
7369 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
7370 polynomial vec_perm_indices element types.
7371 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
7372 * fold-const.c (fold_vec_perm): Likewise.
7373 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
7374 * tree-vect-generic.c (lower_vec_perm): Likewise.
7375 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7376 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
7377 element type to HOST_WIDE_INT.
7379 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7380 Alan Hayward <alan.hayward@arm.com>
7381 David Sherwood <david.sherwood@arm.com>
7383 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
7384 rather than an int. Use plus_constant.
7385 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
7386 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
7388 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7389 Alan Hayward <alan.hayward@arm.com>
7390 David Sherwood <david.sherwood@arm.com>
7392 * calls.c (emit_call_1, expand_call): Change struct_value_size from
7393 a HOST_WIDE_INT to a poly_int64.
7395 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7396 Alan Hayward <alan.hayward@arm.com>
7397 David Sherwood <david.sherwood@arm.com>
7399 * calls.c (load_register_parameters): Cope with polynomial
7400 mode sizes. Require a constant size for BLKmode parameters
7401 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
7402 forces a parameter to be padded at the lsb end in order to
7403 fill a complete number of words, require the parameter size
7404 to be ordered wrt UNITS_PER_WORD.
7406 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7407 Alan Hayward <alan.hayward@arm.com>
7408 David Sherwood <david.sherwood@arm.com>
7410 * reload1.c (spill_stack_slot_width): Change element type
7411 from unsigned int to poly_uint64_pod.
7412 (alter_reg): Treat mode sizes as polynomial.
7414 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7415 Alan Hayward <alan.hayward@arm.com>
7416 David Sherwood <david.sherwood@arm.com>
7418 * reload.c (complex_word_subreg_p): New function.
7419 (reload_inner_reg_of_subreg, push_reload): Use it.
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 * lra-constraints.c (process_alt_operands): Reject matched
7426 operands whose sizes aren't ordered.
7427 (match_reload): Refer to this check here.
7429 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7430 Alan Hayward <alan.hayward@arm.com>
7431 David Sherwood <david.sherwood@arm.com>
7433 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
7434 that the mode size is in the set {1, 2, 4, 8, 16}.
7436 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7437 Alan Hayward <alan.hayward@arm.com>
7438 David Sherwood <david.sherwood@arm.com>
7440 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
7441 Use plus_constant instead of gen_rtx_PLUS.
7443 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7444 Alan Hayward <alan.hayward@arm.com>
7445 David Sherwood <david.sherwood@arm.com>
7447 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
7448 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
7449 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
7450 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
7451 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
7452 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
7453 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
7454 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
7455 * config/i386/i386.c (ix86_push_rounding): ...this new function.
7456 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
7458 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
7459 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
7460 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
7461 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
7462 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
7463 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
7464 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
7465 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
7466 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
7467 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
7469 * expr.c (emit_move_resolve_push): Treat the input and result
7470 of PUSH_ROUNDING as a poly_int64.
7471 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
7472 (emit_push_insn): Likewise.
7473 * lra-eliminations.c (mark_not_eliminable): Likewise.
7474 * recog.c (push_operand): Likewise.
7475 * reload1.c (elimination_effects): Likewise.
7476 * rtlanal.c (nonzero_bits1): Likewise.
7477 * calls.c (store_one_arg): Likewise. Require the padding to be
7478 known at compile time.
7480 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7481 Alan Hayward <alan.hayward@arm.com>
7482 David Sherwood <david.sherwood@arm.com>
7484 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
7485 Use plus_constant instead of gen_rtx_PLUS.
7487 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7488 Alan Hayward <alan.hayward@arm.com>
7489 David Sherwood <david.sherwood@arm.com>
7491 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
7494 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7495 Alan Hayward <alan.hayward@arm.com>
7496 David Sherwood <david.sherwood@arm.com>
7498 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
7499 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
7500 via stack temporaries. Treat the mode size as polynomial too.
7502 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7503 Alan Hayward <alan.hayward@arm.com>
7504 David Sherwood <david.sherwood@arm.com>
7506 * expr.c (expand_expr_real_2): When handling conversions involving
7507 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
7508 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
7509 as a poly_uint64 too.
7511 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7512 Alan Hayward <alan.hayward@arm.com>
7513 David Sherwood <david.sherwood@arm.com>
7515 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
7517 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7518 Alan Hayward <alan.hayward@arm.com>
7519 David Sherwood <david.sherwood@arm.com>
7521 * combine.c (can_change_dest_mode): Handle polynomial
7522 REGMODE_NATURAL_SIZE.
7523 * expmed.c (store_bit_field_1): Likewise.
7524 * expr.c (store_constructor): Likewise.
7525 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
7526 and polynomial REGMODE_NATURAL_SIZE.
7527 (gen_lowpart_common): Likewise.
7528 * reginfo.c (record_subregs_of_mode): Likewise.
7529 * rtlanal.c (read_modify_subreg_p): Likewise.
7531 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7532 Alan Hayward <alan.hayward@arm.com>
7533 David Sherwood <david.sherwood@arm.com>
7535 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
7536 numbers of elements.
7538 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7539 Alan Hayward <alan.hayward@arm.com>
7540 David Sherwood <david.sherwood@arm.com>
7542 * match.pd: Cope with polynomial numbers of vector elements.
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 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
7549 in a POINTER_PLUS_EXPR.
7551 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7552 Alan Hayward <alan.hayward@arm.com>
7553 David Sherwood <david.sherwood@arm.com>
7555 * omp-simd-clone.c (simd_clone_subparts): New function.
7556 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
7557 (ipa_simd_modify_function_body): Likewise.
7559 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7560 Alan Hayward <alan.hayward@arm.com>
7561 David Sherwood <david.sherwood@arm.com>
7563 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
7564 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
7565 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
7566 (expand_vector_condition, vector_element): Likewise.
7567 (subparts_gt): New function.
7568 (get_compute_type): Use subparts_gt.
7569 (count_type_subparts): Delete.
7570 (expand_vector_operations_1): Use subparts_gt instead of
7571 count_type_subparts.
7573 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7574 Alan Hayward <alan.hayward@arm.com>
7575 David Sherwood <david.sherwood@arm.com>
7577 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
7578 (vect_compile_time_alias): ...this new function. Do the calculation
7579 on poly_ints rather than trees.
7580 (vect_prune_runtime_alias_test_list): Update call accordingly.
7582 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7583 Alan Hayward <alan.hayward@arm.com>
7584 David Sherwood <david.sherwood@arm.com>
7586 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
7588 (vect_schedule_slp_instance): Likewise.
7590 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7591 Alan Hayward <alan.hayward@arm.com>
7592 David Sherwood <david.sherwood@arm.com>
7594 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
7595 constant and extern definitions for variable-length vectors.
7596 (vect_get_constant_vectors): Note that the number of units
7597 is known to be constant.
7599 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7600 Alan Hayward <alan.hayward@arm.com>
7601 David Sherwood <david.sherwood@arm.com>
7603 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
7604 of units as polynomial. Choose between WIDE and NARROW based
7607 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7608 Alan Hayward <alan.hayward@arm.com>
7609 David Sherwood <david.sherwood@arm.com>
7611 * tree-vect-stmts.c (simd_clone_subparts): New function.
7612 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
7614 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7615 Alan Hayward <alan.hayward@arm.com>
7616 David Sherwood <david.sherwood@arm.com>
7618 * tree-vect-stmts.c (vectorizable_call): Treat the number of
7619 vectors as polynomial. Use build_index_vector for
7622 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7623 Alan Hayward <alan.hayward@arm.com>
7624 David Sherwood <david.sherwood@arm.com>
7626 * tree-vect-stmts.c (get_load_store_type): Treat the number of
7627 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
7628 for variable-length vectors.
7629 (vectorizable_mask_load_store): Treat the number of units as
7630 polynomial, asserting that it is constant if the condition has
7631 already been enforced.
7632 (vectorizable_store, vectorizable_load): Likewise.
7634 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7635 Alan Hayward <alan.hayward@arm.com>
7636 David Sherwood <david.sherwood@arm.com>
7638 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
7639 of units as polynomial. Punt if we can't tell at compile time
7640 which vector contains the final result.
7642 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7643 Alan Hayward <alan.hayward@arm.com>
7644 David Sherwood <david.sherwood@arm.com>
7646 * tree-vect-loop.c (vectorizable_induction): Treat the number
7647 of units as polynomial. Punt on SLP inductions. Use an integer
7648 VEC_SERIES_EXPR for variable-length integer reductions. Use a
7649 cast of such a series for variable-length floating-point
7652 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7653 Alan Hayward <alan.hayward@arm.com>
7654 David Sherwood <david.sherwood@arm.com>
7656 * tree.h (build_index_vector): Declare.
7657 * tree.c (build_index_vector): New function.
7658 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
7659 of units as polynomial, forcibly converting it to a constant if
7660 vectorizable_reduction has already enforced the condition.
7661 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
7662 to create a {1,2,3,...} vector.
7663 (vectorizable_reduction): Treat the number of units as polynomial.
7664 Choose vectype_in based on the largest scalar element size rather
7665 than the smallest number of units. Enforce the restrictions
7668 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7669 Alan Hayward <alan.hayward@arm.com>
7670 David Sherwood <david.sherwood@arm.com>
7672 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
7673 number of units as polynomial.
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 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
7680 * target.def (autovectorize_vector_sizes): Return the vector sizes
7681 by pointer, using vector_sizes rather than a bitmask.
7682 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
7683 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
7684 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
7686 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
7687 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
7688 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
7689 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
7690 * omp-general.c (omp_max_vf): Likewise.
7691 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7692 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
7693 * tree-vect-loop.c (vect_analyze_loop): Likewise.
7694 * tree-vect-slp.c (vect_slp_bb): Likewise.
7695 * doc/tm.texi: Regenerate.
7696 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
7698 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
7699 the vector size as a poly_uint64 rather than an unsigned int.
7700 (current_vector_size): Change from an unsigned int to a poly_uint64.
7701 (get_vectype_for_scalar_type): Update accordingly.
7702 * tree.h (build_truth_vector_type): Take the size and number of
7703 units as a poly_uint64 rather than an unsigned int.
7704 (build_vector_type): Add a temporary overload that takes
7705 the number of units as a poly_uint64 rather than an unsigned int.
7706 * tree.c (make_vector_type): Likewise.
7707 (build_truth_vector_type): Take the number of units as a poly_uint64
7708 rather than an unsigned int.
7710 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7711 Alan Hayward <alan.hayward@arm.com>
7712 David Sherwood <david.sherwood@arm.com>
7714 * target.def (get_mask_mode): Take the number of units and length
7715 as poly_uint64s rather than unsigned ints.
7716 * targhooks.h (default_get_mask_mode): Update accordingly.
7717 * targhooks.c (default_get_mask_mode): Likewise.
7718 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
7719 * doc/tm.texi: Regenerate.
7721 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7722 Alan Hayward <alan.hayward@arm.com>
7723 David Sherwood <david.sherwood@arm.com>
7725 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
7726 * omp-general.c (omp_max_vf): Likewise.
7727 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
7728 (expand_omp_simd): Handle polynomial safelen.
7729 * omp-low.c (omplow_simd_context): Add a default constructor.
7730 (omplow_simd_context::max_vf): Change from int to poly_uint64.
7731 (lower_rec_simd_input_clauses): Update accordingly.
7732 (lower_rec_input_clauses): Likewise.
7734 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7735 Alan Hayward <alan.hayward@arm.com>
7736 David Sherwood <david.sherwood@arm.com>
7738 * tree-vectorizer.h (vect_nunits_for_cost): New function.
7739 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
7740 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
7741 (vect_analyze_slp_cost): Likewise.
7742 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
7743 (vect_model_load_cost): Likewise.
7745 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7746 Alan Hayward <alan.hayward@arm.com>
7747 David Sherwood <david.sherwood@arm.com>
7749 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
7750 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
7751 from an unsigned int * to a poly_uint64_pod *.
7752 (calculate_unrolling_factor): New function.
7753 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
7755 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7756 Alan Hayward <alan.hayward@arm.com>
7757 David Sherwood <david.sherwood@arm.com>
7759 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
7760 from an unsigned int to a poly_uint64.
7761 (_loop_vec_info::slp_unrolling_factor): Likewise.
7762 (_loop_vec_info::vectorization_factor): Change from an int
7764 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
7765 (vect_get_num_vectors): New function.
7766 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
7767 (vect_get_num_copies): Use vect_get_num_vectors.
7768 (vect_analyze_data_ref_dependences): Change max_vf from an int *
7769 to an unsigned int *.
7770 (vect_analyze_data_refs): Change min_vf from an int * to a
7772 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7773 than an unsigned HOST_WIDE_INT.
7774 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
7775 (vect_analyze_data_ref_dependence): Change max_vf from an int *
7776 to an unsigned int *.
7777 (vect_analyze_data_ref_dependences): Likewise.
7778 (vect_compute_data_ref_alignment): Handle polynomial vf.
7779 (vect_enhance_data_refs_alignment): Likewise.
7780 (vect_prune_runtime_alias_test_list): Likewise.
7781 (vect_shift_permute_load_chain): Likewise.
7782 (vect_supportable_dr_alignment): Likewise.
7783 (dependence_distance_ge_vf): Take the vectorization factor as a
7784 poly_uint64 rather than an unsigned HOST_WIDE_INT.
7785 (vect_analyze_data_refs): Change min_vf from an int * to a
7787 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
7788 vfm1 as a poly_uint64 rather than an int. Make the same change
7789 for the returned bound_scalar.
7790 (vect_gen_vector_loop_niters): Handle polynomial vf.
7791 (vect_do_peeling): Likewise. Update call to
7792 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
7793 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
7795 * tree-vect-loop.c (vect_determine_vectorization_factor)
7796 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
7797 (vect_get_known_peeling_cost): Likewise.
7798 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
7799 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
7800 (vect_transform_loop): Likewise. Use the lowest possible VF when
7801 updating the upper bounds of the loop.
7802 (vect_min_worthwhile_factor): Make static. Return an unsigned int
7804 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
7805 polynomial unroll factors.
7806 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
7807 (vect_make_slp_decision): Likewise.
7808 (vect_supported_load_permutation_p): Likewise, and polynomial
7810 (vect_analyze_slp_cost): Handle polynomial vf.
7811 (vect_slp_analyze_node_operations): Likewise.
7812 (vect_slp_analyze_bb_1): Likewise.
7813 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7814 than an unsigned HOST_WIDE_INT.
7815 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
7816 (vectorizable_load): Handle polynomial vf.
7817 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
7819 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
7821 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7822 Alan Hayward <alan.hayward@arm.com>
7823 David Sherwood <david.sherwood@arm.com>
7825 * match.pd: Handle bit operations involving three constants
7826 and try to fold one pair.
7828 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7830 * tree-vect-loop-manip.c: Include gimple-fold.h.
7831 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
7832 niters_maybe_zero parameters. Handle other cases besides a step of 1.
7833 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
7834 Add a path that uses a step of VF instead of 1, but disable it
7836 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
7837 and niters_no_overflow parameters. Update calls to
7838 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
7839 Create a new SSA name if the latter choses to use a ste other
7840 than zero, and return it via niters_vector_mult_vf_var.
7841 * tree-vect-loop.c (vect_transform_loop): Update calls to
7842 vect_do_peeling, vect_gen_vector_loop_niters and
7843 slpeel_make_loop_iterate_ntimes.
7844 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
7845 (vect_gen_vector_loop_niters): Update declarations after above changes.
7847 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
7849 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
7850 128-bit round to integer instructions.
7851 (ceil<mode>2): Likewise.
7852 (btrunc<mode>2): Likewise.
7853 (round<mode>2): Likewise.
7855 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7857 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
7858 unaligned VSX load/store on P8/P9.
7859 (expand_block_clear): Allow the use of unaligned VSX
7860 load/store on P8/P9.
7862 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7864 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
7866 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
7867 swap associated with both a load and a store.
7869 2018-01-02 Andrew Waterman <andrew@sifive.com>
7871 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
7872 * config/riscv/riscv.md (clear_cache): Use it.
7874 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
7876 * web.c: Remove out-of-date comment.
7878 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7880 * expr.c (fixup_args_size_notes): Check that any existing
7881 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
7882 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
7883 (emit_single_push_insn): ...here.
7885 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7887 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
7888 (const_vector_encoded_nelts): New function.
7889 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
7890 (const_vector_int_elt, const_vector_elt): Declare.
7891 * emit-rtl.c (const_vector_int_elt_1): New function.
7892 (const_vector_elt): Likewise.
7893 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
7894 of CONST_VECTOR_ELT.
7896 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7898 * expr.c: Include rtx-vector-builder.h.
7899 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
7900 directly on the tree encoding.
7901 (const_vector_from_tree): Likewise.
7902 * optabs.c: Include rtx-vector-builder.h.
7903 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
7904 sequence of "u" values.
7905 * vec-perm-indices.c: Include rtx-vector-builder.h.
7906 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
7907 directly on the vec_perm_indices encoding.
7909 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7911 * doc/rtl.texi (const_vector): Describe new encoding scheme.
7912 * Makefile.in (OBJS): Add rtx-vector-builder.o.
7913 * rtx-vector-builder.h: New file.
7914 * rtx-vector-builder.c: Likewise.
7915 * rtl.h (rtx_def::u2): Add a const_vector field.
7916 (CONST_VECTOR_NPATTERNS): New macro.
7917 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
7918 (CONST_VECTOR_DUPLICATE_P): Likewise.
7919 (CONST_VECTOR_STEPPED_P): Likewise.
7920 (CONST_VECTOR_ENCODED_ELT): Likewise.
7921 (const_vec_duplicate_p): Check for a duplicated vector encoding.
7922 (unwrap_const_vec_duplicate): Likewise.
7923 (const_vec_series_p): Check for a non-duplicated vector encoding.
7924 Say that the function only returns true for integer vectors.
7925 * emit-rtl.c: Include rtx-vector-builder.h.
7926 (gen_const_vec_duplicate_1): Delete.
7927 (gen_const_vector): Call gen_const_vec_duplicate instead of
7928 gen_const_vec_duplicate_1.
7929 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
7930 (gen_const_vec_duplicate): Use rtx_vector_builder.
7931 (gen_const_vec_series): Likewise.
7932 (gen_rtx_CONST_VECTOR): Likewise.
7933 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
7934 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7935 Build a new vector rather than modifying a CONST_VECTOR in-place.
7936 (handle_special_swappables): Update call accordingly.
7937 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
7938 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7939 Build a new vector rather than modifying a CONST_VECTOR in-place.
7940 (handle_special_swappables): Update call accordingly.
7942 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7944 * simplify-rtx.c (simplify_const_binary_operation): Use
7945 CONST_VECTOR_ELT instead of XVECEXP.
7947 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7949 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
7950 the selector elements to be different from the data elements
7951 if the selector is a VECTOR_CST.
7952 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
7953 ssizetype for the selector.
7955 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7957 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
7958 before testing each element individually.
7959 * tree-vect-generic.c (lower_vec_perm): Likewise.
7961 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7963 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
7964 * selftest-run-tests.c (selftest::run_tests): Call it.
7965 * vector-builder.h (vector_builder::operator ==): New function.
7966 (vector_builder::operator !=): Likewise.
7967 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
7968 (vec_perm_indices::all_from_input_p): New function.
7969 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7970 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
7971 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
7972 instead of reading the VECTOR_CST directly. Detect whether both
7973 vector inputs are the same before constructing the vec_perm_indices,
7974 and update the number of inputs argument accordingly. Use the
7975 utility functions added above. Only construct sel2 if we need to.
7977 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7979 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
7980 the broadcast of the low byte.
7981 (expand_mult_highpart): Use an explicit encoding for the permutes.
7982 * optabs-query.c (can_mult_highpart_p): Likewise.
7983 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
7984 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7985 (vectorizable_bswap): Likewise.
7986 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
7987 explicit encoding for the power-of-2 permutes.
7988 (vect_permute_store_chain): Likewise.
7989 (vect_grouped_load_supported): Likewise.
7990 (vect_permute_load_chain): Likewise.
7992 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7994 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
7995 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
7996 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
7997 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7998 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
7999 (vect_gen_perm_mask_any): Likewise.
8001 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8003 * int-vector-builder.h: New file.
8004 * vec-perm-indices.h: Include int-vector-builder.h.
8005 (vec_perm_indices): Redefine as an int_vector_builder.
8006 (auto_vec_perm_indices): Delete.
8007 (vec_perm_builder): Redefine as a stand-alone class.
8008 (vec_perm_indices::vec_perm_indices): New function.
8009 (vec_perm_indices::clamp): Likewise.
8010 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
8011 (vec_perm_indices::new_vector): New function.
8012 (vec_perm_indices::new_expanded_vector): Update for new
8013 vec_perm_indices class.
8014 (vec_perm_indices::rotate_inputs): New function.
8015 (vec_perm_indices::all_in_range_p): Operate directly on the
8016 encoded form, without computing elided elements.
8017 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
8018 encoding. Update for new vec_perm_indices class.
8019 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
8020 the given vec_perm_builder.
8021 (expand_vec_perm_var): Update vec_perm_builder constructor.
8022 (expand_mult_highpart): Use vec_perm_builder instead of
8023 auto_vec_perm_indices.
8024 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
8025 vec_perm_indices instead of auto_vec_perm_indices. Use a single
8026 or double series encoding as appropriate.
8027 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
8028 vec_perm_indices instead of auto_vec_perm_indices.
8029 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8030 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8031 (vect_permute_store_chain): Likewise.
8032 (vect_grouped_load_supported): Likewise.
8033 (vect_permute_load_chain): Likewise.
8034 (vect_shift_permute_load_chain): Likewise.
8035 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8036 (vect_transform_slp_perm_load): Likewise.
8037 (vect_schedule_slp_instance): Likewise.
8038 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8039 (vectorizable_mask_load_store): Likewise.
8040 (vectorizable_bswap): Likewise.
8041 (vectorizable_store): Likewise.
8042 (vectorizable_load): Likewise.
8043 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
8044 vec_perm_indices instead of auto_vec_perm_indices. Use
8045 tree_to_vec_perm_builder to read the vector from a tree.
8046 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
8047 vec_perm_builder instead of a vec_perm_indices.
8048 (have_whole_vector_shift): Use vec_perm_builder and
8049 vec_perm_indices instead of auto_vec_perm_indices. Leave the
8050 truncation to calc_vec_perm_mask_for_shift.
8051 (vect_create_epilog_for_reduction): Likewise.
8052 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
8053 from auto_vec_perm_indices to vec_perm_indices.
8054 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8055 instead of changing individual elements.
8056 (aarch64_vectorize_vec_perm_const): Use new_vector to install
8057 the vector in d.perm.
8058 * config/arm/arm.c (expand_vec_perm_d::perm): Change
8059 from auto_vec_perm_indices to vec_perm_indices.
8060 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8061 instead of changing individual elements.
8062 (arm_vectorize_vec_perm_const): Use new_vector to install
8063 the vector in d.perm.
8064 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
8065 Update vec_perm_builder constructor.
8066 (rs6000_expand_interleave): Likewise.
8067 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
8068 (rs6000_expand_interleave): Likewise.
8070 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8072 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
8073 to qimode could truncate the indices.
8074 * optabs.c (expand_vec_perm_var): Likewise.
8076 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8078 * Makefile.in (OBJS): Add vec-perm-indices.o.
8079 * vec-perm-indices.h: New file.
8080 * vec-perm-indices.c: Likewise.
8081 * target.h (vec_perm_indices): Replace with a forward class
8083 (auto_vec_perm_indices): Move to vec-perm-indices.h.
8084 * optabs.h: Include vec-perm-indices.h.
8085 (expand_vec_perm): Delete.
8086 (selector_fits_mode_p, expand_vec_perm_var): Declare.
8087 (expand_vec_perm_const): Declare.
8088 * target.def (vec_perm_const_ok): Replace with...
8089 (vec_perm_const): ...this new hook.
8090 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
8091 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
8092 * doc/tm.texi: Regenerate.
8093 * optabs.def (vec_perm_const): Delete.
8094 * doc/md.texi (vec_perm_const): Likewise.
8095 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
8096 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
8097 expand_vec_perm for constant permutation vectors. Assert that
8098 the mode of variable permutation vectors is the integer equivalent
8099 of the mode that is being permuted.
8100 * optabs-query.h (selector_fits_mode_p): Declare.
8101 * optabs-query.c: Include vec-perm-indices.h.
8102 (selector_fits_mode_p): New function.
8103 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
8104 is defined, instead of checking whether the vec_perm_const_optab
8105 exists. Use targetm.vectorize.vec_perm_const instead of
8106 targetm.vectorize.vec_perm_const_ok. Check whether the indices
8107 fit in the vector mode before using a variable permute.
8108 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
8109 vec_perm_indices instead of an rtx.
8110 (expand_vec_perm): Replace with...
8111 (expand_vec_perm_const): ...this new function. Take the selector
8112 as a vec_perm_indices rather than an rtx. Also take the mode of
8113 the selector. Update call to shift_amt_for_vec_perm_mask.
8114 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
8115 Use vec_perm_indices::new_expanded_vector to expand the original
8116 selector into bytes. Check whether the indices fit in the vector
8117 mode before using a variable permute.
8118 (expand_vec_perm_var): Make global.
8119 (expand_mult_highpart): Use expand_vec_perm_const.
8120 * fold-const.c: Includes vec-perm-indices.h.
8121 * tree-ssa-forwprop.c: Likewise.
8122 * tree-vect-data-refs.c: Likewise.
8123 * tree-vect-generic.c: Likewise.
8124 * tree-vect-loop.c: Likewise.
8125 * tree-vect-slp.c: Likewise.
8126 * tree-vect-stmts.c: Likewise.
8127 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
8129 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
8130 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
8131 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
8132 (aarch64_vectorize_vec_perm_const): ...this new function.
8133 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8134 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8135 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
8136 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
8137 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8138 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8139 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
8141 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
8142 check for NEON modes.
8143 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
8144 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
8145 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
8146 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
8148 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
8149 the old VEC_PERM_CONST conditions.
8150 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
8151 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
8152 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
8153 (ia64_vectorize_vec_perm_const_ok): Merge into...
8154 (ia64_vectorize_vec_perm_const): ...this new function.
8155 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
8156 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
8157 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
8158 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
8159 * config/mips/mips.c (mips_expand_vec_perm_const)
8160 (mips_vectorize_vec_perm_const_ok): Merge into...
8161 (mips_vectorize_vec_perm_const): ...this new function.
8162 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
8163 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
8164 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
8165 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
8166 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
8167 (rs6000_expand_vec_perm_const): Delete.
8168 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
8170 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8171 (altivec_expand_vec_perm_const_le): Take each operand individually.
8172 Operate on constant selectors rather than rtxes.
8173 (altivec_expand_vec_perm_const): Likewise. Update call to
8174 altivec_expand_vec_perm_const_le.
8175 (rs6000_expand_vec_perm_const): Delete.
8176 (rs6000_vectorize_vec_perm_const_ok): Delete.
8177 (rs6000_vectorize_vec_perm_const): New function.
8178 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8179 an element count and rtx array.
8180 (rs6000_expand_extract_even): Update call accordingly.
8181 (rs6000_expand_interleave): Likewise.
8182 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
8183 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
8184 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
8185 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
8186 (rs6000_expand_vec_perm_const): Delete.
8187 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8188 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8189 (altivec_expand_vec_perm_const_le): Take each operand individually.
8190 Operate on constant selectors rather than rtxes.
8191 (altivec_expand_vec_perm_const): Likewise. Update call to
8192 altivec_expand_vec_perm_const_le.
8193 (rs6000_expand_vec_perm_const): Delete.
8194 (rs6000_vectorize_vec_perm_const_ok): Delete.
8195 (rs6000_vectorize_vec_perm_const): New function. Remove stray
8196 reference to the SPE evmerge intructions.
8197 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8198 an element count and rtx array.
8199 (rs6000_expand_extract_even): Update call accordingly.
8200 (rs6000_expand_interleave): Likewise.
8201 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
8202 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
8204 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8206 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8208 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
8209 vector mode and that that mode matches the mode of the data
8211 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
8212 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
8213 directly using expand_vec_perm_1 when forcing selectors into
8215 (expand_vec_perm_var): New function, split out from expand_vec_perm.
8217 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8219 * optabs-query.h (can_vec_perm_p): Delete.
8220 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
8221 * optabs-query.c (can_vec_perm_p): Split into...
8222 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
8223 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
8224 particular selector is valid.
8225 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8226 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8227 (vect_grouped_load_supported): Likewise.
8228 (vect_shift_permute_load_chain): Likewise.
8229 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8230 (vect_transform_slp_perm_load): Likewise.
8231 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8232 (vectorizable_bswap): Likewise.
8233 (vect_gen_perm_mask_checked): Likewise.
8234 * fold-const.c (fold_ternary_loc): Likewise. Don't take
8235 implementations of variable permutation vectors into account
8236 when deciding which selector to use.
8237 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
8238 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
8239 with a false third argument.
8240 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
8241 to test whether the constant selector is valid and can_vec_perm_var_p
8242 to test whether a variable selector is valid.
8244 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8246 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
8247 * optabs-query.c (can_vec_perm_p): Likewise.
8248 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
8249 instead of vec_perm_indices.
8250 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
8251 (vect_gen_perm_mask_checked): Likewise,
8252 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
8253 (vect_gen_perm_mask_checked): Likewise,
8255 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8257 * optabs-query.h (qimode_for_vec_perm): Declare.
8258 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
8259 (qimode_for_vec_perm): ...this new function.
8260 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
8262 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8264 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
8265 does not have a conditional at the top.
8267 2018-01-02 Richard Biener <rguenther@suse.de>
8269 * ipa-inline.c (big_speedup_p): Fix expression.
8271 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8274 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
8277 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8281 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
8282 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
8283 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
8284 cond_taken_branch_cost 3->4.
8286 2018-01-01 Jakub Jelinek <jakub@redhat.com>
8288 PR tree-optimization/83581
8289 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
8290 TODO_cleanup_cfg if any changes have been made.
8293 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
8294 convert_modes if target mode has the right side, but different mode
8298 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
8299 last argument when extracting from CONCAT. If either from_real or
8300 from_imag is NULL, use expansion through memory. If result is not
8301 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
8302 the parts directly to inner mode, if even that fails, use expansion
8306 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
8307 check for bswap in mode rather than HImode and use that in expand_unop
8310 Copyright (C) 2018 Free Software Foundation, Inc.
8312 Copying and distribution of this file, with or without modification,
8313 are permitted in any medium without royalty provided the copyright
8314 notice and this notice are preserved.