1 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
5 * dwarf2out.c (struct dw_line_info_table): Update comments for
7 (FORCE_RESET_NEXT_VIEW): New.
8 (FORCE_RESETTING_VIEW_P): New.
9 (RESETTING_VIEW_P): Check for -1 too.
10 (ZERO_VIEW_P): Likewise.
11 (new_line_info_table): Force-reset next view.
12 (dwarf2out_begin_function): Likewise.
13 (dwarf2out_source_line): Simplify zero_view_p initialization.
14 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
15 view directly. Omit view when omitting .loc at line 0.
17 2018-03-08 Jakub Jelinek <jakub@redhat.com>
19 PR tree-optimization/84740
20 * tree-switch-conversion.c (process_switch): Call build_constructors
21 only if info.phi_count is non-zero.
23 PR tree-optimization/84739
24 * tree-tailcall.c (find_tail_calls): Check call arguments against
25 DECL_ARGUMENTS (current_function_decl) rather than
26 DECL_ARGUMENTS (func) when checking for tail recursion.
28 2018-03-07 Jakub Jelinek <jakub@redhat.com>
30 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
31 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
32 Volker Reichelt's entry and add entries for people that perform
33 GCC fuzzy testing and report numerous bugs.
35 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
38 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
39 readonly data in sdata, if that is disabled.
40 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
41 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
42 -mreadonly-in-sdata option.
44 2018-03-07 Martin Sebor <msebor@redhat.com>
46 PR tree-optimization/84468
47 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
48 basic block when looking for nul assignment.
50 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
53 * except.h (output_function_exception_table): Adjust prototype.
54 * except.c (output_function_exception_table): Remove FNNAME parameter
55 and add SECTION parameter. Ouput one part of the table at a time.
56 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
57 the first part of the exception table and emit unwind directives.
58 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
59 (i386_pe_seh_cold_init): Likewise.
60 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
61 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
62 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
63 (ix86_output_call_insn): Emit a nop in one more case for SEH.
64 * config/i386/winnt.c: Include except.h.
65 (struct seh_frame_state): Add reg_offset, after_prologue and
66 in_cold_section fields.
67 (i386_pe_seh_end_prologue): Set seh->after_prologue.
68 (i386_pe_seh_cold_init): New function.
69 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
70 to seh->in_cold_section.
71 (seh_emit_push): Record the offset of the push.
72 (seh_emit_save): Record the offet of the save.
73 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
74 Test seh->after_prologue to disregard the epilogue.
75 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
76 (i386_pe_end_cold_function): New function.
78 2018-03-07 Jakub Jelinek <jakub@redhat.com>
81 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
82 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
85 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
87 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
88 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
91 * multiple_target.c: Include tree-inline.h and intl.h.
92 (expand_target_clones): Diagnose and fail if node->definition and
93 !tree_versionable_function_p (node->decl).
95 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
97 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
99 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
100 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
101 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
103 2018-03-06 Jakub Jelinek <jakub@redhat.com>
106 * combine.c (try_combine): Use reg_or_subregno instead of handling
107 just paradoxical SUBREGs and REGs.
109 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
111 * config/arc/arc.c (arc_finalize_pic): Remove function.
112 (arc_must_save_register): We use single base PIC register, remove
113 checks to save/restore the PIC register.
114 (arc_expand_prologue): Likewise.
115 * config/arc/arc-protos.h (arc_set_default_type_attributes):
117 (arc_verify_short): Likewise.
118 (arc_attr_type): Likewise.
119 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
120 (walk_stores): Likewise.
121 (arc_address_cost): Make it static.
122 (arc_verify_short): Likewise.
123 (branch_dest): Likewise.
124 (arc_attr_type): Likewise.
125 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
126 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
127 (arc_final_prescan_insn): Remove inserting the nops due to
128 hardware hazards. It is done in reorg step.
129 (insn_length_variant_t): Remove.
130 (insn_length_parameters_t): Likewise.
131 (arc_insn_length_parameters): Likewise.
132 (arc_get_insn_variants): Likewise.
133 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
135 2018-03-06 Jakub Jelinek <jakub@redhat.com>
138 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
141 PR tree-optimization/84687
142 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
144 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
146 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
148 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
149 Rename to ppc_speculation_barrier.
150 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
151 __builtin_ppc_speculation_barrier.
153 2018-03-05 Jakub Jelinek <jakub@redhat.com>
156 * combine.c (combine_simplify_rtx): Don't try to simplify if
157 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
160 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
162 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
163 to 32 bytes when compiling for POWER9.
165 2018-03-05 Jakub Jelinek <jakub@redhat.com>
168 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
169 regparm >= 3 with no arg reg available also for calls with
170 flag_force_indirect_call. Pass decl to ix86_function_regparm.
173 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
175 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
177 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
180 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
182 2018-03-05 Richard Biener <rguenther@suse.de>
184 PR tree-optimization/84486
185 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
186 When inserting a __builtin_assume_aligned call set the LHS
187 SSA name alignment info accordingly.
189 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
191 PR tree-optimization/84114
192 * config/aarch64/aarch64.c (aarch64_reassociation_width)
193 Avoid reassociation of FLOAT_MODE addition.
195 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
197 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
198 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
199 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
200 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
201 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
202 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
203 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
205 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
206 __builtin_ia32_wbinvd): New builtins.
207 (SPECIAL_ARGS2): New.
208 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
209 (SPECIAL_ARGS2): New.
210 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
211 (ix86_valid_target_attribute_inner_p): Ditto.
212 (ix86_init_mmx_sse_builtins): Add special_args2.
213 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
214 TARGET_WBNOINVD_P): New.
215 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
216 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
217 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
218 * config/i386/immintrin.h (_wbinvd): New intrinsic.
219 * config/i386/pconfigintrin.h: New file.
220 * config/i386/wbnoinvdintrin.h: Ditto.
221 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
222 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
224 2018-03-05 Richard Biener <rguenther@suse.de>
226 PR tree-optimization/84670
227 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
229 (BB_VISITED_WITH_VISITED_SUCCS): New define.
230 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
231 (compute_antic_aux): Only assert the number of values in ANTIC_IN
232 doesn't grow if all successors (recursively) were visited at least
235 2018-03-05 Richard Biener <rguenther@suse.de>
237 PR tree-optimization/84650
238 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
239 if executed in the loop pipeline.
241 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
243 * doc/configfiles.texi (Configuration Files): Move info about
244 conditionalizing $target-protos.h to...
245 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
246 differs from $target-protos.h.
248 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
249 Chung-Ju Wu <jasonwucj@gmail.com>
251 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
252 * config/nds32/nds32-multiple.md (setmemsi): Define.
253 * config/nds32/nds32-memory-manipulation.c
254 (nds32_gen_dup_4_byte_to_word_value): New.
255 (emit_setmem_word_loop): New.
256 (emit_setmem_byte_loop): New.
257 (nds32_expand_setmem_loop): New.
258 (nds32_expand_setmem_loop_v3m): New.
259 (nds32_expand_setmem_unroll): New.
260 (nds32_expand_setmem): New.
262 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
263 Chung-Ju Wu <jasonwucj@gmail.com>
265 * config/nds32/nds32-memory-manipulation.c
266 (nds32_emit_load_store): New.
267 (nds32_emit_post_inc_load_store): New.
268 (nds32_emit_mem_move): New.
269 (nds32_emit_mem_move_block): New.
270 (nds32_expand_movmemsi_loop_unknown_size): New.
271 (nds32_expand_movmemsi_loop_known_size): New.
272 (nds32_expand_movmemsi_loop): New.
273 (nds32_expand_movmemsi_unroll): New.
274 (nds32_expand_movmemqi): Rename ...
275 (nds32_expand_movmemsi): ... to this.
276 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
277 (movmemsi): ... to this.
278 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
279 (nds32_expand_movmemsi): ... to this.
281 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
282 Monk Chiang <sh.chiang04@gmail.com>
283 Chung-Ju Wu <jasonwucj@gmail.com>
285 * config/nds32/nds32-protos.h
286 (nds32_expand_load_multiple): New arguments.
287 (nds32_expand_store_multiple): Ditto.
288 (nds32_valid_multiple_load_store): Rename ...
289 (nds32_valid_multiple_load_store_p): ... to this.
290 * config/nds32/nds32-memory-manipulation.c
291 (nds32_expand_load_multiple): Refine implementation.
292 (nds32_expand_store_multiple): Ditto.
293 * config/nds32/nds32-multiple.md
294 (load_multiple): Update nds32_expand_load_multiple interface.
295 (store_multiple): Update nds32_expand_store_multiple interface.
296 * config/nds32/nds32-predicates.c
297 (nds32_valid_multiple_load_store): Rename ...
298 (nds32_valid_multiple_load_store_p): ... to this and refine
300 * config/nds32/predicates.md
301 (nds32_load_multiple_and_update_address_operation): New predicate.
302 (nds32_store_multiple_and_update_address_operation): New predicate.
304 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
305 Chung-Ju Wu <jasonwucj@gmail.com>
307 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
308 (combo): New attribute.
309 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
311 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
313 * config/nds32/nds32.opt: Change -mcmodel= default value.
315 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
316 Monk Chiang <sh.chiang04@gmail.com>
317 Chung-Ju Wu <jasonwucj@gmail.com>
319 * config/nds32/constants.md (unspec_element): New enum.
320 * config/nds32/constraints.md (Umw): New constraint.
321 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
322 * config/nds32/nds32-intrinsic.md: Likewise.
323 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
324 (nds32_valid_smw_lwm_base_p): New.
325 (nds32_output_smw_single_word): New.
326 (nds32_output_lmw_single_word): New.
327 (nds32_expand_unaligned_load): New.
328 (nds32_expand_unaligned_store): New.
329 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
330 (nds32_output_smw_single_word): Declare.
331 (nds32_output_lmw_single_word): Declare.
332 (nds32_expand_unaligned_load): Declare.
333 (nds32_expand_unaligned_store): Declare.
334 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
335 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
336 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
337 NDS32_BUILTIN_UASTORE_DW.
338 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
341 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
342 Kito Cheng <kito.cheng@gmail.com>
343 Chung-Ju Wu <jasonwucj@gmail.com>
345 * config/nds32/nds32-intrinsic.c
346 (nds32_expand_builtin_null_ftype_reg): Delete.
347 (nds32_expand_builtin_reg_ftype_imm): Ditto.
348 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
349 (nds32_read_argument): New.
350 (nds32_legitimize_target): Ditto.
351 (nds32_legitimize_argument): Ditto.
352 (nds32_check_constant_argument): Ditto.
353 (nds32_expand_unop_builtin): Ditto.
354 (nds32_expand_unopimm_builtin): Ditto.
355 (nds32_expand_binop_builtin): Ditto.
356 (nds32_builtin_decl_impl): Ditto.
357 (builtin_description): Ditto.
358 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
359 (nds32_init_builtins_impl): Ditto.
360 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
361 (nds32_builtin_decl): New.
362 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
363 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
365 2018-03-02 Jeff Law <law@redhat.com>
367 * reorg.c (stop_search_p): Handle DEBUG_INSN.
368 (redundant_insn, fill_simple_delay_slots): Likewise.
369 (fill_slots_from_thread): Likewise.
370 * resource.c (mark_referenced_resources): Likewise.
371 (mark_set_resources, find_dead_or_set_registers): Likewise.
373 2018-03-02 Jakub Jelinek <jakub@redhat.com>
375 * substring-locations.h (format_warning_va): Formatting fix for
377 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
379 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
380 * substring-locations.c: Include intl.h.
381 (format_warning_va): Turned into small wrapper around
382 format_warning_n_va, renamed to ...
383 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
384 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
386 (format_warning_at_substring_n): New function.
387 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
388 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
389 format_warning_at_substring with just a shorter name instead of
390 const function pointer.
391 (fmtwarn_n): New function.
392 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
393 appropriate, get rid of all the fmtstr temporaries, move conditionals
394 with G_() wrapped string literals directly into fmtwarn arguments,
395 cast dir.len to (int), formatting fixes.
397 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
399 * doc/invoke.texi: Remove "Cilk Plus" references.
401 2018-03-02 Jakub Jelinek <jakub@redhat.com>
402 Richard Biener <rguenther@suse.de>
405 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
406 for error or warning attributes if CALL_FROM_THUNK_P is set.
409 2018-03-02 Jakub Jelinek <jakub@redhat.com>
412 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
413 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
416 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
417 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
419 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
420 instead of -1U in last predictors element's probability member.
422 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
425 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
426 arguments if they are comparable.
428 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
430 PR tree-optimization/84634
431 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
432 masks and masked_loop_p with a single loop_masks, making sure it's
433 null for bb vectorization.
435 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
437 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
438 (vect_analyze_data_ref_access): Use loop->safe_len rather than
439 loop->force_vectorize to check whether there is no alias.
441 2018-03-02 Jakub Jelinek <jakub@redhat.com>
444 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
446 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
448 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
449 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
450 instead of a loop around prev_real_insn.
451 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
455 * config/i386/i386.c (ix86_print_operand): Use conditional
456 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
459 2018-03-02 Richard Biener <rguenther@suse.de>
461 PR tree-optimization/84427
462 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
463 (bitmap_set_subtract_values): Rewrite to handle multiple
466 (prune_clobbered_mems): Likewise.
467 (phi_translate): Take edge instead of pred/phiblock.
468 (phi_translate_1): Likewise.
469 (phi_translate_set): Likewise. Insert all translated
470 exprs for a value into the set, keeping possibly multiple
471 expressions per value.
472 (compute_antic_aux): Adjust for phi_translate changes.
473 When intersecting union the expressions and prune those
474 not in the final value set, keeping possibly multiple
475 expressions per value. Do not use value-insertion
476 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
477 all expressions. Add verification that the value-sets
478 only shrink during iteration.
479 (compute_partial_antic_aux): Adjust for the phi_translate changes.
480 (do_pre_regular_insertion): Likewise.
481 (do_pre_partial_partial_insertion): Likewise.
483 2018-03-02 Richard Biener <rguenther@suse.de>
486 * config/darwin.c (saved_debug_info_level): New static global.
487 (darwin_asm_lto_start): Disable debug info generation for LTO out.
488 (darwin_asm_lto_end): Restore debug info generation settings.
490 2018-03-01 Martin Liska <mliska@suse.cz>
493 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
496 2018-03-01 Richard Biener <rguenther@suse.de>
499 * dwarf2out.c (gen_variable_die): Properly handle late VLA
500 type annotation with LTO when debug was disabled at compile-time.
502 2018-03-01 Matthew Fortune <mfortune@gmail.com>
504 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
506 (mips_final_postscan_insn): Likewise.
508 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
510 PR rtl-optimization/84528
511 * alias.c (init_alias_target): Add commentary.
512 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
513 a unique base value if the frame pointer is not eliminated
514 to the stack pointer.
516 2018-03-01 Tom de Vries <tom@codesourcery.com>
518 PR rtl-optimization/83327
519 * lra-int.h (hard_regs_spilled_into): Declare.
520 * lra.c (hard_regs_spilled_into): Define.
521 (init_reg_info): Init hard_regs_spilled_into.
522 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
523 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
524 (process_bb_lives): Handle hard_regs_spilled_into.
525 (lra_create_live_ranges_1): Before doing liveness propagation, clear
526 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
528 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
530 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
531 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
532 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
533 * config/rs6000/aix72.h: New file.
535 2018-02-28 Jakub Jelinek <jakub@redhat.com>
537 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
538 instead of warning_at with conditional singular and plural messages
542 * stor-layout.c (update_alignment_for_field): For
543 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
544 && !DECL_PACKED (field), do the alignment update, just use
545 only desired_align instead of MAX (type_align, desired_align)
547 (place_field): Don't do known_align < desired_align handling
548 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
549 is non-NULL, instead do it after rli->prev_field handling and
550 only if not within a bitfield word. For DECL_PACKED (field)
551 use type_align of BITS_PER_UNIT.
553 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
555 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
556 superfluous parentheses and trailing spaces.
558 2018-02-28 Richard Biener <rguenther@suse.de>
560 PR tree-optimization/84584
561 * graphite-scop-detection.c (scop_detection::add_scop): Discard
562 SCoPs with fake exit edge.
564 2018-02-28 Martin Liska <mliska@suse.cz>
567 * timevar.c (timer::print): Fix format to properly print 100%
570 2018-02-28 Richard Biener <rguenther@suse.de>
573 * genmatch.c (capture_info::walk_match): Do not mark
574 captured expressions without operands as expr_p given
575 they act more like predicates and should be subject to
576 "lost tail" side-effect preserving.
578 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
580 PR rtl-optimization/81611
581 * auto-inc-dec.c (attempt_change): Move dead note from
582 mem_insn if it's the next use of regno
583 (find_address): Take address use of reg holding
584 non-incremented value. Add parm to limit search to the named
586 (merge_in_block): Attempt to use a mem insn that is the next
587 use of the original regno.
589 2018-02-27 Martin Sebor <msebor@redhat.com>
592 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
593 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
595 2018-02-27 Martin Sebor <msebor@redhat.com>
598 * diagnostic-core.h (warning_n, error_n, inform_n): Change
599 n argument to unsigned HOST_WIDE_INT.
600 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
601 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
602 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
603 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
605 2018-02-27 Richard Biener <rguenther@suse.de>
607 PR tree-optimization/84512
608 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
609 Do not use the estimate returned from record_stmt_cost for
610 the scalar iteration cost but sum properly using add_stmt_cost.
612 2018-02-27 Richard Biener <rguenther@suse.de>
614 PR tree-optimization/84466
615 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
616 Adjust last change to less strictly validate use operands.
618 2018-02-27 Martin Liska <mliska@suse.cz>
620 PR gcov-profile/84548
621 * gcov.c (process_file): Allow partial overlap and consider it
622 also as group functions.
623 (output_lines): Properly calculate range of lines for a group.
625 2018-02-27 Martin Liska <mliska@suse.cz>
627 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
628 'ggc' suffixes. Change first column width.
629 (timer::print): Fix formatting of the column.
631 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
633 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
634 preserve inline entry blocks for the sake of debug inline
635 entry point markers alone.
636 (remove_unused_locals): Suggest in comments a better place to
637 force the preservation of inline entry blocks that are
638 otherwise unused, but do not preserve them.
640 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
642 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
644 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
647 * config/i386/constraints.md (Bs): Replace
648 ix86_indirect_branch_register with
649 TARGET_INDIRECT_BRANCH_REGISTER.
651 * config/i386/i386.md (indirect_jump): Likewise.
652 (tablejump): Likewise.
653 (*sibcall_memory): Likewise.
654 (*sibcall_value_memory): Likewise.
655 Peepholes of indirect call and jump via memory: Likewise.
656 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
657 (*sibcall_value_GOT_32): Likewise.
658 * config/i386/predicates.md (indirect_branch_operand): Likewise.
659 (GOT_memory_operand): Likewise.
660 (call_insn_operand): Likewise.
661 (sibcall_insn_operand): Likewise.
662 (GOT32_symbol_operand): Likewise.
663 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
665 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
667 PR rtl-optimization/83496
668 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
669 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
670 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
671 redundant insn, if any.
672 (relax_delay_slots): Likewise.
673 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
675 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
677 PR tree-optimization/83965
678 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
679 that grouped statements are part of a reduction chain. Return
680 true if the statement is not marked as a reduction itself but
682 (vect_recog_dot_prod_pattern): Don't check whether the statement
683 is part of a group here.
684 (vect_recog_sad_pattern): Likewise.
685 (vect_recog_widen_sum_pattern): Likewise.
687 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
690 * final.c (rest_of_clean_state): Also look for calls inside sequences.
692 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
695 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
697 (ix86_output_indirect_function_return): New prototype.
698 (ix86_split_simple_return_pop_internal): Likewise.
699 * config/i386/i386.c (indirect_return_via_cx): New.
700 (indirect_return_via_cx_bnd): Likewise.
701 (indirect_thunk_name): Handle return va CX_REG.
702 (output_indirect_thunk_function): Create alias for
703 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
704 (ix86_output_indirect_jmp): Remove the bool argument.
705 (ix86_output_indirect_function_return): New function.
706 (ix86_split_simple_return_pop_internal): Likewise.
707 * config/i386/i386.md (*indirect_jump): Don't pass false
708 to ix86_output_indirect_jmp.
709 (*tablejump_1): Likewise.
710 (simple_return_pop_internal): Change it to define_insn_and_split.
711 Call ix86_split_simple_return_pop_internal to split it for
713 (simple_return_indirect_internal): Call
714 ix86_output_indirect_function_return instead of
715 ix86_output_indirect_jmp.
717 2018-02-26 Jakub Jelinek <jakub@redhat.com>
720 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
721 memset and value initialization afterwards.
723 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
725 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
727 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
730 * common/config/aarch64/aarch64-common.c
731 (aarch_option_optimization_table[]): Switch
732 off fomit-frame-pointer
734 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
735 Chung-Ju Wu <jasonwucj@gmail.com>
737 * config/nds32/nds32-multiple.md (load_multiple): Disallow
739 (store_multiple): Ditto.
741 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
743 * config.gcc: Add --with-cpu support for nds32 target.
744 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
745 * config/nds32/nds32.opt: Add -mcpu= option.
747 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
749 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
750 isel=yes): Warn for these deprecated options.
752 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
754 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
755 ISA_2_5_MASKS_EMBEDDED.
757 2018-02-23 Jakub Jelinek <jakub@redhat.com>
759 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
760 p->max as pointers rather than using iterative_hash_expr.
762 2018-02-23 Carl Love <cel@us.ibm.com>
764 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
765 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
767 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
768 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
769 P8V_BUILTIN_VEC_VUNSIGNED2.
771 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
774 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
775 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
777 * lra-constraints.c (curr_insn_transform): Set up
778 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
780 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
781 * lra-eliminations.c (spill_pseudos): Ditto.
782 (process_insn_for_elimination): Ditto.
783 * lra-lives.c (reg_early_clobber_p): Use the new macros.
784 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
785 LRA_NON_CLOBBERED_ALT.
787 2018-02-22 Martin Sebor <msebor@redhat.com>
789 PR tree-optimization/84480
790 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
791 to maybe_diag_stxncpy_trunc. Call it.
792 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
793 from gimple_fold_builtin_strcpy. Print inlining stack.
794 (handle_builtin_stxncpy): Print inlining stack.
795 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
797 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
800 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
801 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
802 and -fcheck-pointer-bounds are used together.
803 (indirect_thunk_prefix): New enum.
804 (indirect_thunk_need_prefix): New function.
805 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
806 "_nt" instead of "_bnd" for NOTRACK prefix.
807 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
808 (output_indirect_thunk_function): Likewise.
810 (ix86_code_end): Update output_indirect_thunk_function calls.
811 (ix86_output_indirect_branch_via_reg): Replace
812 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
813 (ix86_output_indirect_branch_via_push): Likewise.
814 (ix86_output_function_return): Likewise.
815 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
816 incompatible with -fcf-protection=branch and
817 -fcheck-pointer-bounds.
819 2018-02-22 Steve Ellcey <sellcey@cavium.com>
822 * config/aarch64/aarch64.c (aarch64_print_address_internal):
823 Change gcc_assert call to output_operand_lossage.
825 2018-02-22 Steve Ellcey <sellcey@cavium.com>
827 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
829 2018-02-22 DJ Delorie <dj@redhat.com>
830 Sebastian Perta <sebastian.perta@renesas.com>
831 Oleg Endo <olegendo@gcc.gnu.org>
833 * config/rx/rx.c (rx_rtx_costs): New function.
834 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
836 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
838 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
840 2018-02-22 Martin Liska <mliska@suse.cz>
843 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
844 Add "native" as a possible value.
846 2018-02-22 Martin Liska <mliska@suse.cz>
849 * config/i386/i386.c (ix86_option_override_internal):
850 Add "native" as a possible value for -march and -mtune.
852 2018-02-22 Jakub Jelinek <jakub@redhat.com>
855 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
856 to all type variants.
858 PR tree-optimization/84503
859 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
860 width as info->bitpos + info->bitsize - start.
861 (merged_store_group::merge_overlapping): Simplify width computation.
862 (check_no_overlap): New function.
863 (imm_store_chain_info::try_coalesce_bswap): Compute expected
864 start + width and last_order of the group, fail if check_no_overlap
866 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
867 to group if check_no_overlap fails.
869 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
871 * config/rs6000/altivec.md: Delete contraint arguments to
872 define_expand, define_split, and define_peephole2, and in
873 define_insn_and_split if always unused.
874 * config/rs6000/darwin.md: Ditto.
875 * config/rs6000/dfp.md: Ditto.
876 * config/rs6000/rs6000.md: Ditto.
877 * config/rs6000/sync.md: Ditto.
878 * config/rs6000/vector.md: Ditto.
879 * config/rs6000/vsx.md: Ditto.
881 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
883 * config/rs6000/altivec.md: Write output control strings as braced
884 blocks instead of double-quoted strings.
885 * config/rs6000/darwin.md: Ditto.
886 * config/rs6000/rs6000.md: Ditto.
887 * config/rs6000/vector.md: Ditto.
888 * config/rs6000/vsx.md: Ditto.
890 2018-02-21 Jason Merrill <jason@redhat.com>
892 PR c++/84314 - ICE with templates and fastcall attribute.
893 * attribs.c (build_type_attribute_qual_variant): Remove assert.
895 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
897 * ipa-cp.c (determine_versionability): Fix comment typos.
899 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
902 * ipa-cp.c (determine_versionability): Do not version functions caling
905 2018-02-21 Martin Liska <mliska@suse.cz>
908 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
909 Add "native" as a possible value.
910 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
911 the macro when native cpu detection is available.
913 2018-02-21 Martin Liska <mliska@suse.cz>
916 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
917 Add "native" as a possible value.
918 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
919 when native cpu detection is available.
921 2018-02-21 Jakub Jelinek <jakub@redhat.com>
922 Martin Sebor <msebor@redhat.com>
924 PR tree-optimization/84478
925 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
927 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
928 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
929 support which is conservatively correct, for 2 only stay conservative
930 for maxlen. Formatting and comment capitalization fixes. Add STRICT
931 argument to the 2 argument get_range_strlen, adjust 6 arg
932 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
934 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
935 (gimple_fold_builtin_strlen): Pass true as last argument to
938 2018-02-20 Martin Sebor <msebor@redhat.com>
941 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
942 (builtin_memref::set_base_and_offset): Same. Handle inner references.
943 (builtin_memref::builtin_memref): Factor out parts into
944 set_base_and_offset and call it.
946 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
949 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
950 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
951 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
952 search at the associated MODE_INT.
954 2018-02-20 Jeff Law <law@redhat.com>
957 PR tree-optimization/81592
959 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
960 for range data rather than using global data.
961 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
962 range data rather than using global data.
963 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
964 pass it to children as needed.
965 (struct directive::fmtresult): Similarly.
966 (struct directive::set_width): Similarly.
967 (struct directive::set_precision): Similarly.
968 (format_integer, format_directive, parse_directive): Similarly.
969 (format_none): Accept unnamed vr_values parameter.
970 (format_percent, format_floating, format_character): Similarly.
971 (format_string, format_plain): Similarly.
972 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
973 the EVRP range analyzer for range data rather than using global data.
974 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
975 gimple-ssa-evrp-analyze.h
976 (class sprintf_dom_walker): Add after_dom_children member function.
977 Add evrp_range_analyzer member.
978 (sprintf_dom_walker::before_dom_children): Call into the EVRP
979 range analyzer as needed.
980 (sprintf_dom_walker::after_dom_children): New member function.
981 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
983 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
984 (evrp_range_analyzer::pop_to_marker): Likewise.
986 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
988 PR tree-optimization/84419
989 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
990 with the required type if its current type is compatible but
993 2018-02-20 Jakub Jelinek <jakub@redhat.com>
996 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
999 2018-02-20 Martin Liska <mliska@suse.cz>
1002 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
1003 possible values if we don't have a hint.
1005 2018-02-20 Martin Liska <mliska@suse.cz>
1009 * final.c (shorten_branches): Build align_tab array with one
1011 * opts.c (finish_options): Add alignment option limit check.
1012 (MAX_CODE_ALIGN): Likewise.
1013 (MAX_CODE_ALIGN_VALUE): Likewise.
1014 * doc/invoke.texi: Document maximum allowed option value for
1015 all -falign-* options.
1017 2018-02-19 Jakub Jelinek <jakub@redhat.com>
1020 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
1021 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
1022 * var-tracking.c (emit_note_insn_var_location): Remove all references
1023 to NOTE_INSN_CALL_ARG_LOCATION.
1024 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
1025 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
1026 Use copy_rtx_if_shared.
1027 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
1028 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
1029 (dwarf2out_var_location): Remove handling of
1030 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
1032 * final.c (final_scan_insn): Remove all references to
1033 NOTE_INSN_CALL_ARG_LOCATION.
1034 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
1035 before dumping final insns.
1036 * except.c (emit_note_eh_region_end): Remove all references to
1037 NOTE_INSN_CALL_ARG_LOCATION.
1038 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
1039 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
1040 * config/arc/arc.c (hwloop_optimize): Likewise.
1041 * config/arm/arm.c (create_fix_barrier): Likewise.
1042 * config/s390/s390.c (s390_chunkify_start): Likewise.
1043 * config/sh/sh.c (find_barrier): Likewise.
1044 * config/i386/i386.c (rest_of_insert_endbranch,
1045 ix86_seh_fixup_eh_fallthru): Likewise.
1046 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
1047 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
1048 * config/frv/frv.c (frv_function_prologue): Likewise.
1049 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
1051 (note_outside_basic_block_p): Remove all references to
1052 NOTE_INSN_CALL_ARG_LOCATION.
1053 * gengtype.c (adjust_field_rtx_def): Likewise.
1054 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
1056 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
1057 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
1060 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
1063 PR tree-optimization/84452
1064 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
1065 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
1068 2018-02-19 Martin Liska <mliska@suse.cz>
1071 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
1073 2018-02-19 Martin Liska <mliska@suse.cz>
1074 Richard Sandiford <richard.sandiford@linaro.org>
1076 PR tree-optimization/82491
1077 * gimple-fold.c (get_base_constructor): Make earlier bail out
1080 2018-02-19 Carl Love <cel@us.ibm.com>
1082 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
1083 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
1085 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
1086 P8V_BUILTIN_VEC_NEG.
1088 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
1090 * config/rl78/rl78.md (movdf): New define expand.
1092 2018-02-19 Martin Liska <mliska@suse.cz>
1095 * doc/invoke.texi: Fix typo.
1096 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
1098 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
1100 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
1101 handle rs6000_single_float and rs6000_double_float specially for
1104 2018-02-16 Jeff Law <law@redhat.com>
1106 * config/rx/rx.c (add_pop_cfi_notes): New function.;
1109 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1112 * ipa-inline.c (inline_small_functions): Fix a typo.
1114 2018-02-16 Nathan Sidwell <nathan@acm.org>
1116 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
1118 2018-02-16 Carl Love <cel@us.ibm.com>
1120 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
1121 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
1122 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
1123 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
1124 expansion to P8V_BUILTIN_VEC_FLOAT2.
1126 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
1128 PR rtl-optimization/70023
1129 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
1130 src_regno into account.
1132 2018-02-16 Carl Love <cel@us.ibm.com>
1134 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
1135 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
1136 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
1137 * config/rs6000/rs6000.c: Remove case statements for
1138 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
1139 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
1140 and P9V_BUILTIN_VEC_VINSERT4B.
1141 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
1142 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
1143 * config/rs6000/vsx.md:
1144 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
1147 2018-02-16 Carl Love <cel@us.ibm.com>
1149 * config/rs6000/altivec.h: Add builtin names vec_extract4b
1151 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
1153 * config/rs6000/rs6000-c.c: Add the definitions for
1154 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
1155 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
1156 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
1157 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
1158 definition for insert4b and define insn *insert3b_internal.
1159 * doc/extend.texi: Add documentation for vec_extract4b.
1161 2018-02-16 Nathan Sidwell <nathan@acm.org>
1163 * doc/extend.texi (Backwards Compatibility): Mention friend
1164 injection. Note for-scope is deprecated.
1165 * doc/invoke.texi (-ffriend-injection): Deprecate.
1167 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
1169 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
1170 that moved to I2, also allow destinations that are a paradoxical
1171 subreg (instead of a normal reg).
1173 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
1176 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
1179 2018-02-16 Richard Biener <rguenther@suse.de>
1181 PR tree-optimization/84037
1182 PR tree-optimization/84016
1184 * config/i386/i386.c (ix86_builtin_vectorization_cost):
1185 Adjust vec_construct for the fact we need additional higher latency
1186 128bit inserts for AVX256 and AVX512 vector builds.
1187 (ix86_add_stmt_cost): Scale vector construction cost for
1190 2018-02-16 Richard Biener <rguenther@suse.de>
1192 PR tree-optimization/84417
1193 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
1194 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
1195 (non_rewritable_lvalue_p): Likewise, use poly-ints.
1197 2018-02-16 Martin Liska <mliska@suse.cz>
1200 * internal-fn.def (ASAN_CHECK): Set proper flags.
1201 (ASAN_MARK): Likewise.
1203 2018-02-16 Julia Koval <julia.koval@intel.com>
1205 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
1206 from PTA_CANNONLAKE.
1208 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1211 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
1212 Use ++iter rather than iter++ for std::list iterators.
1213 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
1214 defer deleting them until all nodes in the forest are processed. Do
1215 free even leaf nodes. Change to_process into auto_vec.
1218 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
1219 * vec.h (vec_default_construct): Use memset instead of placement new
1220 if BROKEN_VALUE_INITIALIZATION is defined.
1221 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
1222 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
1225 PR rtl-optimization/83723
1226 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
1227 * lra.c (lra_substitute_pseudo): Likewise. If true, use
1228 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
1230 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
1232 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
1234 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
1236 PR rtl-optimization/81443
1237 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
1238 from inner REGs to paradoxical SUBREGs.
1240 2018-02-16 Richard Biener <rguenther@suse.de>
1242 PR tree-optimization/84399
1243 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1244 For operands we can analyze at their definition make sure we can
1245 analyze them at each use as well.
1247 2018-02-16 Richard Biener <rguenther@suse.de>
1249 PR tree-optimization/84190
1250 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
1251 volatile accesses if the decl isn't volatile.
1253 2018-02-15 Jason Merrill <jason@redhat.com>
1255 PR c++/84314 - ICE with templates and fastcall attribute.
1256 * attribs.c (build_type_attribute_qual_variant): Don't clobber
1257 TYPE_CANONICAL on an existing type.
1259 2018-02-15 Jakub Jelinek <jakub@redhat.com>
1261 PR tree-optimization/84383
1262 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
1263 dstoff nor call operand_equal_p if dstbase is NULL.
1265 PR tree-optimization/84334
1266 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
1267 also a CONSTANT_CLASS_P, punt.
1269 2018-02-14 Jim Wilson <jimw@sifive.com>
1271 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
1272 first SMALL_OPERAND check. New local min_second_step. Move assert
1273 to where locals are set. Add TARGET_RVC support.
1274 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
1276 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
1278 * doc/invoke.texi: Correct -Wformat-overflow code sample.
1280 2018-02-14 Martin Sebor <msebor@redhat.com>
1282 PR tree-optimization/83698
1283 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
1284 arrays constrain the offset range to their bounds.
1285 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
1286 (builtin_access::overlap): Avoid setting the size of overlap if it's
1288 (maybe_diag_overlap): Also consider arrays when deciding what values
1289 of offsets to include in diagnostics.
1291 2018-02-14 Martin Sebor <msebor@redhat.com>
1294 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
1295 that correspond to the kind of a declaration.
1297 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
1300 * config/pa/pa.md: Load address of PIC label using the linkage table
1301 if the label is nonlocal.
1303 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
1305 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
1306 warning message if user requests -maltivec=be.
1307 * doc/invoke.texi: Document deprecation of -maltivec=be.
1309 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
1312 * config/rs6000/rs6000-c.c: Update definitions for
1313 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
1314 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
1316 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1319 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
1320 add _get_ssp intrinsics. Remove argument from
1321 __builtin_ia32_rdssp[d|q].
1322 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
1323 * config/i386/i386-builtin.def: Remove argument from
1324 __builtin_ia32_rdssp[d|q].
1325 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
1326 ix86_expand_special_args_builtin for _rdssp[d|q].
1327 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
1328 Clear register before usage.
1329 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
1330 Add documentation for new _get_ssp and _inc_ssp intrinsics.
1332 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
1334 PR tree-optimization/84357
1335 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
1336 operand 1 of an ARRAY_REF too.
1338 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
1341 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
1342 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1344 (set_of_reg): New struct.
1345 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
1346 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
1347 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1349 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
1350 Split into bitclr, bitset, bitinvert patterns if appropriate.
1351 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
1352 use rx_fuse_in_memory_bitop.
1353 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
1354 to named insn, correct maximum insn length.
1356 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
1359 * machmode.def: Define a complex mode for PARTIAL_INT.
1360 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
1362 * doc/rtl.texi: Document CSPImode.
1363 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
1365 (msp430_hard_regno_nregs_with_padding): Likewise.
1367 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
1370 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
1372 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
1374 PR rtl-optimization/84169
1375 * combine.c (try_combine): New variable split_i2i3. Set it to true if
1376 we generated a parallel as new i3 and we split that to new i2 and i3
1377 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
1378 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
1379 those to i2, not i1. Partially rewrite this scan code.
1381 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1384 * stor-layout.c (place_field): For variable length fields, adjust
1385 offset_align afterwards not just based on the field's alignment,
1386 but also on the size.
1389 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
1390 of exps and logs in the use_exp2 case.
1392 2018-02-13 Jeff Law <law@redhat.com>
1394 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
1397 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
1400 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
1404 * common.opt (gas-loc-support, gas-locview-support): New.
1405 (ginline-points, ginternal-reset-location-views): New.
1406 * doc/invoke.texi: Document them. Use @itemx where intended.
1407 (gvariable-location-views): Adjust.
1408 * target.def (reset_location_view): New.
1409 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
1410 (TARGET_RESET_LOCATION_VIEW): New.
1411 * doc/tm.texi: Rebuilt.
1412 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
1413 (dwarf2out_default_as_locview_support): New.
1414 (output_asm_line_debug_info): Use option variables.
1415 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
1416 (output_loc_list): Likewise.
1417 (add_high_low_attributes): Check option variables.
1418 Don't output entry view attribute in strict mode.
1419 (gen_inlined_subroutine_die): Check option variables.
1420 (dwarf2out_inline_entry): Likewise.
1421 (init_sections_and_labels): Likewise.
1422 (dwarf2out_early_finish): Likewise.
1423 (maybe_reset_location_view): New, from...
1424 (dwarf2out_var_location): ... here. Call it.
1425 * debug.h (dwarf2out_default_as_loc_support): Declare.
1426 (dwarf2out_default_as_locview_support): Declare.
1427 * hooks.c (hook_int_rtx_insn_0): New.
1428 * hooks.h (hook_int_rtx_insn_0): Declare.
1429 * toplev.c (process_options): Take -gas-loc-support and
1430 -gas-locview-support from dwarf2out. Enable
1431 -gvariable-location-views by default only with locview
1432 assembler support. Enable -ginternal-reset-location-views by
1433 default only if the target defines the corresponding hook.
1434 Enable -ginline-points by default if location views are
1435 enabled; force it disabled if statement frontiers are
1437 * tree-inline.c (expand_call_inline): Check option variables.
1438 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
1440 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
1442 PR tree-optimization/84321
1443 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
1444 handling. Also check whether the anti-range contains any values
1445 that satisfy the mask; switch to a VR_RANGE if not.
1447 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
1450 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
1452 2018-02-13 Martin Jambor <mjambor@suse.cz>
1455 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
1456 of call statements, also set location of a load to a temporary.
1458 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
1460 * config/rl78/rl78.c (add_vector_labels): New function.
1461 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
1462 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
1463 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
1464 which checks that no arguments are passed.
1465 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
1466 * doc/extend.texi: Documentation for the new attribute.
1468 2018-02-13 Andreas Schwab <schwab@suse.de>
1470 * config/riscv/linux.h (CPP_SPEC): Define.
1472 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1475 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
1476 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
1477 OPTION_MASK_ISA_AES as first argument to def_builtin_const
1478 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
1479 instead of OPTION_MASK_ISA_PCLMUL as first argument to
1480 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
1481 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
1482 temporarily for AES and PCLMUL builtins.
1484 PR tree-optimization/84339
1485 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
1486 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
1490 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
1491 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
1492 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
1494 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
1496 * omp-simd-clone.h: New file.
1497 * omp-simd-clone.c: Include omp-simd-clone.h.
1498 (expand_simd_clones): No longer static.
1499 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
1500 cgraph.h and omp-simd-clone.h.
1501 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
1502 (vect_recog_widen_shift_pattern): Formatting fix.
1503 (vect_pattern_recog_1): Don't check optab for calls.
1506 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
1507 operands[2] into a REG before using gen_lowpart on it.
1509 2018-02-12 Jeff Law <law@redhat.com>
1512 * config/sh/sh.c (find_barrier): Consider a sibling call
1515 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
1516 successfully back substituting a reg.
1518 2018-02-12 Richard Biener <rguenther@suse.de>
1520 PR tree-optimization/84037
1521 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
1522 parameter, move visited init to caller.
1523 (vect_slp_analyze_operations): Separate cost from validity
1524 check, initialize visited once for all instances.
1525 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
1527 * tree-vect-stmts.c (vect_model_simple_cost): Make early
1529 (vect_model_promotion_demotion_cost): Likewise.
1530 (vectorizable_bswap): Guard cost modeling with !slp_node
1531 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
1533 (vectorizable_call): Likewise.
1534 (vectorizable_conversion): Likewise.
1535 (vectorizable_assignment): Likewise.
1536 (vectorizable_shift): Likewise.
1537 (vectorizable_operation): Likewise.
1538 (vectorizable_store): Likewise.
1539 (vectorizable_load): Likewise.
1540 (vectorizable_condition): Likewise.
1541 (vectorizable_comparison): Likewise.
1543 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
1546 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
1547 (ASAN_MARK): Fix fnspec to account for return value, change pointer
1548 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
1550 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1553 * params.def (inline-min-speedup): Increase from 8 to 15.
1554 (max-inline-insns-auto): Decrease from 40 to 30.
1555 * ipa-split.c (consider_split): Add some buffer for function to
1556 be considered inlining candidate.
1557 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
1560 2018-02-12 Richard Biener <rguenther@suse.de>
1562 PR tree-optimization/84037
1563 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
1564 matched stmts if we cannot swap the non-matched ones.
1566 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1568 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
1569 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
1570 _mm_maskz_scalef_round_ss): New intrinsics.
1571 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
1572 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
1573 __builtin_ia32_scalefss_round): Remove.
1574 (__builtin_ia32_scalefsd_mask_round,
1575 __builtin_ia32_scalefss_mask_round): New intrinsics.
1576 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
1577 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
1578 ((match_operand:VF_128 2 "<round_nimm_predicate>"
1579 "<round_constraint>")): Changed to ...
1580 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
1581 "<round_scalar_constraint>")): ... this.
1582 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
1583 %0, %1, %2<round_op3>}"): Changed to ...
1584 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
1585 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
1586 %2<round_scalar_mask_op3>}"): ... this.
1587 * config/i386/subst.md (round_scalar_nimm_predicate): New.
1589 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1591 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
1592 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
1593 (_mm_maskz_sqrt_round_ss): New intrinsics.
1594 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
1595 (__builtin_ia32_sqrtsd_mask_round)
1596 (__builtin_ia32_sqrtss_mask_round): New builtins.
1597 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
1598 (__builtin_ia32_sqrtss_round): Remove.
1599 (__builtin_ia32_sqrtsd_mask_round)
1600 (__builtin_ia32_sqrtss_mask_round): New builtins.
1601 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
1602 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
1603 ((match_operand:VF_128 1 "vector_operand"
1604 "xBm,<round_constraint>")): Changed to ...
1605 ((match_operand:VF_128 1 "vector_operand"
1606 "xBm,<round_scalar_constraint>")): ... this.
1607 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
1608 %0, %2, %<iptr>1<round_op3>}): Changed to ...
1609 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
1610 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
1611 %<iptr>1<round_scalar_mask_op3>}): ... this.
1612 ((set_attr "prefix" "<round_prefix>")): Changed to ...
1613 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
1615 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
1618 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
1619 Cast vec_cmpeq result to correct type.
1620 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
1621 Cast vec_cmpgt result to correct type.
1623 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
1625 * final.c (final_scan_insn_1): Renamed from...
1626 (final_scan_insn): ... this. New wrapper, to recover
1627 seen from the outermost call in recursive ones.
1628 * config/sparc/sparc.c (output_return): Drop seen from call.
1629 (output_sibcall): Likewise.
1630 * config/visium/visium.c (output_branch): Likewise.
1632 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
1634 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
1637 2018-02-10 Alan Modra <amodra@gmail.com>
1640 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
1641 Specify LR as an input.
1643 2018-02-10 Jakub Jelinek <jakub@redhat.com>
1646 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
1647 remove_member_access_dummy_vars): New functions.
1648 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
1649 lower_omp_1, execute_lower_omp): Use them.
1651 PR rtl-optimization/84308
1652 * shrink-wrap.c (spread_components): Release todo vector.
1654 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
1656 PR rtl-optimization/57193
1657 * ira-color.c (struct allocno_color_data): Add member
1658 conflict_allocno_hard_prefs.
1659 (update_conflict_allocno_hard_prefs): New.
1660 (bucket_allocno_compare_func): Add a preference based on
1661 conflict_allocno_hard_prefs.
1662 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
1663 (color_allocnos): Remove a dead code. Initiate
1664 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
1666 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1669 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
1670 constraint from =wa to wa. Avoid a subreg on the output operand,
1671 instead use a pseudo and subreg it in a move.
1672 (p9_xxbrd_<mode>): Changed to ...
1673 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
1674 (p9_xxbrd_v2df): New expander.
1675 (p9_xxbrw_<mode>): Changed to ...
1676 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
1677 (p9_xxbrw_v4sf): New expander.
1679 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1681 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
1683 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
1686 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
1687 multiply in 32-bit mode.
1688 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
1689 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
1692 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1694 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
1695 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
1696 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
1697 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
1699 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
1702 * dwarf2out.c (is_trivial_indirect_ref): New function.
1703 (dwarf2out_late_global_decl): Do not generate a location
1704 attribute for variables that have a non-trivial DECL_VALUE_EXPR
1705 and that are not defined in the current unit.
1707 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1709 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
1710 instead of a libcall for UNORDERED.
1712 2018-02-09 Tamar Christina <tamar.christina@arm.com>
1715 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
1716 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
1718 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1721 * config/s390/s390.c (s390_set_current_function): Invoke
1722 s390_indirect_branch_settings also if fndecl didn't change.
1724 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1726 * config/rs6000/rs6000.md (blockage): Set length to zero.
1728 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1730 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
1732 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1735 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
1736 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
1740 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
1741 PARALLEL incoming that failed vt_get_decl_and_offset check.
1744 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
1745 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
1747 (get_variable_section): For decls in named .bss* sections pass true as
1748 second argument to bss_initializer_p.
1750 2018-02-09 Marek Polacek <polacek@redhat.com>
1751 Jakub Jelinek <jakub@redhat.com>
1754 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
1755 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
1756 Sync some changes from cxx_fold_indirect_ref.
1758 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1760 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
1762 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
1763 (BLOCK_INLINE_ENTRY_LABEL): New.
1764 (dwarf2out_var_location): Disregard inline entry markers.
1765 (inline_entry_data): New struct.
1766 (inline_entry_data_hasher): New hashtable type.
1767 (inline_entry_data_hasher::hash): New.
1768 (inline_entry_data_hasher::equal): New.
1769 (inline_entry_data_table): New variable.
1770 (add_high_low_attributes): Add DW_AT_entry_pc and
1771 DW_AT_GNU_entry_view attributes if a pending entry is found
1772 in inline_entry_data_table. Add old entry_pc attribute only
1773 if debug nonbinding markers are disabled.
1774 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
1775 markers are enabled.
1776 (block_within_block_p, dwarf2out_inline_entry): New.
1777 (dwarf2out_finish): Check that no entries remained in
1778 inline_entry_data_table.
1779 * final.c (reemit_insn_block_notes): Handle inline entry notes.
1780 (final_scan_insn, notice_source_line): Likewise.
1781 (rest_of_clean_state): Skip inline entry markers.
1782 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
1784 * gimple.c (gimple_build_debug_inline_entry): New.
1785 * gimple.h (enum gimple_debug_subcode): Add
1786 GIMPLE_DEBUG_INLINE_ENTRY.
1787 (gimple_build_debug_inline_entry): Declare.
1788 (gimple_debug_inline_entry_p): New.
1789 (gimple_debug_nonbind_marker_p): Adjust.
1790 * insn-notes.def (INLINE_ENTRY): New.
1791 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
1792 inline entry marker notes.
1793 (print_insn): Likewise.
1794 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
1795 (INSN_DEBUG_MARKER_KIND): Likewise.
1796 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
1797 * tree-inline.c (expand_call_inline): Build and insert
1798 debug_inline_entry stmt.
1799 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
1800 inline entry blocks early, if nonbind markers are enabled.
1801 (dump_scope_block): Dump fragment info.
1802 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
1803 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
1804 (gimple_build_debug_inline_entry): New.
1805 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
1806 Enable/disable inline entry points too.
1807 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
1808 (DEBUG_INSN): Describe inline entry markers.
1810 * common.opt (gvariable-location-views): New.
1811 (gvariable-location-views=incompat5): New.
1812 * config.in: Rebuilt.
1813 * configure: Rebuilt.
1814 * configure.ac: Test assembler for view support.
1815 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
1816 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
1817 * dwarf2out.c (var_loc_view): New typedef.
1818 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
1819 (dwarf2out_locviews_in_attribute): New.
1820 (dwarf2out_locviews_in_loclist): New.
1821 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
1822 (enum dw_line_info_opcode): Add LI_adv_address.
1823 (struct dw_line_info_table): Add view.
1824 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
1825 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
1826 (zero_view_p): New variable.
1827 (ZERO_VIEW_P): New macro.
1828 (output_asm_line_debug_info): New.
1829 (struct var_loc_node): Add view.
1830 (add_AT_view_list, AT_loc_list): New.
1831 (add_var_loc_to_decl): Add view param. Test it against last.
1832 (new_loc_list): Add view params. Record them.
1833 (AT_loc_list_ptr): Handle loc and view lists.
1834 (view_list_to_loc_list_val_node): New.
1835 (print_dw_val): Handle dw_val_class_view_list.
1836 (size_of_die): Likewise.
1837 (value_format): Likewise.
1838 (loc_list_has_views): New.
1839 (gen_llsym): Set vl_symbol too.
1840 (maybe_gen_llsym, skip_loc_list_entry): New.
1841 (dwarf2out_maybe_output_loclist_view_pair): New.
1842 (output_loc_list): Output view list or entries too.
1843 (output_view_list_offset): New.
1844 (output_die): Handle dw_val_class_view_list.
1845 (output_dwarf_version): New.
1846 (output_compilation_unit_header): Use it.
1847 (output_skeleton_debug_sections): Likewise.
1848 (output_rnglists, output_line_info): Likewise.
1849 (output_pubnames, output_aranges): Update version comments.
1850 (output_one_line_info_table): Output view numbers in asm comments.
1851 (dw_loc_list): Determine current endview, pass it to new_loc_list.
1852 Call maybe_gen_llsym.
1853 (loc_list_from_tree_1): Adjust.
1854 (add_AT_location_description): Create view list attribute if
1855 needed, check it's absent otherwise.
1856 (convert_cfa_to_fb_loc_list): Adjust.
1857 (maybe_emit_file): Call output_asm_line_debug_info for test.
1858 (dwarf2out_var_location): Reset views as needed. Precompute
1859 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
1860 attribute. Set view.
1861 (new_line_info_table): Reset next view.
1862 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
1863 (dwarf2out_source_line): Likewise. Output view resets and labels to
1864 the assembler, or select appropriate line info opcodes.
1865 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
1866 (optimize_string_length): Catch it. Adjust.
1867 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
1868 dw_val_class_view_list, and remove it if no longer needed.
1869 (hash_loc_list): Hash view numbers.
1870 (loc_list_hasher::equal): Compare them.
1871 (optimize_location_lists): Check whether a view list symbol is
1872 needed, and whether the locview attribute is present, and
1873 whether they match. Remove the locview attribute if no longer
1875 (index_location_lists): Call skip_loc_list_entry for test.
1876 (dwarf2out_finish): Call output_asm_line_debug_info for test.
1877 Use output_dwarf_version.
1878 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
1879 (struct dw_val_node): Add val_view_list.
1880 * final.c (SEEN_NEXT_VIEW): New.
1881 (set_next_view_needed): New.
1882 (clear_next_view_needed): New.
1883 (maybe_output_next_view): New.
1884 (final_start_function): Rename to...
1885 (final_start_function_1): ... this. Take pointer to FIRST,
1886 add SEEN parameter. Emit param bindings in the initial view.
1887 (final_start_function): Reintroduce SEEN-less interface.
1888 (final): Rename to...
1889 (final_1): ... this. Take SEEN parameter. Output final pending
1890 next view at the end.
1891 (final): Reintroduce seen-less interface.
1892 (final_scan_insn): Output pending next view before switching
1893 sections or ending a block. Mark the next view as needed when
1894 outputting variable locations. Notify debug backend of section
1895 changes, and of location view changes.
1896 (rest_of_handle_final): Adjust.
1897 * toplev.c (process_options): Autodetect value for debug variable
1898 location views option. Warn on incompat5 without -gdwarf-5.
1899 * doc/invoke.texi (gvariable-location-views): New.
1900 (gvariable-location-views=incompat5): New.
1901 (gno-variable-location-views): New.
1903 2018-02-08 David Malcolm <dmalcolm@redhat.com>
1905 PR tree-optimization/84136
1906 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
1907 that the result of find_edge is non-NULL.
1909 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1912 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
1913 storing integer register in SImode. Fix cost of 256 and 512
1914 byte aligned SSE register store.
1916 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1918 * config/i386/i386.c (ix86_multiplication_cost): Fix
1919 multiplication cost for TARGET_AVX512DQ.
1921 2018-02-08 Marek Polacek <polacek@redhat.com>
1923 PR tree-optimization/84238
1924 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
1927 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1929 PR tree-optimization/84265
1930 * tree-vect-stmts.c (vectorizable_store): Don't treat
1931 VMAT_CONTIGUOUS accesses as grouped.
1932 (vectorizable_load): Likewise.
1934 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1936 PR tree-optimization/81635
1937 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
1938 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
1939 (test_round_for_mask): New functions.
1940 (wide_int_cc_tests): Call test_round_for_mask.
1941 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
1942 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
1943 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
1944 range returned by get_range_info.
1946 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1949 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
1950 * symtab.c: Include builtins.h
1951 (symtab_node::output_to_lto_symbol_table_p): Move here
1952 from lto-streamer-out.c:output_symbol_p.
1953 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
1954 (output_symbol_p): Move all logic to symtab.c
1955 (produce_symtab): Update.
1957 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1959 * config/s390/s390-opts.h (enum indirect_branch): Define.
1960 * config/s390/s390-protos.h (s390_return_addr_from_memory)
1961 (s390_indirect_branch_via_thunk)
1962 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
1963 (enum s390_indirect_branch_type): Define.
1964 * config/s390/s390.c (struct s390_frame_layout, struct
1965 machine_function): Remove.
1966 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
1967 (indirect_branch_table_label_no, indirect_branch_table_name):
1969 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
1970 (enum s390_indirect_branch_option): Define.
1971 (s390_return_addr_from_memory): New function.
1972 (s390_handle_string_attribute): New function.
1973 (s390_attribute_table): Add new attribute handler.
1974 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
1975 (s390_indirect_branch_via_thunk): New function.
1976 (s390_indirect_branch_via_inline_thunk): New function.
1977 (s390_function_ok_for_sibcall): When jumping via thunk disallow
1978 sibling call optimization for non z10 compiles.
1979 (s390_emit_call): Force indirect branch target to be a single
1980 register. Add r1 clobber for non-z10 compiles.
1981 (s390_emit_epilogue): Emit return jump via return_use expander.
1982 (s390_reorg): Handle JUMP_INSNs as execute targets.
1983 (s390_option_override_internal): Perform validity checks for the
1984 new command line options.
1985 (s390_indirect_branch_attrvalue): New function.
1986 (s390_indirect_branch_settings): New function.
1987 (s390_set_current_function): Invoke s390_indirect_branch_settings.
1988 (s390_output_indirect_thunk_function): New function.
1989 (s390_code_end): Implement target hook.
1990 (s390_case_values_threshold): Implement target hook.
1991 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
1993 * config/s390/s390.h (struct s390_frame_layout)
1994 (struct machine_function): Move here from s390.c.
1995 (TARGET_INDIRECT_BRANCH_NOBP_RET)
1996 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
1997 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
1998 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
1999 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
2000 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
2001 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
2002 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
2003 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
2004 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
2005 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
2006 (mnemonic attribute): Add values which aren't recognized
2008 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
2009 pattern for branch conversion. Fix mnemonic attribute.
2010 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
2011 indirect branch via thunk if requested.
2012 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
2013 ("*indirect_jump"): Disable for branch conversion using out of
2015 ("indirect_jump_via_thunk<mode>_z10")
2016 ("indirect_jump_via_thunk<mode>")
2017 ("indirect_jump_via_inlinethunk<mode>_z10")
2018 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
2019 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
2020 ("casesi_jump_via_inlinethunk<mode>_z10")
2021 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
2022 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
2023 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
2024 ("*indirect2_jump"): Disable for branch conversion.
2025 ("casesi_jump"): Turn into expander and expand patterns for branch
2027 ("return_use"): New expander.
2028 ("*return"): Emit return via thunk and rename it to ...
2029 ("*return<mode>"): ... this one.
2030 * config/s390/s390.opt: Add new options and and enum for the
2033 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
2035 * lra-constraints.c (match_reload): Unconditionally use
2036 gen_lowpart_SUBREG, rather than selecting between that
2037 and equivalent gen_rtx_SUBREG code.
2039 2018-02-08 Richard Biener <rguenther@suse.de>
2041 PR tree-optimization/84233
2042 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
2043 changed flag instead of boguously re-using phi_inserted.
2045 2018-02-08 Martin Jambor <mjambor@suse.cz>
2047 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
2048 static local variables.
2050 2018-02-08 Richard Biener <rguenther@suse.de>
2052 PR tree-optimization/84278
2053 * tree-vect-stmts.c (vectorizable_store): When looking for
2054 smaller vector types to perform grouped strided loads/stores
2055 make sure the mode is supported by the target.
2056 (vectorizable_load): Likewise.
2058 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2060 * config/aarch64/aarch64.c (aarch64_components_for_bb):
2061 Increase LDP/STP opportunities by adding adjacent callee-saves.
2063 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
2065 PR rtl-optimization/84068
2066 PR rtl-optimization/83459
2067 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
2069 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
2071 PR tree-optimization/84224
2072 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
2073 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
2076 2018-02-07 Iain Sandoe <iain@codesourcery.com>
2079 * config/rs6000/altivec.md (*restore_world): Remove LR use.
2080 * config/rs6000/predicates.md (restore_world_operation): Adjust op
2081 count, remove one USE.
2083 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
2085 * doc/install.texi (Configuration): Document the
2086 --with-long-double-format={ibm,ieee} PowerPC configuration
2090 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
2091 Convert from define_expand to be define_insn_and_split. Rework
2092 float/double/_Float128 conversions to QI/HI/SImode to work with
2093 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
2094 conversions to QI/HImode types did a store and then a load to
2095 truncate the value. For conversions to VSX registers, don't split
2096 the insn, instead emit the code directly. Use the code iterator
2097 any_fix to combine signed and unsigned conversions.
2098 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
2099 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
2100 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2101 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
2102 (fix_<mode>di2_hw): Likewise.
2103 (fixuns_<mode>di2_hw): Likewise.
2104 (fix_<mode>si2_hw): Likewise.
2105 (fixuns_<mode>si2_hw): Likewise.
2106 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
2107 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
2108 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
2109 fix<uns>_trunc<SFDF:mode>si2_p8.
2110 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
2112 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
2113 (fix<uns>_<mode>_mem): Likewise.
2114 (fctiw<u>z_<mode>_mem): Likewise.
2115 (fix<uns>_<mode>_mem): Likewise.
2116 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
2117 the register allocator from doing a direct move to the GPRs to do
2118 a store, and instead use the ISA 3.0 store byte/half-word from
2119 vector register instruction. For IEEE 128-bit floating point,
2120 also optimize stores of 32-bit ints.
2121 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
2123 2018-02-07 Alan Hayward <alan.hayward@arm.com>
2125 * genextract.c (push_pathstr_operand): New function to support
2127 (walk_rtx): Call push_pathstr_operand.
2128 (print_path): Support [a-zA-Z].
2130 2018-02-07 Richard Biener <rguenther@suse.de>
2132 PR tree-optimization/84037
2133 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
2134 (cse_and_gimplify_to_preheader): Declare.
2135 (vect_get_place_in_interleaving_chain): Likewise.
2136 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
2138 (_loop_vec_info::~_loop_vec_info): Delete it.
2139 (cse_and_gimplify_to_preheader): New function.
2140 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
2141 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
2142 (vectorizable_load): Likewise. For grouped stores always base
2143 the IV on the first element.
2144 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
2145 condition before gimplifying.
2147 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2149 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
2150 *DIV_EXPR and *MOD_EXPR.
2152 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
2155 * config/i386/i386.c (ix86_option_override_internal): Mask out
2156 the CF_SET bit when checking -fcf-protection.
2158 2018-02-07 Tom de Vries <tom@codesourcery.com>
2161 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
2164 2018-02-07 Richard Biener <rguenther@suse.de>
2166 PR tree-optimization/84204
2167 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
2170 PR tree-optimization/84205
2171 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
2172 special-case isl_ast_op_zdiv_r.
2174 PR tree-optimization/84223
2175 * graphite-scop-detection.c (gather_bbs::before_dom_children):
2176 Only add conditions from within the region.
2177 (gather_bbs::after_dom_children): Adjust.
2179 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
2182 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
2183 * config/avr/avr.md: Only post-reload split REG-REG moves if
2184 either register is GENERAL_REG_P.
2186 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2188 PR tree-optimization/84235
2189 * tree-ssa-scopedtables.c
2190 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
2191 if the subtraction is performed in floating point type where NaNs are
2192 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
2193 build 1. Formatting fix.
2195 2018-02-06 Jakub Jelinek <jakub@redhat.com>
2198 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
2199 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
2200 and skip it regardless of bb boundaries. Use CALL_P macro,
2201 don't test INSN_P (insn) together with CALL_P or JUMP_P check
2202 unnecessarily, formatting fix.
2204 2018-02-06 Michael Collison <michael.collison@arm.com>
2206 * config/arm/thumb2.md:
2207 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
2208 (*thumb_mov_notscc): Ditto.
2210 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
2213 * config/rs6000/rs6000.md (su code attribute): Use "u" for
2214 unsigned_fix, not "s".
2216 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2218 * configure.ac (gcc_fn_eh_frame_ro): New function.
2219 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
2220 correct .eh_frame permissions.
2221 * configure: Regenerate.
2223 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
2225 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
2228 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2230 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2231 Display warning message for -mno-speculate-indirect-jumps.
2233 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
2235 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
2237 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
2239 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
2241 PR tree-optimization/84225
2242 * tree-eh.c (find_trapping_overflow): Only call
2243 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
2245 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2248 * config/i386/i386.c: Reimplement the check of possible options
2249 -mibt/-mshstk conbination. Change error messages.
2250 * doc/invoke.texi: Fix a typo: remove extra '='.
2252 2018-02-06 Marek Polacek <polacek@redhat.com>
2254 PR tree-optimization/84228
2255 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
2257 2018-02-06 Tamar Christina <tamar.christina@arm.com>
2260 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
2261 emitted arch directives.
2262 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
2263 __ARM_FEATURE_COPROC before changing architectures.
2265 2018-02-06 Richard Biener <rguenther@suse.de>
2267 * config/i386/i386.c (print_reg): Fix typo.
2268 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
2270 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
2272 * configure: Regenerate.
2274 2018-02-05 Martin Sebor <msebor@redhat.com>
2276 PR tree-optimization/83369
2277 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
2280 2018-02-05 Martin Liska <mliska@suse.cz>
2282 * doc/invoke.texi: Cherry-pick upstream r323995.
2284 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
2286 * ira.c (ira_init_register_move_cost): Adjust comment.
2288 2018-02-05 Martin Liska <mliska@suse.cz>
2290 PR gcov-profile/84137
2291 * doc/gcov.texi: Fix typo in documentation.
2293 2018-02-05 Martin Liska <mliska@suse.cz>
2295 PR gcov-profile/83879
2296 * doc/gcov.texi: Document necessity of --dynamic-list-data when
2297 using dlopen functionality.
2299 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
2301 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
2302 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
2303 _mm_maskz_range_ss, _mm_mask_range_round_ss,
2304 _mm_maskz_range_round_ss): New intrinsics.
2305 (__builtin_ia32_rangesd128_round)
2306 (__builtin_ia32_rangess128_round): Remove.
2307 (__builtin_ia32_rangesd128_mask_round,
2308 __builtin_ia32_rangess128_mask_round): New builtins.
2309 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
2310 __builtin_ia32_rangess128_round): Remove.
2311 (__builtin_ia32_rangesd128_mask_round,
2312 __builtin_ia32_rangess128_mask_round): New builtins.
2313 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
2314 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
2315 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
2316 "<round_saeonly_constraint>")): Changed to ...
2317 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
2318 "<round_saeonly_scalar_constraint>")): ... this.
2319 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
2320 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
2321 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
2322 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
2323 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
2325 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
2327 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
2329 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
2330 Remove all values except native, 8540 and 8548.
2332 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
2334 * config/i386/i386.c (ix86_output_function_return): Pass
2335 INVALID_REGNUM, instead of -1, as invalid register number to
2336 indirect_thunk_name and output_indirect_thunk.
2338 2018-02-02 Julia Koval <julia.koval@intel.com>
2340 * config.gcc: Add -march=icelake.
2341 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
2342 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
2343 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
2344 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
2345 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
2346 (processor_target_table): Add icelake.
2347 (ix86_option_override_internal): Handle new PTAs.
2348 (get_builtin_code_for_version): Handle icelake.
2349 (M_INTEL_COREI7_ICELAKE): New.
2350 (fold_builtin_cpu): Handle icelake.
2351 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
2352 * doc/invoke.texi: Add -march=icelake.
2354 2018-02-02 Julia Koval <julia.koval@intel.com>
2356 * config/i386/i386.c (ix86_option_override_internal): Change flags type
2357 to wide_int_bitmask.
2358 * wide-int-bitmask.h: New.
2360 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2363 * config/i386/i386.md: Replace Pmode with word_mode in
2364 builtin_setjmp_setup and builtin_longjmp to support x32.
2366 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
2370 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
2372 (rs6000_supported_cpu_names): New static variable.
2373 (linux_cpu_translation_table): Likewise.
2374 (elf_platform) <cpu>: Define new static variable and use it.
2375 Translate kernel AT_PLATFORM name to canonical name if needed.
2376 Error if platform name is unknown.
2378 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
2381 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
2383 2018-02-01 Jeff Law <law@redhat.com>
2386 * config/i386/i386.c (release_scratch_register_on_entry): Add new
2387 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
2388 the scratch if RELEASE_VIA_POP is false.
2389 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
2390 If we have to save a temporary register, decrement SIZE appropriately.
2391 Pass new arguments to release_scratch_register_on_entry.
2392 (ix86_adjust_stack_and_probe): Likewise.
2393 (ix86_emit_probe_stack_range): Pass new arguments to
2394 release_scratch_register_on_entry.
2396 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
2398 PR rtl-optimization/84157
2399 * combine.c (change_zero_ext): Use REG_P predicate in
2400 front of HARD_REGISTER_P predicate.
2402 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
2404 * config/avr/avr.c (avr_option_override): Move disabling of
2405 -fdelete-null-pointer-checks to...
2406 * common/config/avr/avr-common.c (avr_option_optimization_table):
2409 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2411 PR tree-optimization/81635
2412 * tree-data-ref.c (split_constant_offset_1): For types that
2413 wrap on overflow, try to use range info to prove that wrapping
2416 2018-02-01 Renlin Li <renlin.li@arm.com>
2419 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
2421 (aarch64_register_move_cost): Likewise.
2422 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
2424 (REG_CLASS_NAMES): Likewise.
2425 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
2426 TAILCALL_ADDR_REGS. Remove IP registers.
2427 * config/aarch64/aarch64.md (Ucs): Update register constraint.
2429 2018-02-01 Richard Biener <rguenther@suse.de>
2431 * domwalk.h (dom_walker::dom_walker): Add additional constructor
2432 for specifying RPO order and allow NULL for that.
2433 * domwalk.c (dom_walker::dom_walker): Likewise.
2434 (dom_walker::walk): Handle NULL RPO order.
2435 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
2437 (rewrite_update_dom_walker): Likewise.
2438 (mark_def_dom_walker): Likewise.
2440 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2442 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
2443 (aarch64_maybe_expand_sve_subreg_move): Declare.
2444 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
2445 * config/aarch64/predicates.md (aarch64_any_register_operand): New
2447 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
2448 that are semantically a reverse operation.
2449 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
2450 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
2451 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
2453 (aarch64_can_change_mode_class): For big-endian, forbid changes
2454 between two SVE modes if they have different element sizes.
2456 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2458 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
2459 the TImode handling for big-endian targets.
2461 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2463 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
2464 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
2466 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
2467 Remove BSWAP handing for big-endian targets and use the form of
2468 LD1RQ appropariate for the mode.
2470 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2472 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
2473 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
2476 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2479 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
2480 check for operands that need to go through aarch64_sve_reload_be.
2482 2018-02-01 Jakub Jelinek <jakub@redhat.com>
2484 PR tree-optimization/81661
2485 PR tree-optimization/84117
2486 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
2487 * tree-eh.c: Include gimplify.h.
2488 (find_trapping_overflow, replace_trapping_overflow,
2489 rewrite_to_non_trapping_overflow): New functions.
2490 * tree-vect-loop.c: Include tree-eh.h.
2491 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
2492 * tree-data-ref.c: Include tree-eh.h.
2493 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
2495 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
2497 PR rtl-optimization/84123
2498 * combine.c (change_zero_ext): Check if hard register satisfies
2499 can_change_dest_mode before calling gen_lowpart_SUBREG.
2501 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
2504 * ira.c (ira_init_register_move_cost): Remove assert.
2506 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2508 PR rtl-optimization/84071
2509 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
2510 * doc/tm.texi: Regenerate.
2512 2018-01-31 Richard Biener <rguenther@suse.de>
2514 PR tree-optimization/84132
2515 * tree-data-ref.c (analyze_miv_subscript): Properly
2516 check whether evolution_function_is_affine_multivariate_p
2517 before calling gcd_of_steps_may_divide_p.
2519 2018-01-31 Julia Koval <julia.koval@intel.com>
2522 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
2523 * config/i386/i386.md (rdpid_rex64) New.
2524 (rdpid): Make 32bit only.
2526 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
2529 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
2530 an IDENTIFIER_NODE for FUNCTION_TYPE's.
2532 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2535 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
2537 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
2539 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2541 PR rtl-optimization/84071
2542 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
2543 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
2545 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2547 * config/arc/arc.c (arc_handle_aux_attribute): New function.
2548 (arc_attribute_table): Add 'aux' attribute.
2549 (arc_in_small_data_p): Consider aux like variables.
2550 (arc_is_aux_reg_p): New function.
2551 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
2552 (arc_get_aux_arg): New function.
2553 (prepare_move_operands): Handle aux-register access.
2554 (arc_handle_aux_attribute): New function.
2555 * doc/extend.texi (ARC Variable attributes): Add subsection.
2557 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2559 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
2560 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
2561 (arc_attribute_table): Add 'uncached' attribute.
2562 (arc_print_operand): Print '.di' flag for uncached memory
2564 (arc_in_small_data_p): Do not consider for small data the uncached
2566 (arc_is_uncached_mem_p): New function.
2567 * config/arc/predicates.md (compact_store_memory_operand): Check
2568 for uncached memory accesses.
2569 (nonvol_nonimm_operand): Likewise.
2570 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
2572 2018-01-31 Jakub Jelinek <jakub@redhat.com>
2575 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
2576 falign-loops=): Add Optimization flag.
2578 2018-01-30 Jeff Law <law@redhat.com>
2581 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
2582 INT_REGISTERS_SAVED. Check it prior to calling
2583 get_scratch_register_on_entry.
2584 (ix86_adjust_stack_and_probe): Similarly.
2585 (ix86_emit_probe_stack_range): Similarly.
2586 (ix86_expand_prologue): Corresponding changes.
2588 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2591 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
2592 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
2594 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
2597 * lra-constraints.c (curr_insn_transform): Process AND in the
2600 2018-01-30 Jakub Jelinek <jakub@redhat.com>
2602 PR rtl-optimization/83986
2603 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
2604 dependence against last_pending_memory_flush in addition to
2607 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
2609 PR tree-optimization/81611
2610 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
2613 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2616 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
2619 2018-01-30 Richard Biener <rguenther@suse.de>
2620 Jakub Jelinek <jakub@redhat.com>
2622 PR tree-optimization/84111
2623 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
2624 inner loops added during recursion, as they don't have up-to-date
2627 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2630 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
2631 (can_inline_edge_by_limits_p): ... here.
2632 (can_early_inline_edge_p, check_callers,
2633 update_caller_keys, update_callee_keys, recursive_inlining,
2634 add_new_edges_to_heap, speculation_useful_p,
2635 inline_small_functions,
2636 inline_small_functions, flatten_function,
2637 inline_to_all_callers_1): Update.
2639 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2641 * profile-count.c (profile_count::combine_with_ipa_count): Handle
2644 2018-01-30 Richard Biener <rguenther@suse.de>
2646 PR tree-optimization/83008
2647 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
2648 invariant and constant vector uses in stmts when they need
2651 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2654 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
2655 * configure: Regenerate.
2657 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2659 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
2661 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
2662 Use gen_rtx_REG rather than gen_lowpart.
2664 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2666 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
2667 rather than 0 when creating partial subregs.
2669 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2671 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
2674 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
2677 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
2678 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
2679 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
2680 flags. This restores the settings used before the 2017-07-24.
2681 Turning off pre increment/decrement/modify allows IVOPTS to
2682 optimize DF/SF loops where the index is an int.
2684 2018-01-29 Richard Biener <rguenther@suse.de>
2685 Kelvin Nilsen <kelvin@gcc.gnu.org>
2688 * tree-vect-stmts.c (vectorizable_call): Don't call
2689 targetm.vectorize_builtin_md_vectorized_function if callee is
2692 2018-01-22 Carl Love <cel@us.ibm.com>
2694 * doc/extend.tex: Fix typo in second arg in
2695 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
2697 2018-01-29 Richard Biener <rguenther@suse.de>
2699 PR tree-optimization/84086
2700 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
2701 (flush_ssaname_freelist): When SSA names were released reset
2702 the SCEV hash table.
2704 2018-01-29 Richard Biener <rguenther@suse.de>
2706 PR tree-optimization/84057
2707 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
2708 removed paths when removing edges.
2710 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
2712 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
2713 -mfunction-return=@var{choice}.
2715 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
2718 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
2720 (layout::print_source_line): Likewise.
2721 (test_get_line_width_without_trailing_whitespace): Add test cases.
2723 2018-01-27 Jakub Jelinek <jakub@redhat.com>
2726 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
2729 2018-01-26 Jim Wilson <jimw@sifive.com>
2731 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
2733 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
2736 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2738 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
2739 and CMP + SUB-immediate -> SUBS.
2741 2018-01-26 Martin Sebor <msebor@redhat.com>
2743 PR tree-optimization/83896
2744 * tree-ssa-strlen.c (get_string_len): Rename...
2745 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
2746 Avoid assuming length is constant.
2747 (handle_char_store): Use HOST_WIDE_INT for string length.
2749 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
2752 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
2753 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
2755 2018-01-26 Richard Biener <rguenther@suse.de>
2757 PR rtl-optimization/84003
2758 * dse.c (record_store): Only record redundant stores when
2759 the earlier store aliases at least all accesses the later one does.
2761 2018-01-26 Jakub Jelinek <jakub@redhat.com>
2763 PR rtl-optimization/83985
2764 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
2765 REG_CFA_RESTORE insns.
2766 (delete_unmarked_insns): Don't ignore separate shrink wrapping
2767 REG_CFA_RESTORE insns here.
2770 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
2771 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
2773 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2775 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
2776 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
2777 (arc_init): Likewise.
2778 (arc_override_options): Likewise.
2779 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
2781 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
2783 * config/arc/arc.h (TARGET_DBNZ): Define.
2784 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
2785 properly set the tune attribute.
2786 (dbnz): Use TARGET_DBNZ guard.
2787 * config/arc/arc.opt (mtune): Add core3 option.
2789 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2791 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
2792 recognize new pic like addresses.
2793 (arc_delegitimize_address): Clean up.
2795 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2797 * config/arc/arc-arches.def: Option mrf16 valid for all
2799 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
2800 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
2801 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
2802 * config/arc/arc-tables.opt: Regenerate.
2803 * config/arc/arc.c (arc_conditional_register_usage): Handle
2804 reduced register file case.
2805 (arc_file_start): Set must have build attributes.
2806 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
2808 * config/arc/arc.opt (mrf16): Add new option.
2809 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
2810 * config/arc/genmultilib.awk: Handle new mrf16 option.
2811 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
2812 * config/arc/t-multilib: Regenerate.
2813 * doc/invoke.texi (ARC Options): Document mrf16 option.
2815 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2817 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
2818 * config/arc/arc.c (arc_handle_secure_attribute): New function.
2819 (arc_attribute_table): Add 'secure_call' attribute.
2820 (arc_print_operand): Print secure call operand.
2821 (arc_function_ok_for_sibcall): Don't optimize tail calls when
2823 (arc_is_secure_call_p): New function. * config/arc/arc.md
2824 (call_i): Add support for sjli instruction.
2825 (call_value_i): Likewise.
2826 * config/arc/constraints.md (Csc): New constraint.
2828 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2829 John Eric Martin <John.Martin@emmicro-us.com>
2831 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
2832 * config/arc/arc.c (_arc_jli_section): New struct.
2833 (arc_jli_section): New type.
2834 (rc_jli_sections): New static variable.
2835 (arc_handle_jli_attribute): New function.
2836 (arc_attribute_table): Add jli_always and jli_fixed attribute.
2837 (arc_file_end): New function.
2838 (TARGET_ASM_FILE_END): Define.
2839 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
2840 (arc_add_jli_section): New function.
2841 (jli_call_scan): Likewise.
2842 (arc_reorg): Call jli_call_scan.
2843 (arc_output_addsi): Remove 'S' from printing asm operand.
2844 (arc_is_jli_call_p): New function.
2845 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
2847 (movhi_insn): Likewise.
2848 (movsi_insn): Likewise.
2849 (movsi_set_cc_insn): Likewise.
2850 (loadqi_update): Likewise.
2851 (load_zeroextendqisi_update): Likewise.
2852 (load_signextendqisi_update): Likewise.
2853 (loadhi_update): Likewise.
2854 (load_zeroextendhisi_update): Likewise.
2855 (load_signextendhisi_update): Likewise.
2856 (loadsi_update): Likewise.
2857 (loadsf_update): Likewise.
2858 (movsicc_insn): Likewise.
2859 (bset_insn): Likewise.
2860 (bxor_insn): Likewise.
2861 (bclr_insn): Likewise.
2862 (bmsk_insn): Likewise.
2863 (bicsi3_insn): Likewise.
2864 (cmpsi_cc_c_insn): Likewise.
2865 (movsi_ne): Likewise.
2866 (movsi_cond_exec): Likewise.
2867 (clrsbsi2): Likewise.
2878 (call_i): Remove 'S' asm letter, add jli instruction.
2879 (call_value_i): Likewise.
2880 * config/arc/arc.op (mjli-always): New option.
2881 * config/arc/constraints.md (Cji): New constraint.
2882 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
2884 (subsf3_fpx): Likewise.
2885 (mulsf3_fpx): Likewise.
2886 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
2888 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
2890 * doc/invoke.texi (ARC): Document mjli-always option.
2892 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
2894 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
2895 avoid addition with 0 and use incw and decw where possible.
2897 2018-01-26 Richard Biener <rguenther@suse.de>
2899 PR tree-optimization/81082
2900 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
2901 association if it requires casting to unsigned.
2902 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
2903 from fold_plusminus_mult_expr to catch important cases late when
2904 range info is available.
2906 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2908 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
2909 * configure.ac (hidden_linkonce): New test.
2910 * configure: Regenerate.
2911 * config.in: Regenerate.
2913 2018-01-26 Julia Koval <julia.koval@intel.com>
2915 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
2916 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
2917 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
2918 _mm_mask_bitshuffle_epi64_mask): Fix type.
2919 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
2920 USI_FTYPE_V4DI_V4DI_USI): Remove.
2921 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
2922 __builtin_ia32_vpshufbitqmb256_mask,
2923 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
2924 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
2925 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
2927 2018-01-26 Alan Modra <amodra@gmail.com>
2930 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
2931 UNSPEC_VBPERMQ. Sort other unspecs.
2933 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
2935 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
2937 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
2940 * predict.c (drop_profile): Do not push/pop cfun; update also
2942 (handle_missing_profiles): Fix logic looking for zero profiles.
2944 2018-01-25 Jakub Jelinek <jakub@redhat.com>
2947 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
2948 on functions with #pragma omp declare simd or functions with simd
2950 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
2951 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
2952 Remove trailing \n from warning_at calls.
2954 2018-01-25 Tom de Vries <tom@codesourcery.com>
2957 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
2958 for neutered workers.
2960 2018-01-24 Joseph Myers <joseph@codesourcery.com>
2963 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
2964 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
2966 2018-01-24 Jeff Law <law@redhat.com>
2969 * i386.c (get_probe_interval): Move to earlier point.
2970 (ix86_compute_frame_layout): If -fstack-clash-protection and
2971 the frame is larger than the probe interval, then use pushes
2972 to save registers rather than reg->mem moves.
2973 (ix86_expand_prologue): Remove conditional for int_registers_saved
2976 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
2979 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
2980 min/max for never referenced object.
2982 2018-01-24 Jakub Jelinek <jakub@redhat.com>
2985 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
2987 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
2988 attributes from DECL_ATTRIBUTES (decl) without affecting
2989 DECL_ATTRIBUTES (current_function_decl).
2990 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
2991 functions with non-NULL DECL_ABSTRACT_ORIGIN.
2993 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
2995 PR tree-optimization/83979
2996 * fold-const.c (fold_comparison): Use constant_boolean_node
2997 instead of boolean_{true,false}_node.
2999 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
3001 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
3004 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3006 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3007 Simplify the clause that sets the length attribute.
3008 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3009 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
3010 clause that sets the length attribute.
3011 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3013 2018-01-24 Tom de Vries <tom@codesourcery.com>
3016 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
3017 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
3018 Add strict parameter.
3019 (prevent_branch_around_nothing): Insert dummy insn between branch to
3020 label and label with no ptx insn inbetween.
3021 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
3023 2018-01-24 Tom de Vries <tom@codesourcery.com>
3026 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
3027 for neutered threads in warp.
3028 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
3030 2018-01-24 Richard Biener <rguenther@suse.de>
3032 PR tree-optimization/83176
3033 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
3036 2018-01-24 Richard Biener <rguenther@suse.de>
3038 PR tree-optimization/82819
3039 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
3040 code generating pluses that are no-ops in the target precision.
3042 2018-01-24 Richard Biener <rguenther@suse.de>
3045 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
3047 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3049 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
3050 to merge probabilities.
3051 * predict.c (probably_never_executed): Also mark as cold functions
3052 with global 0 profile and guessed local profile.
3053 * profile-count.c (profile_probability::combine_with_count): New
3055 * profile-count.h (profile_probability::operator*,
3056 profile_probability::operator*=, profile_probability::operator/,
3057 profile_probability::operator/=): Reduce precision to adjusted
3058 and set value to guessed on contradictory divisions.
3059 (profile_probability::combine_with_freq): Remove.
3060 (profile_probability::combine_wiht_count): Declare.
3061 (profile_count::force_nonzero):: Set to adjusted.
3062 (profile_count::probability_in):: Set quality to adjusted.
3063 * tree-ssa-tail-merge.c (replace_block_by): Use
3066 2018-01-23 Andrew Waterman <andrew@sifive.com>
3067 Jim Wilson <jimw@sifive.com>
3069 * config/riscv/riscv.c (riscv_stack_boundary): New.
3070 (riscv_option_override): Set riscv_stack_boundary. Handle
3071 riscv_preferred_stack_boundary_arg.
3072 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
3073 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
3074 (STACK_BOUNDARY): Set to riscv_stack_boundary.
3075 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
3076 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
3077 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
3079 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
3082 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
3083 of struct ix86_frame.
3084 (ix86_expand_epilogue): Likewise. Add a local variable for
3085 the reg_save_offset field in struct ix86_frame.
3087 2018-01-23 Bin Cheng <bin.cheng@arm.com>
3089 PR tree-optimization/82604
3090 * tree-loop-distribution.c (enum partition_kind): New enum item
3091 PKIND_PARTIAL_MEMSET.
3092 (partition_builtin_p): Support above new enum item.
3093 (generate_code_for_partition): Ditto.
3094 (compute_access_range): Differentiate cases that equality can be
3095 proven at all loops, the innermost loops or no loops.
3096 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
3097 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
3098 (finalize_partitions, distribute_loop): Don't fuse partition of
3099 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
3100 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
3103 2018-01-23 Martin Liska <mliska@suse.cz>
3105 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
3106 order to ignore the predictor.
3107 (PRED_POLYMORPHIC_CALL): Likewise.
3108 (PRED_RECURSIVE_CALL): Likewise.
3110 2018-01-23 Martin Liska <mliska@suse.cz>
3112 * tree-profile.c (tree_profiling): Print function header to
3113 aware reader which function we are working on.
3114 * value-prof.c (gimple_find_values_to_profile): Do not print
3115 not interesting value histograms.
3117 2018-01-23 Martin Liska <mliska@suse.cz>
3119 * profile-count.h (enum profile_quality): Add
3120 profile_uninitialized as the first value. Do not number values
3121 as they are zero based.
3122 (profile_count::verify): Update sanity check.
3123 (profile_probability::verify): Likewise.
3125 2018-01-23 Nathan Sidwell <nathan@acm.org>
3127 * doc/invoke.texi (ffor-scope): Deprecate.
3129 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3131 PR tree-optimization/83510
3132 * domwalk.c (set_all_edges_as_executable): New function.
3133 (dom_walker::dom_walker): Convert bool param
3134 "skip_unreachable_blocks" to enum reachability. Move setup of
3135 edge flags to set_all_edges_as_executable and only do it when
3136 reachability is REACHABLE_BLOCKS.
3137 * domwalk.h (enum dom_walker::reachability): New enum.
3138 (dom_walker::dom_walker): Convert bool param
3139 "skip_unreachable_blocks" to enum reachability.
3140 (set_all_edges_as_executable): New decl.
3141 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
3142 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
3144 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
3145 but converting true to REACHABLE_BLOCKS.
3146 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
3148 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
3149 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
3150 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
3152 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
3153 if check_all_array_refs will be called.
3155 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3157 * tree.c (selftest::test_location_wrappers): Add more test
3160 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3162 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
3163 (selftest::test_bit_in_range): Likewise.
3165 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3168 * doc/sourcebuild.texi (vect_float): Say that the selector
3169 only describes the situation when -funsafe-math-optimizations is on.
3170 (vect_float_strict): Document.
3172 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3174 PR tree-optimization/83965
3175 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
3176 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
3177 instead of checking only for a reduction.
3178 (vect_recog_widen_sum_pattern): Likewise.
3180 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3182 * predict.c (probably_never_executed): Only use precise profile info.
3183 (compute_function_frequency): Skip after inlining hack since we now
3184 have quality checking.
3186 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3188 * profile-count.h (profile_probability::very_unlikely,
3189 profile_probability::unlikely, profile_probability::even): Set
3190 precision to guessed.
3192 2018-01-23 Richard Biener <rguenther@suse.de>
3194 PR tree-optimization/83963
3195 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
3196 Properly terminate dominator walk when crossing the exit edge not
3197 when visiting its source block.
3199 2018-01-23 Jakub Jelinek <jakub@redhat.com>
3202 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
3203 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
3205 2018-01-22 Jakub Jelinek <jakub@redhat.com>
3207 PR tree-optimization/83957
3208 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
3209 semicolon after for body surrounded by braces.
3211 PR tree-optimization/83081
3212 * profile-count.h (profile_probability::split): New method.
3213 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
3214 Use profile_probability::split.
3215 (do_compare_rtx_and_jump): Fix adjustment of probabilities
3216 when splitting a single conditional jump into 2.
3218 2018-01-22 David Malcolm <dmalcolm@redhat.com>
3220 PR tree-optimization/69452
3221 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
3224 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3226 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
3227 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
3228 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
3230 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3232 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
3233 * config/rl78/rl78.md (movdi): New define_expand.
3234 * config/rl78/rl78.c (rl78_split_movdi): New function.
3236 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
3239 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
3241 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
3242 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
3243 128-bit to produce an UNSPEC move to get the double word with the
3244 signbit and then a shift directly to do signbit.
3245 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
3246 implementation with a new version that just does either a direct
3247 move or a regular move. Move memory interface to separate insns.
3248 Move insns so they are next to the expander.
3249 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
3250 with signbit move. Split big and little endian case.
3251 (signbit<mode>2_dm_mem_le): Likewise.
3252 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
3253 (signbit<mode>2_dm2): Likewise.
3255 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3257 * config/rl78/rl78.md (anddi3): New define_expand.
3259 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3261 * config/rl78/rl78.md (umindi3): New define_expand.
3263 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3265 * config/rl78/rl78.md (smindi3): New define_expand.
3267 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3269 * config/rl78/rl78.md (smaxdi3): New define_expand.
3271 2018-01-22 Carl Love <cel@us.ibm.com>
3273 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
3274 LVX_V1TI): Add macro expansion.
3275 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
3276 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
3277 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
3278 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
3279 Change check to determine if the instruction is a byte reversing
3280 entry. Fix typo in comment.
3281 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
3282 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
3283 Add def_builtin calls for new builtins.
3284 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
3285 Add define_insn expansion.
3287 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3289 * config/rl78/rl78.md (umaxdi3): New define_expand.
3291 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3293 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
3294 for non-QImode registers.
3296 2018-01-22 Richard Biener <rguenther@suse.de>
3298 PR tree-optimization/83963
3299 * graphite-scop-detection.c (scop_detection::get_sese): Delay
3300 including the loop exit block.
3301 (scop_detection::merge_sese): Likewise.
3302 (scop_detection::add_scop): Do it here instead.
3304 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3306 * doc/sourcebuild.texi (arm_softfloat): Document.
3308 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
3311 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
3312 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
3313 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
3315 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3316 David Edelsohn <dje.gcc@gmail.com>
3319 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3320 Change "crset eq" to "crset 2".
3321 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3322 (*call_indirect_aix<mode>_nospec): Likewise.
3323 (*call_value_indirect_aix<mode>_nospec): Likewise.
3324 (*call_indirect_elfv2<mode>_nospec): Likewise.
3325 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
3326 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
3327 change assembly output from . to $.
3328 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3329 (indirect_jump<mode>_nospec): Change assembly output from . to $.
3330 (*tablejump<mode>_internal1_nospec): Likewise.
3332 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
3335 * config/sh/sh_optimize_sett_clrt.cc:
3336 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
3338 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
3340 PR tree-optimization/83940
3341 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
3342 offset_dt to vect_constant_def rather than vect_unknown_def_type.
3343 (vect_check_load_store_mask): Add a mask_dt_out parameter and
3344 use it to pass back the definition type.
3345 (vect_check_store_rhs): Likewise rhs_dt_out.
3346 (vect_build_gather_load_calls): Add a mask_dt argument and use
3347 it instead of a call to vect_is_simple_use.
3348 (vectorizable_store): Update calls to vect_check_load_store_mask
3349 and vect_check_store_rhs. Use the dt returned by the latter instead
3350 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
3351 instead of calls to vect_is_simple_use. Pass the scalar rather
3352 than the vector operand to vect_is_simple_use when handling
3353 second and subsequent copies of an rhs value.
3354 (vectorizable_load): Update calls to vect_check_load_store_mask
3355 and vect_build_gather_load_calls. Use the cached mask_dt and
3356 gs_info.offset_dt instead of calls to vect_is_simple_use.
3358 2018-01-20 Jakub Jelinek <jakub@redhat.com>
3361 * tree-emutls.c: Include gimplify.h.
3362 (lower_emutls_2): New function.
3363 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
3364 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
3365 it before further processing.
3368 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
3369 UINTVAL (trueop1) instead of INTVAL (op1).
3371 2018-01-19 Jakub Jelinek <jakub@redhat.com>
3375 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
3376 INCOMING_FRAME_SP_OFFSET if not defined.
3377 (scan_trace): Add ENTRY argument. If true and
3378 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
3379 emit a note to adjust the CFA offset.
3380 (create_cfi_notes): Adjust scan_trace callers.
3381 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
3382 INCOMING_FRAME_SP_OFFSET in the CIE.
3383 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
3384 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
3386 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
3387 * doc/tm.texi: Regenerated.
3389 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3391 PR rtl-optimization/83147
3392 * lra-constraints.c (remove_inheritance_pseudos): Use
3393 lra_substitute_pseudo_within_insn.
3395 2018-01-19 Tom de Vries <tom@codesourcery.com>
3396 Cesar Philippidis <cesar@codesourcery.com>
3399 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
3401 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
3404 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
3405 spaces for function labels.
3407 2018-01-19 Martin Liska <mliska@suse.cz>
3409 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
3410 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
3411 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
3412 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
3413 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
3414 (PRED_CONST_RETURN): Change from 69 to 65.
3415 (PRED_NULL_RETURN): Change from 91 to 71.
3416 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
3417 (PRED_LOOP_GUARD): Change from 66 to 73.
3419 2018-01-19 Martin Liska <mliska@suse.cz>
3421 * predict.c (predict_insn_def): Add new assert.
3422 (struct branch_predictor): Change type to signed integer.
3423 (test_prediction_value_range): Amend test to cover
3425 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
3426 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
3427 (PRED_LOOP_ITERATIONS_MAX): Likewise.
3428 (PRED_LOOP_IV_COMPARE): Likewise.
3429 * predict.h (PROB_UNINITIALIZED): Define new constant.
3431 2018-01-19 Martin Liska <mliska@suse.cz>
3433 * predict.c (dump_prediction): Add new format for
3434 analyze_brprob.py script which is enabled with -details
3436 * profile-count.h (precise_p): New function.
3438 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3440 PR tree-optimization/83922
3441 * tree-vect-loop.c (vect_verify_full_masking): Return false if
3442 there are no statements that need masking.
3443 (vect_active_double_reduction_p): New function.
3444 (vect_analyze_loop_operations): Use it when handling phis that
3445 are not in the loop header.
3447 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3449 PR tree-optimization/83914
3450 * tree-vect-loop.c (vectorizable_induction): Don't convert
3451 init_expr or apply the peeling adjustment for inductions
3452 that are nested within the vectorized loop.
3454 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3456 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
3459 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3463 * function.h (gimplify_parameters): Add gimple_seq * argument.
3464 * function.c: Include gimple.h and options.h.
3465 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
3466 for the added local temporaries if needed.
3467 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
3468 if there are any parameter cleanups, wrap whole body into a
3469 try/finally with the cleanups.
3471 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
3474 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
3475 Use GET_MODE_CLASS for scalar floating point.
3477 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3481 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
3482 Fix call of call_cgraph_insertion_hooks.
3484 2018-01-18 Martin Sebor <msebor@redhat.com>
3486 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
3488 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3491 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
3494 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
3497 * common.opt: (-ffile-prefix-map): New option.
3498 * opts.c (common_handle_option): Defer it.
3499 * opts-global.c (handle_common_deferred_options): Handle it.
3500 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
3501 * file-prefix-map.h: New file.
3502 (remap_debug_filename, add_debug_prefix_map): ...here.
3503 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
3504 * final.c (debug_prefix_map, add_debug_prefix_map
3505 remap_debug_filename): Move to...
3506 * file-prefix-map.c: New file.
3507 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
3508 generalize, get rid of alloca(), use strrchr() instead of strchr().
3509 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
3510 Implement in terms of add_prefix_map().
3511 (remap_macro_filename, remap_debug_filename): Implement in term of
3513 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
3514 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
3515 * dbxout.c: Include file-prefix-map.h.
3516 * varasm.c: Likewise.
3517 * vmsdbgout.c: Likewise.
3518 * xcoffout.c: Likewise.
3519 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
3520 * doc/cppopts.texi (-fmacro-prefix-map): Document.
3521 * doc/invoke.texi (-ffile-prefix-map): Document.
3522 (-fdebug-prefix-map): Update description.
3524 2018-01-18 Martin Liska <mliska@suse.cz>
3526 * config/i386/i386.c (indirect_thunk_name): Document that also
3528 (output_indirect_thunk): Document why both instructions
3529 (pause and lfence) are generated.
3531 2018-01-18 Richard Biener <rguenther@suse.de>
3533 PR tree-optimization/83887
3534 * graphite-scop-detection.c
3535 (scop_detection::get_nearest_dom_with_single_entry): Remove.
3536 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
3537 (scop_detection::merge_sese): Re-implement with a flood-fill
3538 algorithm that properly finds a SESE region if it exists.
3540 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3543 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
3544 pointer_diff optimizations use view_convert instead of convert.
3546 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3548 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3549 Generate different code for -mno-speculate-indirect-jumps.
3550 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3551 (*call_indirect_aix<mode>): Disable for
3552 -mno-speculate-indirect-jumps.
3553 (*call_indirect_aix<mode>_nospec): New define_insn.
3554 (*call_value_indirect_aix<mode>): Disable for
3555 -mno-speculate-indirect-jumps.
3556 (*call_value_indirect_aix<mode>_nospec): New define_insn.
3557 (*sibcall_nonlocal_sysv<mode>): Generate different code for
3558 -mno-speculate-indirect-jumps.
3559 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3561 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3563 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
3564 long double type, set the flags for noting the default long double
3565 type, even if we don't pass or return a long double type.
3567 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
3570 * ipa-inline.c (flatten_function): Do not overwrite final inlining
3573 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
3575 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
3576 support for merge[hl].
3577 (fold_mergehl_helper): New helper function.
3578 (tree-vector-builder.h): New #include for tree_vector_builder usage.
3579 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
3580 (altivec_vmrglw_direct): Add xxmrglw insn.
3582 2018-01-17 Andrew Waterman <andrew@sifive.com>
3584 * config/riscv/riscv.c (riscv_conditional_register_usage): If
3585 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
3587 2018-01-17 David Malcolm <dmalcolm@redhat.com>
3590 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
3591 call the lto_location_cache before reading the
3592 DECL_SOURCE_LOCATION of the types.
3594 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3595 Richard Sandiford <richard.sandiford@linaro.org>
3597 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
3598 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
3599 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
3600 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
3601 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
3603 * config/aarch64/constraints.md (aarch64_movti_operand):
3605 * config/aarch64/predicates.md (Uti): Add new constraint.
3607 2018-01-17 Carl Love <cel@us.ibm.com>
3609 * config/rs6000/vsx.md (define_expand xl_len_r,
3610 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
3611 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
3613 (define_expand, define_insn): Move the shift left from the
3614 define_insn to the define_expand for lxvl and stxvl instructions.
3615 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
3616 and XL_LEN_R definitions to PURE.
3618 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
3620 * config/i386/i386.c (indirect_thunk_name): Declare regno
3621 as unsigned int. Compare regno with INVALID_REGNUM.
3622 (output_indirect_thunk): Ditto.
3623 (output_indirect_thunk_function): Ditto.
3624 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
3625 in the call to output_indirect_thunk_function.
3627 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
3630 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
3631 rather than the size of inner_type to determine the stack slot size
3632 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
3634 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
3637 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
3640 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
3642 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
3643 endian Linux systems to optionally enable multilibs for selecting
3644 the long double type if the user configured an explicit type.
3645 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
3646 have no long double multilibs if not defined.
3647 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
3648 warn if the user used -mabi={ieee,ibm}longdouble and we built
3649 multilibs for long double.
3650 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
3651 appropriate multilib option.
3652 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
3654 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
3655 for building long double multilibs.
3656 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
3658 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
3660 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
3663 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
3665 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
3668 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
3671 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
3674 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3676 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
3677 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
3679 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
3681 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
3682 different rtl trees depending on TARGET_64BIT.
3683 (rs6000_gen_lvx): Likewise.
3685 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3687 * config/visium/visium.md (nop): Tweak comment.
3688 (hazard_nop): Likewise.
3690 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3692 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
3693 -mspeculate-indirect-jumps.
3694 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
3695 for -mno-speculate-indirect-jumps.
3696 (*call_indirect_elfv2<mode>_nospec): New define_insn.
3697 (*call_value_indirect_elfv2<mode>): Disable for
3698 -mno-speculate-indirect-jumps.
3699 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
3700 (indirect_jump): Emit different RTL for
3701 -mno-speculate-indirect-jumps.
3702 (*indirect_jump<mode>): Disable for
3703 -mno-speculate-indirect-jumps.
3704 (*indirect_jump<mode>_nospec): New define_insn.
3705 (tablejump): Emit different RTL for
3706 -mno-speculate-indirect-jumps.
3707 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
3708 (tablejumpsi_nospec): New define_expand.
3709 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
3710 (tablejumpdi_nospec): New define_expand.
3711 (*tablejump<mode>_internal1): Disable for
3712 -mno-speculate-indirect-jumps.
3713 (*tablejump<mode>_internal1_nospec): New define_insn.
3714 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
3717 2018-01-16 Artyom Skrobov tyomitch@gmail.com
3719 * caller-save.c (insert_save): Drop unnecessary parameter. All
3722 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3723 Richard Biener <rguenth@suse.de>
3726 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
3727 return early, inline manually is_gimple_sizepos. Make sure if we
3728 call gimplify_expr we don't end up with a gimple constant.
3729 * tree.c (variably_modified_type_p): Don't return true for
3730 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
3731 * gimplify.h (is_gimple_sizepos): Remove.
3733 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3735 PR tree-optimization/83857
3736 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
3737 vectorizable_live_operation for pure SLP statements.
3738 (vectorizable_live_operation): Handle PHIs.
3740 2018-01-16 Richard Biener <rguenther@suse.de>
3742 PR tree-optimization/83867
3743 * tree-vect-stmts.c (vect_transform_stmt): Precompute
3744 nested_in_vect_loop_p since the scalar stmt may get invalidated.
3746 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3749 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
3750 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
3751 If off is not INTEGER_CST, issue a may not be aligned warning
3752 rather than isn't aligned. Use isn%'t rather than isn't.
3753 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
3755 <case MULT_EXPR>: Improve the case when bottom and one of the
3756 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
3757 operand, in that case check if the other operand is multiple of
3758 bottom divided by the INTEGER_CST operand.
3760 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3763 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
3764 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
3765 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
3766 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
3767 * config/pa/pa.c (pa_function_arg_advance): Likewise.
3768 (pa_function_arg, pa_arg_partial_bytes): Likewise.
3769 (pa_function_arg_size): New function.
3771 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3773 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
3774 in a separate statement.
3776 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3778 PR tree-optimization/83847
3779 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
3780 group gathers and scatters.
3782 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3784 PR rtl-optimization/86620
3785 * params.def (max-sched-ready-insns): Bump minimum value to 1.
3787 PR rtl-optimization/83213
3788 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
3789 to last if both are JUMP_INSNs.
3791 PR tree-optimization/83843
3792 * gimple-ssa-store-merging.c
3793 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
3794 store_immediate_info for bswap/nop orig_stores.
3796 2018-01-15 Andrew Waterman <andrew@sifive.com>
3798 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
3800 <UDIV>: Increase cost if !TARGET_DIV.
3802 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
3804 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
3805 (define_attr "cr_logical_3op"): New.
3806 (cceq_ior_compare): Adjust.
3807 (cceq_ior_compare_complement): Adjust.
3808 (*cceq_rev_compare): Adjust.
3809 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
3810 (is_cracked_insn): Adjust.
3811 (insn_must_be_first_in_group): Adjust.
3812 * config/rs6000/40x.md: Adjust.
3813 * config/rs6000/440.md: Adjust.
3814 * config/rs6000/476.md: Adjust.
3815 * config/rs6000/601.md: Adjust.
3816 * config/rs6000/603.md: Adjust.
3817 * config/rs6000/6xx.md: Adjust.
3818 * config/rs6000/7450.md: Adjust.
3819 * config/rs6000/7xx.md: Adjust.
3820 * config/rs6000/8540.md: Adjust.
3821 * config/rs6000/cell.md: Adjust.
3822 * config/rs6000/e300c2c3.md: Adjust.
3823 * config/rs6000/e500mc.md: Adjust.
3824 * config/rs6000/e500mc64.md: Adjust.
3825 * config/rs6000/e5500.md: Adjust.
3826 * config/rs6000/e6500.md: Adjust.
3827 * config/rs6000/mpc.md: Adjust.
3828 * config/rs6000/power4.md: Adjust.
3829 * config/rs6000/power5.md: Adjust.
3830 * config/rs6000/power6.md: Adjust.
3831 * config/rs6000/power7.md: Adjust.
3832 * config/rs6000/power8.md: Adjust.
3833 * config/rs6000/power9.md: Adjust.
3834 * config/rs6000/rs64.md: Adjust.
3835 * config/rs6000/titan.md: Adjust.
3837 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3839 * config/i386/predicates.md (indirect_branch_operand): Rewrite
3840 ix86_indirect_branch_register logic.
3842 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3844 * config/i386/constraints.md (Bs): Update
3845 ix86_indirect_branch_register check. Don't check
3846 ix86_indirect_branch_register with GOT_memory_operand.
3848 * config/i386/predicates.md (GOT_memory_operand): Don't check
3849 ix86_indirect_branch_register here.
3850 (GOT32_symbol_operand): Likewise.
3852 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3854 * config/i386/predicates.md (constant_call_address_operand):
3855 Rewrite ix86_indirect_branch_register logic.
3856 (sibcall_insn_operand): Likewise.
3858 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3860 * config/i386/constraints.md (Bs): Replace
3861 ix86_indirect_branch_thunk_register with
3862 ix86_indirect_branch_register.
3864 * config/i386/i386.md (indirect_jump): Likewise.
3865 (tablejump): Likewise.
3866 (*sibcall_memory): Likewise.
3867 (*sibcall_value_memory): Likewise.
3868 Peepholes of indirect call and jump via memory: Likewise.
3869 * config/i386/i386.opt: Likewise.
3870 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3871 (GOT_memory_operand): Likewise.
3872 (call_insn_operand): Likewise.
3873 (sibcall_insn_operand): Likewise.
3874 (GOT32_symbol_operand): Likewise.
3876 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3879 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
3880 type rather than type addr's type points to.
3881 (expand_omp_atomic_mutex): Likewise.
3882 (expand_omp_atomic): Likewise.
3884 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3887 * config/i386/i386.c (output_indirect_thunk_function): Use
3888 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
3889 for __x86_return_thunk.
3891 2018-01-15 Richard Biener <rguenther@suse.de>
3894 * expmed.c (extract_bit_field_1): Fix typo.
3896 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3899 * config/arm/iterators.md (VF): New mode iterator.
3900 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
3901 Remove integer-related logic from pattern.
3902 (neon_vabd<mode>_3): Likewise.
3904 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3907 * common.opt (fstrict-overflow): No longer an alias.
3908 (fwrapv-pointer): New option.
3909 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
3910 also for pointer types based on flag_wrapv_pointer.
3911 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
3912 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
3913 opts->x_flag_wrapv got set.
3914 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
3915 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
3916 POINTER_TYPE_OVERFLOW_UNDEFINED.
3917 * match.pd: Likewise in address comparison pattern.
3918 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
3920 2018-01-15 Richard Biener <rguenther@suse.de>
3923 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
3924 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
3925 Reset type names to their identifier if their TYPE_DECL doesn't
3926 have linkage (and thus is used for ODR and devirt).
3927 (save_debug_info_for_decl): Remove.
3928 (save_debug_info_for_type): Likewise.
3929 (add_tree_to_fld_list): Adjust.
3930 * tree-pretty-print.c (dump_generic_node): Make dumping of
3931 type names more robust.
3933 2018-01-15 Richard Biener <rguenther@suse.de>
3935 * BASE-VER: Bump to 8.0.1.
3937 2018-01-14 Martin Sebor <msebor@redhat.com>
3940 * builtins.c (check_access): Avoid warning when the no-warning bit
3943 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
3945 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
3946 * ira-color (allocno_hard_regs_compare): Likewise.
3948 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
3951 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
3952 Use .pushsection/.popsection.
3954 2018-01-14 Martin Sebor <msebor@redhat.com>
3957 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
3959 2018-01-14 Jakub Jelinek <jakub@redhat.com>
3961 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
3962 entry from extra_headers.
3963 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
3964 extra_headers, make the list bitwise identical to the i?86-*-* one.
3966 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3968 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
3969 -mcmodel=large with -mindirect-branch=thunk,
3970 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
3971 -mfunction-return=thunk-extern.
3972 * doc/invoke.texi: Document -mcmodel=large is incompatible with
3973 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
3974 -mfunction-return=thunk and -mfunction-return=thunk-extern.
3976 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3978 * config/i386/i386.c (print_reg): Print the name of the full
3979 integer register without '%'.
3980 (ix86_print_operand): Handle 'V'.
3981 * doc/extend.texi: Document 'V' modifier.
3983 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3985 * config/i386/constraints.md (Bs): Disallow memory operand for
3986 -mindirect-branch-register.
3988 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3989 (GOT_memory_operand): Likewise.
3990 (call_insn_operand): Likewise.
3991 (sibcall_insn_operand): Likewise.
3992 (GOT32_symbol_operand): Likewise.
3993 * config/i386/i386.md (indirect_jump): Call convert_memory_address
3994 for -mindirect-branch-register.
3995 (tablejump): Likewise.
3996 (*sibcall_memory): Likewise.
3997 (*sibcall_value_memory): Likewise.
3998 Disallow peepholes of indirect call and jump via memory for
3999 -mindirect-branch-register.
4000 (*call_pop): Replace m with Bw.
4001 (*call_value_pop): Likewise.
4002 (*sibcall_pop_memory): Replace m with Bs.
4003 * config/i386/i386.opt (mindirect-branch-register): New option.
4004 * doc/invoke.texi: Document -mindirect-branch-register option.
4006 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4008 * config/i386/i386-protos.h (ix86_output_function_return): New.
4009 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
4010 set function_return_type.
4011 (indirect_thunk_name): Add ret_p to indicate thunk for function
4013 (output_indirect_thunk_function): Pass false to
4014 indirect_thunk_name.
4015 (ix86_output_indirect_branch_via_reg): Likewise.
4016 (ix86_output_indirect_branch_via_push): Likewise.
4017 (output_indirect_thunk_function): Create alias for function
4018 return thunk if regno < 0.
4019 (ix86_output_function_return): New function.
4020 (ix86_handle_fndecl_attribute): Handle function_return.
4021 (ix86_attribute_table): Add function_return.
4022 * config/i386/i386.h (machine_function): Add
4023 function_return_type.
4024 * config/i386/i386.md (simple_return_internal): Use
4025 ix86_output_function_return.
4026 (simple_return_internal_long): Likewise.
4027 * config/i386/i386.opt (mfunction-return=): New option.
4028 (indirect_branch): Mention -mfunction-return=.
4029 * doc/extend.texi: Document function_return function attribute.
4030 * doc/invoke.texi: Document -mfunction-return= option.
4032 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
4034 * config/i386/i386-opts.h (indirect_branch): New.
4035 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
4036 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
4037 with local indirect jump when converting indirect call and jump.
4038 (ix86_set_indirect_branch_type): New.
4039 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
4040 (indirectlabelno): New.
4041 (indirect_thunk_needed): Likewise.
4042 (indirect_thunk_bnd_needed): Likewise.
4043 (indirect_thunks_used): Likewise.
4044 (indirect_thunks_bnd_used): Likewise.
4045 (INDIRECT_LABEL): Likewise.
4046 (indirect_thunk_name): Likewise.
4047 (output_indirect_thunk): Likewise.
4048 (output_indirect_thunk_function): Likewise.
4049 (ix86_output_indirect_branch_via_reg): Likewise.
4050 (ix86_output_indirect_branch_via_push): Likewise.
4051 (ix86_output_indirect_branch): Likewise.
4052 (ix86_output_indirect_jmp): Likewise.
4053 (ix86_code_end): Call output_indirect_thunk_function if needed.
4054 (ix86_output_call_insn): Call ix86_output_indirect_branch if
4056 (ix86_handle_fndecl_attribute): Handle indirect_branch.
4057 (ix86_attribute_table): Add indirect_branch.
4058 * config/i386/i386.h (machine_function): Add indirect_branch_type
4059 and has_local_indirect_jump.
4060 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
4062 (tablejump): Likewise.
4063 (*indirect_jump): Use ix86_output_indirect_jmp.
4064 (*tablejump_1): Likewise.
4065 (simple_return_indirect_internal): Likewise.
4066 * config/i386/i386.opt (mindirect-branch=): New option.
4067 (indirect_branch): New.
4070 (thunk-inline): Likewise.
4071 (thunk-extern): Likewise.
4072 * doc/extend.texi: Document indirect_branch function attribute.
4073 * doc/invoke.texi: Document -mindirect-branch= option.
4075 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
4078 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
4080 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
4082 * ipa-inline.c (want_inline_small_function_p): Return false if
4083 inlining has already failed with CIF_FINAL_ERROR.
4084 (update_caller_keys): Call want_inline_small_function_p before
4086 (update_callee_keys): Likewise.
4088 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
4090 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
4092 (rs6000_quadword_masked_address_p): Likewise.
4093 (quad_aligned_load_p): Likewise.
4094 (quad_aligned_store_p): Likewise.
4095 (const_load_sequence_p): Add comment to describe the outer-most loop.
4096 (mimic_memory_attributes_and_flags): New function.
4097 (rs6000_gen_stvx): Likewise.
4098 (replace_swapped_aligned_store): Likewise.
4099 (rs6000_gen_lvx): Likewise.
4100 (replace_swapped_aligned_load): Likewise.
4101 (replace_swapped_load_constant): Capitalize argument name in
4102 comment describing this function.
4103 (rs6000_analyze_swaps): Add a third pass to search for vector loads
4104 and stores that access quad-word aligned addresses and replace
4105 with stvx or lvx instructions when appropriate.
4106 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
4107 New function prototype.
4108 (rs6000_quadword_masked_address_p): Likewise.
4109 (rs6000_gen_lvx): Likewise.
4110 (rs6000_gen_stvx): Likewise.
4111 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
4112 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
4113 when memory address is aligned.
4114 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
4115 this split to select lvx instruction when memory address is aligned.
4116 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
4117 instruction when memory address is aligned.
4118 (*vsx_le_perm_load_v16qi): Likewise.
4119 (four unnamed splitters): Modify to select the stvx instruction
4120 when memory is aligned.
4122 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
4124 * predict.c (determine_unlikely_bbs): Handle correctly BBs
4125 which appears in the queue multiple times.
4127 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4128 Alan Hayward <alan.hayward@arm.com>
4129 David Sherwood <david.sherwood@arm.com>
4131 * tree-vectorizer.h (vec_lower_bound): New structure.
4132 (_loop_vec_info): Add check_nonzero and lower_bounds.
4133 (LOOP_VINFO_CHECK_NONZERO): New macro.
4134 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
4135 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
4136 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
4137 fields. Make seg_len the distance travelled, not including the
4139 (dr_direction_indicator): Declare.
4140 (dr_zero_step_indicator): Likewise.
4141 (dr_known_forward_stride_p): Likewise.
4142 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
4144 (runtime_alias_check_p): Allow runtime alias checks with
4146 (operator ==): Compare access_size and align.
4147 (prune_runtime_alias_test_list): Rework for new distinction between
4148 the access_size and seg_len.
4149 (create_intersect_range_checks_index): Likewise. Cope with polynomial
4151 (get_segment_min_max): New function.
4152 (create_intersect_range_checks): Use it.
4153 (dr_step_indicator): New function.
4154 (dr_direction_indicator): Likewise.
4155 (dr_zero_step_indicator): Likewise.
4156 (dr_known_forward_stride_p): Likewise.
4157 * tree-loop-distribution.c (data_ref_segment_size): Return
4158 DR_STEP * (niters - 1).
4159 (compute_alias_check_pairs): Update call to the dr_with_seg_len
4161 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
4162 (vect_preserves_scalar_order_p): New function, split out from...
4163 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
4164 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
4165 (vect_vfa_access_size): New function.
4166 (vect_vfa_align): Likewise.
4167 (vect_compile_time_alias): Take access_size_a and access_b arguments.
4168 (dump_lower_bound): New function.
4169 (vect_check_lower_bound): Likewise.
4170 (vect_small_gap_p): Likewise.
4171 (vectorizable_with_step_bound_p): Likewise.
4172 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
4173 depencies if the vectorization factor is 1. Convert the checks
4174 for nonzero steps into checks on the bounds of DR_STEP. Try using
4175 a bunds check for variable steps if the minimum required step is
4176 relatively small. Update calls to the dr_with_seg_len
4177 constructor and to vect_compile_time_alias.
4178 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
4180 (vect_loop_versioning): Call it.
4181 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
4183 (vect_estimate_min_profitable_iters): Account for any bounds checks.
4185 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4186 Alan Hayward <alan.hayward@arm.com>
4187 David Sherwood <david.sherwood@arm.com>
4189 * doc/sourcebuild.texi (vect_scatter_store): Document.
4190 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
4192 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
4194 * genopinit.c (main): Add supports_vec_scatter_store and
4195 supports_vec_scatter_store_cached to target_optabs.
4196 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
4197 IFN_MASK_SCATTER_STORE.
4198 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
4200 * internal-fn.h (internal_store_fn_p): Declare.
4201 (internal_fn_stored_value_index): Likewise.
4202 * internal-fn.c (scatter_store_direct): New macro.
4203 (expand_scatter_store_optab_fn): New function.
4204 (direct_scatter_store_optab_supported_p): New macro.
4205 (internal_store_fn_p): New function.
4206 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
4207 IFN_MASK_SCATTER_STORE.
4208 (internal_fn_mask_index): Likewise.
4209 (internal_fn_stored_value_index): New function.
4210 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
4212 * optabs-query.h (supports_vec_scatter_store_p): Declare.
4213 * optabs-query.c (supports_vec_scatter_store_p): New function.
4214 * tree-vectorizer.h (vect_get_store_rhs): Declare.
4215 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
4216 true for scatter stores.
4217 (vect_gather_scatter_fn_p): Handle scatter stores too.
4218 (vect_check_gather_scatter): Consider using scatter stores if
4219 supports_vec_scatter_store_p.
4220 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
4222 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4223 internal_fn_stored_value_index.
4224 (check_load_store_masking): Handle scatter stores too.
4225 (vect_get_store_rhs): Make public.
4226 (vectorizable_call): Use internal_store_fn_p.
4227 (vectorizable_store): Handle scatter store internal functions.
4228 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
4229 when deciding whether the end of the group has been reached.
4230 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
4231 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
4232 (mask_scatter_store<mode>): New insns.
4234 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4235 Alan Hayward <alan.hayward@arm.com>
4236 David Sherwood <david.sherwood@arm.com>
4238 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
4239 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
4240 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
4242 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
4243 Use vect_truncate_gather_scatter_offset if we can't treat the
4244 operation as a normal gather load or scatter store.
4245 (get_group_load_store_type): Take the gather_scatter_info
4246 as argument. Try using a gather load or scatter store for
4247 single-element groups.
4248 (get_load_store_type): Update calls to get_group_load_store_type
4249 and vect_use_strided_gather_scatters_p.
4251 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4252 Alan Hayward <alan.hayward@arm.com>
4253 David Sherwood <david.sherwood@arm.com>
4255 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
4256 optional tree argument.
4257 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
4259 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
4260 but continue to use the current value as a fallback.
4261 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
4262 to compare the updates.
4263 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
4264 (get_load_store_type): Use it when handling a strided access.
4265 (vect_get_strided_load_store_ops): New function.
4266 (vect_get_data_ptr_increment): Likewise.
4267 (vectorizable_load): Handle strided gather loads. Always pass
4268 a step to vect_create_data_ref_ptr and bump_vector_ptr.
4270 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4271 Alan Hayward <alan.hayward@arm.com>
4272 David Sherwood <david.sherwood@arm.com>
4274 * doc/md.texi (gather_load@var{m}): Document.
4275 (mask_gather_load@var{m}): Likewise.
4276 * genopinit.c (main): Add supports_vec_gather_load and
4277 supports_vec_gather_load_cached to target_optabs.
4278 * optabs-tree.c (init_tree_optimization_optabs): Use
4279 ggc_cleared_alloc to allocate target_optabs.
4280 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
4281 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
4283 * internal-fn.h (internal_load_fn_p): Declare.
4284 (internal_gather_scatter_fn_p): Likewise.
4285 (internal_fn_mask_index): Likewise.
4286 (internal_gather_scatter_fn_supported_p): Likewise.
4287 * internal-fn.c (gather_load_direct): New macro.
4288 (expand_gather_load_optab_fn): New function.
4289 (direct_gather_load_optab_supported_p): New macro.
4290 (direct_internal_fn_optab): New function.
4291 (internal_load_fn_p): Likewise.
4292 (internal_gather_scatter_fn_p): Likewise.
4293 (internal_fn_mask_index): Likewise.
4294 (internal_gather_scatter_fn_supported_p): Likewise.
4295 * optabs-query.c (supports_at_least_one_mode_p): New function.
4296 (supports_vec_gather_load_p): Likewise.
4297 * optabs-query.h (supports_vec_gather_load_p): Declare.
4298 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
4299 and memory_type field.
4300 (NUM_PATTERNS): Bump to 15.
4301 * tree-vect-data-refs.c: Include internal-fn.h.
4302 (vect_gather_scatter_fn_p): New function.
4303 (vect_describe_gather_scatter_call): Likewise.
4304 (vect_check_gather_scatter): Try using internal functions for
4305 gather loads. Recognize existing calls to a gather load function.
4306 (vect_analyze_data_refs): Consider using gather loads if
4307 supports_vec_gather_load_p.
4308 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
4309 (vect_get_gather_scatter_offset_type): Likewise.
4310 (vect_convert_mask_for_vectype): Likewise.
4311 (vect_add_conversion_to_patterm): Likewise.
4312 (vect_try_gather_scatter_pattern): Likewise.
4313 (vect_recog_gather_scatter_pattern): New pattern recognizer.
4314 (vect_vect_recog_func_ptrs): Add it.
4315 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4316 internal_fn_mask_index and internal_gather_scatter_fn_p.
4317 (check_load_store_masking): Take the gather_scatter_info as an
4318 argument and handle gather loads.
4319 (vect_get_gather_scatter_ops): New function.
4320 (vectorizable_call): Check internal_load_fn_p.
4321 (vectorizable_load): Likewise. Handle gather load internal
4323 (vectorizable_store): Update call to check_load_store_masking.
4324 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
4325 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
4326 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
4327 (aarch64_gather_scale_operand_d): New predicates.
4328 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
4329 (mask_gather_load<mode>): New insns.
4331 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4332 Alan Hayward <alan.hayward@arm.com>
4333 David Sherwood <david.sherwood@arm.com>
4335 * optabs.def (fold_left_plus_optab): New optab.
4336 * doc/md.texi (fold_left_plus_@var{m}): Document.
4337 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
4338 * internal-fn.c (fold_left_direct): Define.
4339 (expand_fold_left_optab_fn): Likewise.
4340 (direct_fold_left_optab_supported_p): Likewise.
4341 * fold-const-call.c (fold_const_fold_left): New function.
4342 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
4343 * tree-parloops.c (valid_reduction_p): New function.
4344 (gather_scalar_reductions): Use it.
4345 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
4346 (vect_finish_replace_stmt): Declare.
4347 * tree-vect-loop.c (fold_left_reduction_fn): New function.
4348 (needs_fold_left_reduction_p): New function, split out from...
4349 (vect_is_simple_reduction): ...here. Accept reductions that
4350 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
4351 (vect_force_simple_reduction): Also store the reduction type in
4352 the assignment's STMT_VINFO_REDUC_TYPE.
4353 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
4354 (merge_with_identity): New function.
4355 (vect_expand_fold_left): Likewise.
4356 (vectorize_fold_left_reduction): Likewise.
4357 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
4358 scalar phi in place for it. Check for target support and reject
4359 cases that would reassociate the operation. Defer the transform
4360 phase to vectorize_fold_left_reduction.
4361 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
4362 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
4363 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
4365 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4367 * tree-if-conv.c (predicate_mem_writes): Remove redundant
4368 call to ifc_temp_var.
4370 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4371 Alan Hayward <alan.hayward@arm.com>
4372 David Sherwood <david.sherwood@arm.com>
4374 * target.def (legitimize_address_displacement): Take the original
4375 offset as a poly_int.
4376 * targhooks.h (default_legitimize_address_displacement): Update
4378 * targhooks.c (default_legitimize_address_displacement): Likewise.
4379 * doc/tm.texi: Regenerate.
4380 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
4381 as an argument, moving assert of ad->disp == ad->disp_term to...
4382 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
4383 Try calling targetm.legitimize_address_displacement before expanding
4384 the address rather than afterwards, and adjust for the new interface.
4385 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
4386 Match the new hook interface. Handle SVE addresses.
4387 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
4390 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4392 * Makefile.in (OBJS): Add early-remat.o.
4393 * target.def (select_early_remat_modes): New hook.
4394 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
4395 * doc/tm.texi: Regenerate.
4396 * targhooks.h (default_select_early_remat_modes): Declare.
4397 * targhooks.c (default_select_early_remat_modes): New function.
4398 * timevar.def (TV_EARLY_REMAT): New timevar.
4399 * passes.def (pass_early_remat): New pass.
4400 * tree-pass.h (make_pass_early_remat): Declare.
4401 * early-remat.c: New file.
4402 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
4404 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
4406 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4407 Alan Hayward <alan.hayward@arm.com>
4408 David Sherwood <david.sherwood@arm.com>
4410 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
4411 vfm1 with a bound_epilog parameter.
4412 (vect_do_peeling): Update calls accordingly, and move the prologue
4413 call earlier in the function. Treat the base bound_epilog as 0 for
4414 fully-masked loops and retain vf - 1 for other loops. Add 1 to
4415 this base when peeling for gaps.
4416 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
4417 with fully-masked loops.
4418 (vect_estimate_min_profitable_iters): Handle the single peeled
4419 iteration in that case.
4421 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4422 Alan Hayward <alan.hayward@arm.com>
4423 David Sherwood <david.sherwood@arm.com>
4425 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
4426 single-element interleaving even if the size is not a power of 2.
4427 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
4428 accesses for single-element interleaving if the group size is
4431 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4432 Alan Hayward <alan.hayward@arm.com>
4433 David Sherwood <david.sherwood@arm.com>
4435 * doc/md.texi (fold_extract_last_@var{m}): Document.
4436 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
4437 * optabs.def (fold_extract_last_optab): New optab.
4438 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
4439 * internal-fn.c (fold_extract_direct): New macro.
4440 (expand_fold_extract_optab_fn): Likewise.
4441 (direct_fold_extract_optab_supported_p): Likewise.
4442 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
4443 * tree-vect-loop.c (vect_model_reduction_cost): Handle
4444 EXTRACT_LAST_REDUCTION.
4445 (get_initial_def_for_reduction): Do not create an initial vector
4446 for EXTRACT_LAST_REDUCTION reductions.
4447 (vectorizable_reduction): Leave the scalar phi in place for
4448 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
4449 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
4450 epilogue code for EXTRACT_LAST_REDUCTION and defer the
4451 transform phase to vectorizable_condition.
4452 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
4454 (vect_finish_stmt_generation): ...here.
4455 (vect_finish_replace_stmt): New function.
4456 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
4457 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
4459 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
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 * doc/md.texi (extract_last_@var{m}): Document.
4466 * optabs.def (extract_last_optab): New optab.
4467 * internal-fn.def (EXTRACT_LAST): New internal function.
4468 * internal-fn.c (cond_unary_direct): New macro.
4469 (expand_cond_unary_optab_fn): Likewise.
4470 (direct_cond_unary_optab_supported_p): Likewise.
4471 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
4472 loops using EXTRACT_LAST.
4473 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
4474 (extract_last_<mode>): ...this optab.
4475 (vec_extract<mode><Vel>): Update accordingly.
4477 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4478 Alan Hayward <alan.hayward@arm.com>
4479 David Sherwood <david.sherwood@arm.com>
4481 * target.def (empty_mask_is_expensive): New hook.
4482 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
4483 * doc/tm.texi: Regenerate.
4484 * targhooks.h (default_empty_mask_is_expensive): Declare.
4485 * targhooks.c (default_empty_mask_is_expensive): New function.
4486 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
4487 if the target says that empty masks are expensive.
4488 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
4490 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
4492 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4493 Alan Hayward <alan.hayward@arm.com>
4494 David Sherwood <david.sherwood@arm.com>
4496 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
4497 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
4498 (vect_use_loop_mask_for_alignment_p): New function.
4499 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
4500 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
4501 niters_skip argument. Make sure that the first niters_skip elements
4502 of the first iteration are inactive.
4503 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
4504 Update call to vect_set_loop_masks_directly.
4505 (get_misalign_in_elems): New function, split out from...
4506 (vect_gen_prolog_loop_niters): ...here.
4507 (vect_update_init_of_dr): Take a code argument that specifies whether
4508 the adjustment should be added or subtracted.
4509 (vect_update_init_of_drs): Likewise.
4510 (vect_prepare_for_masked_peels): New function.
4511 (vect_do_peeling): Skip prologue peeling if we're using a mask
4512 instead. Update call to vect_update_inits_of_drs.
4513 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4515 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
4516 alignment. Do not include the number of peeled iterations in
4517 the minimum threshold in that case.
4518 (vectorizable_induction): Adjust the start value down by
4519 LOOP_VINFO_MASK_SKIP_NITERS iterations.
4520 (vect_transform_loop): Call vect_prepare_for_masked_peels.
4521 Take the number of skipped iterations into account when calculating
4523 * tree-vect-stmts.c (vect_gen_while_not): New function.
4525 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4526 Alan Hayward <alan.hayward@arm.com>
4527 David Sherwood <david.sherwood@arm.com>
4529 * doc/sourcebuild.texi (vect_fully_masked): Document.
4530 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
4532 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
4534 (vect_analyze_loop_2): ...here. Don't check the vectorization
4535 factor against the number of loop iterations if the loop is
4538 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4539 Alan Hayward <alan.hayward@arm.com>
4540 David Sherwood <david.sherwood@arm.com>
4542 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
4543 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
4544 (dump_groups): Update accordingly.
4545 (iv_use::mem_type): New member variable.
4546 (address_p): New function.
4547 (record_use): Add a mem_type argument and initialize the new
4549 (record_group_use): Add a mem_type argument. Use address_p.
4550 Remove obsolete null checks of base_object. Update call to record_use.
4551 (find_interesting_uses_op): Update call to record_group_use.
4552 (find_interesting_uses_cond): Likewise.
4553 (find_interesting_uses_address): Likewise.
4554 (get_mem_type_for_internal_fn): New function.
4555 (find_address_like_use): Likewise.
4556 (find_interesting_uses_stmt): Try find_address_like_use before
4557 calling find_interesting_uses_op.
4558 (addr_offset_valid_p): Use the iv mem_type field as the type
4559 of the addressed memory.
4560 (add_autoinc_candidates): Likewise.
4561 (get_address_cost): Likewise.
4562 (split_small_address_groups_p): Use address_p.
4563 (split_address_groups): Likewise.
4564 (add_iv_candidate_for_use): Likewise.
4565 (autoinc_possible_for_pair): Likewise.
4566 (rewrite_groups): Likewise.
4567 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
4568 (determine_group_iv_cost): Update after split of USE_ADDRESS.
4569 (get_alias_ptr_type_for_ptr_address): New function.
4570 (rewrite_use_address): Rewrite address uses in calls that were
4571 identified by find_address_like_use.
4573 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4574 Alan Hayward <alan.hayward@arm.com>
4575 David Sherwood <david.sherwood@arm.com>
4577 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
4579 * gimple-expr.h (is_gimple_addressable: Likewise.
4580 * gimple-expr.c (is_gimple_address): Likewise.
4581 * internal-fn.c (expand_call_mem_ref): New function.
4582 (expand_mask_load_optab_fn): Use it.
4583 (expand_mask_store_optab_fn): Likewise.
4585 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4586 Alan Hayward <alan.hayward@arm.com>
4587 David Sherwood <david.sherwood@arm.com>
4589 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
4590 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
4591 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
4592 (cond_umax@var{mode}): Document.
4593 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
4594 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
4595 (cond_umin_optab, cond_umax_optab): New optabs.
4596 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
4597 (COND_IOR, COND_XOR): New internal functions.
4598 * internal-fn.h (get_conditional_internal_fn): Declare.
4599 * internal-fn.c (cond_binary_direct): New macro.
4600 (expand_cond_binary_optab_fn): Likewise.
4601 (direct_cond_binary_optab_supported_p): Likewise.
4602 (get_conditional_internal_fn): New function.
4603 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
4604 Cope with reduction statements that are vectorized as calls rather
4606 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
4607 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
4608 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
4609 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4610 (UNSPEC_COND_EOR): New unspecs.
4611 (optab): Add mappings for them.
4612 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
4613 (sve_int_op, sve_fp_op): New int attributes.
4615 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4616 Alan Hayward <alan.hayward@arm.com>
4617 David Sherwood <david.sherwood@arm.com>
4619 * optabs.def (while_ult_optab): New optab.
4620 * doc/md.texi (while_ult@var{m}@var{n}): Document.
4621 * internal-fn.def (WHILE_ULT): New internal function.
4622 * internal-fn.h (direct_internal_fn_supported_p): New override
4623 that takes two types as argument.
4624 * internal-fn.c (while_direct): New macro.
4625 (expand_while_optab_fn): New function.
4626 (convert_optab_supported_p): Likewise.
4627 (direct_while_optab_supported_p): New macro.
4628 * wide-int.h (wi::udiv_ceil): New function.
4629 * tree-vectorizer.h (rgroup_masks): New structure.
4630 (vec_loop_masks): New typedef.
4631 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
4633 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
4634 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
4635 (vect_max_vf): New function.
4636 (slpeel_make_loop_iterate_ntimes): Delete.
4637 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
4638 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
4639 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
4640 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
4641 internal-fn.h, stor-layout.h and optabs-query.h.
4642 (vect_set_loop_mask): New function.
4643 (add_preheader_seq): Likewise.
4644 (add_header_seq): Likewise.
4645 (interleave_supported_p): Likewise.
4646 (vect_maybe_permute_loop_masks): Likewise.
4647 (vect_set_loop_masks_directly): Likewise.
4648 (vect_set_loop_condition_masked): Likewise.
4649 (vect_set_loop_condition_unmasked): New function, split out from
4650 slpeel_make_loop_iterate_ntimes.
4651 (slpeel_make_loop_iterate_ntimes): Rename to..
4652 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
4653 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
4654 (vect_do_peeling): Update call accordingly.
4655 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
4657 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4658 mask_compare_type, can_fully_mask_p and fully_masked_p.
4659 (release_vec_loop_masks): New function.
4660 (_loop_vec_info): Use it to free the loop masks.
4661 (can_produce_all_loop_masks_p): New function.
4662 (vect_get_max_nscalars_per_iter): Likewise.
4663 (vect_verify_full_masking): Likewise.
4664 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
4665 retries, and free the mask rgroups before retrying. Check loop-wide
4666 reasons for disallowing fully-masked loops. Make the final decision
4667 about whether use a fully-masked loop or not.
4668 (vect_estimate_min_profitable_iters): Do not assume that peeling
4669 for the number of iterations will be needed for fully-masked loops.
4670 (vectorizable_reduction): Disable fully-masked loops.
4671 (vectorizable_live_operation): Likewise.
4672 (vect_halve_mask_nunits): New function.
4673 (vect_double_mask_nunits): Likewise.
4674 (vect_record_loop_mask): Likewise.
4675 (vect_get_loop_mask): Likewise.
4676 (vect_transform_loop): Handle the case in which the final loop
4677 iteration might handle a partial vector. Call vect_set_loop_condition
4678 instead of slpeel_make_loop_iterate_ntimes.
4679 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
4680 (check_load_store_masking): New function.
4681 (prepare_load_store_mask): Likewise.
4682 (vectorizable_store): Handle fully-masked loops.
4683 (vectorizable_load): Likewise.
4684 (supportable_widening_operation): Use vect_halve_mask_nunits for
4686 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
4687 (vect_gen_while): New function.
4688 * config/aarch64/aarch64.md (umax<mode>3): New expander.
4689 (aarch64_uqdec<mode>): New insn.
4691 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4692 Alan Hayward <alan.hayward@arm.com>
4693 David Sherwood <david.sherwood@arm.com>
4695 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
4696 (reduc_xor_scal_optab): New optabs.
4697 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
4698 (reduc_xor_scal_@var{m}): Document.
4699 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
4700 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
4702 * fold-const-call.c (fold_const_call): Handle them.
4703 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
4704 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
4705 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
4706 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
4707 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
4708 (UNSPEC_XORV): New unspecs.
4709 (optab): Add entries for them.
4710 (BITWISEV): New int iterator.
4711 (bit_reduc_op): New int attributes.
4713 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4714 Alan Hayward <alan.hayward@arm.com>
4715 David Sherwood <david.sherwood@arm.com>
4717 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
4718 * internal-fn.def (VEC_SHL_INSERT): New internal function.
4719 * optabs.def (vec_shl_insert_optab): New optab.
4720 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
4721 (duplicate_and_interleave): Likewise.
4722 * tree-vect-loop.c: Include internal-fn.h.
4723 (neutral_op_for_slp_reduction): New function, split out from
4724 get_initial_defs_for_reduction.
4725 (get_initial_def_for_reduction): Handle option 2 for variable-length
4726 vectors by loading the neutral value into a vector and then shifting
4727 the initial value into element 0.
4728 (get_initial_defs_for_reduction): Replace the code argument with
4729 the neutral value calculated by neutral_op_for_slp_reduction.
4730 Use gimple_build_vector for constant-length vectors.
4731 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
4732 but the first group_size elements have a neutral value.
4733 Use duplicate_and_interleave otherwise.
4734 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
4735 Update call to get_initial_defs_for_reduction. Handle SLP
4736 reductions for variable-length vectors by creating one vector
4737 result for each scalar result, with the elements associated
4738 with other scalar results stubbed out with the neutral value.
4739 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
4740 Require IFN_VEC_SHL_INSERT for double reductions on
4741 variable-length vectors, or SLP reductions that have
4742 a neutral value. Require can_duplicate_and_interleave_p
4743 support for variable-length unchained SLP reductions if there
4744 is no neutral value, such as for MIN/MAX reductions. Also require
4745 the number of vector elements to be a multiple of the number of
4746 SLP statements when doing variable-length unchained SLP reductions.
4747 Update call to vect_create_epilog_for_reduction.
4748 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
4749 and remove initial values.
4750 (duplicate_and_interleave): Make public.
4751 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
4752 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
4754 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4755 Alan Hayward <alan.hayward@arm.com>
4756 David Sherwood <david.sherwood@arm.com>
4758 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
4759 (can_duplicate_and_interleave_p): New function.
4760 (vect_get_and_check_slp_defs): Take the vector of statements
4761 rather than just the current one. Remove excess parentheses.
4762 Restriction rejectinon of vect_constant_def and vect_external_def
4763 for variable-length vectors to boolean types, or types for which
4764 can_duplicate_and_interleave_p is false.
4765 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
4766 (duplicate_and_interleave): New function.
4767 (vect_get_constant_vectors): Use gimple_build_vector for
4768 constant-length vectors and suitable variable-length constant
4769 vectors. Use duplicate_and_interleave for other variable-length
4770 vectors. Don't defer the update when inserting new statements.
4772 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4773 Alan Hayward <alan.hayward@arm.com>
4774 David Sherwood <david.sherwood@arm.com>
4776 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
4777 min_profitable_iters doesn't go negative.
4779 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4780 Alan Hayward <alan.hayward@arm.com>
4781 David Sherwood <david.sherwood@arm.com>
4783 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
4784 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
4785 * optabs.def (vec_mask_load_lanes_optab): New optab.
4786 (vec_mask_store_lanes_optab): Likewise.
4787 * internal-fn.def (MASK_LOAD_LANES): New internal function.
4788 (MASK_STORE_LANES): Likewise.
4789 * internal-fn.c (mask_load_lanes_direct): New macro.
4790 (mask_store_lanes_direct): Likewise.
4791 (expand_mask_load_optab_fn): Handle masked operations.
4792 (expand_mask_load_lanes_optab_fn): New macro.
4793 (expand_mask_store_optab_fn): Handle masked operations.
4794 (expand_mask_store_lanes_optab_fn): New macro.
4795 (direct_mask_load_lanes_optab_supported_p): Likewise.
4796 (direct_mask_store_lanes_optab_supported_p): Likewise.
4797 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
4799 (vect_load_lanes_supported): Likewise.
4800 * tree-vect-data-refs.c (strip_conversion): New function.
4801 (can_group_stmts_p): Likewise.
4802 (vect_analyze_data_ref_accesses): Use it instead of checking
4803 for a pair of assignments.
4804 (vect_store_lanes_supported): Take a masked_p parameter.
4805 (vect_load_lanes_supported): Likewise.
4806 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
4807 vect_store_lanes_supported and vect_load_lanes_supported.
4808 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4809 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
4810 parameter. Don't allow gaps for masked accesses.
4811 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
4812 and vect_load_lanes_supported.
4813 (get_load_store_type): Take a masked_p parameter and update
4814 call to get_group_load_store_type.
4815 (vectorizable_store): Update call to get_load_store_type.
4816 Handle IFN_MASK_STORE_LANES.
4817 (vectorizable_load): Update call to get_load_store_type.
4818 Handle IFN_MASK_LOAD_LANES.
4820 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4821 Alan Hayward <alan.hayward@arm.com>
4822 David Sherwood <david.sherwood@arm.com>
4824 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
4826 * config/aarch64/aarch64-protos.h
4827 (aarch64_sve_struct_memory_operand_p): Declare.
4828 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
4829 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
4830 (VPRED, vpred): Handle SVE structure modes.
4831 * config/aarch64/constraints.md (Utx): New constraint.
4832 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
4833 (aarch64_sve_struct_nonimmediate_operand): New predicates.
4834 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
4835 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
4836 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
4837 structure modes. Split into pieces after RA.
4838 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
4839 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
4841 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
4842 SVE structure modes.
4843 (aarch64_classify_address): Likewise.
4844 (sizetochar): Move earlier in file.
4845 (aarch64_print_operand): Handle SVE register lists.
4846 (aarch64_array_mode): New function.
4847 (aarch64_sve_struct_memory_operand_p): Likewise.
4848 (TARGET_ARRAY_MODE): Redefine.
4850 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4851 Alan Hayward <alan.hayward@arm.com>
4852 David Sherwood <david.sherwood@arm.com>
4854 * target.def (array_mode): New target hook.
4855 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
4856 * doc/tm.texi: Regenerate.
4857 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
4858 * hooks.c (hook_optmode_mode_uhwi_none): New function.
4859 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
4861 * stor-layout.c (mode_for_array): Likewise. Support polynomial
4864 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4865 Alan Hayward <alan.hayward@arm.com>
4866 David Sherwood <david.sherwood@arm.com>
4868 * fold-const.c (fold_binary_loc): Check the argument types
4869 rather than the result type when testing for a vector operation.
4871 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4873 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
4874 * doc/tm.texi: Regenerate.
4876 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4877 Alan Hayward <alan.hayward@arm.com>
4878 David Sherwood <david.sherwood@arm.com>
4880 * doc/invoke.texi (-msve-vector-bits=): Document new option.
4881 (sve): Document new AArch64 extension.
4882 * doc/md.texi (w): Extend the description of the AArch64
4883 constraint to include SVE vectors.
4884 (Upl, Upa): Document new AArch64 predicate constraints.
4885 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
4887 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
4888 (msve-vector-bits=): New option.
4889 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
4890 SVE when these are disabled.
4891 (sve): New extension.
4892 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
4893 modes. Adjust their number of units based on aarch64_sve_vg.
4894 (MAX_BITSIZE_MODE_ANY_MODE): Define.
4895 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
4896 aarch64_addr_query_type.
4897 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
4898 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
4899 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
4900 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
4901 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
4902 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
4903 (aarch64_simd_imm_zero_p): Delete.
4904 (aarch64_check_zero_based_sve_index_immediate): Declare.
4905 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4906 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4907 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4908 (aarch64_sve_float_mul_immediate_p): Likewise.
4909 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4911 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
4912 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
4913 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
4914 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
4915 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
4916 (aarch64_regmode_natural_size): Likewise.
4917 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
4918 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
4920 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
4921 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
4922 for VG and the SVE predicate registers.
4923 (V_ALIASES): Add a "z"-prefixed alias.
4924 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
4925 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
4926 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
4927 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
4928 (REG_CLASS_NAMES): Add entries for them.
4929 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
4930 and the predicate registers.
4931 (aarch64_sve_vg): Declare.
4932 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
4933 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
4934 (REGMODE_NATURAL_SIZE): Define.
4935 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
4937 * config/aarch64/aarch64.c: Include cfgrtl.h.
4938 (simd_immediate_info): Add a constructor for series vectors,
4939 and an associated step field.
4940 (aarch64_sve_vg): New variable.
4941 (aarch64_dbx_register_number): Handle VG and the predicate registers.
4942 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
4943 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
4944 (VEC_ANY_DATA, VEC_STRUCT): New constants.
4945 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
4946 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
4947 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
4948 (aarch64_get_mask_mode): New functions.
4949 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
4950 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
4951 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
4952 predicate modes and predicate registers. Explicitly restrict
4953 GPRs to modes of 16 bytes or smaller. Only allow FP registers
4954 to store a vector mode if it is recognized by
4955 aarch64_classify_vector_mode.
4956 (aarch64_regmode_natural_size): New function.
4957 (aarch64_hard_regno_caller_save_mode): Return the original mode
4959 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
4960 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
4961 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
4962 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
4964 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
4965 does not overlap dest if the function is frame-related. Handle
4967 (aarch64_split_add_offset): New function.
4968 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
4969 them aarch64_add_offset.
4970 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
4971 and update call to aarch64_sub_sp.
4972 (aarch64_add_cfa_expression): New function.
4973 (aarch64_expand_prologue): Pass extra temporary registers to the
4974 functions above. Handle the case in which we need to emit new
4975 DW_CFA_expressions for registers that were originally saved
4976 relative to the stack pointer, but now have to be expressed
4977 relative to the frame pointer.
4978 (aarch64_output_mi_thunk): Pass extra temporary registers to the
4980 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
4981 IP0 and IP1 values for SVE frames.
4982 (aarch64_expand_vec_series): New function.
4983 (aarch64_expand_sve_widened_duplicate): Likewise.
4984 (aarch64_expand_sve_const_vector): Likewise.
4985 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
4986 Handle SVE constants. Use emit_move_insn to move a force_const_mem
4987 into the register, rather than emitting a SET directly.
4988 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
4989 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
4990 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
4991 (offset_9bit_signed_scaled_p): New functions.
4992 (aarch64_replicate_bitmask_imm): New function.
4993 (aarch64_bitmask_imm): Use it.
4994 (aarch64_cannot_force_const_mem): Reject expressions involving
4995 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
4996 (aarch64_classify_index): Handle SVE indices, by requiring
4997 a plain register index with a scale that matches the element size.
4998 (aarch64_classify_address): Handle SVE addresses. Assert that
4999 the mode of the address is VOIDmode or an integer mode.
5000 Update call to aarch64_classify_symbol.
5001 (aarch64_classify_symbolic_expression): Update call to
5002 aarch64_classify_symbol.
5003 (aarch64_const_vec_all_in_range_p): New function.
5004 (aarch64_print_vector_float_operand): Likewise.
5005 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
5006 "vN" for FP registers with SVE modes. Handle (const ...) vectors
5007 and the FP immediates 1.0 and 0.5.
5008 (aarch64_print_address_internal): Handle SVE addresses.
5009 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
5010 (aarch64_regno_regclass): Handle predicate registers.
5011 (aarch64_secondary_reload): Handle big-endian reloads of SVE
5013 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
5014 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
5015 (aarch64_convert_sve_vector_bits): New function.
5016 (aarch64_override_options): Use it to handle -msve-vector-bits=.
5017 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
5019 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
5020 Handle SVE vector and predicate modes. Accept VL-based constants
5021 that need only one temporary register, and VL offsets that require
5022 no temporary registers.
5023 (aarch64_conditional_register_usage): Mark the predicate registers
5024 as fixed if SVE isn't available.
5025 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
5026 Return true for SVE vector and predicate modes.
5027 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
5028 rather than an unsigned int. Handle SVE modes.
5029 (aarch64_preferred_simd_mode): Update call accordingly. Handle
5031 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
5033 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
5034 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
5035 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
5036 (aarch64_sve_float_mul_immediate_p): New functions.
5037 (aarch64_sve_valid_immediate): New function.
5038 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
5039 Explicitly reject structure modes. Check for INDEX constants.
5040 Handle PTRUE and PFALSE constants.
5041 (aarch64_check_zero_based_sve_index_immediate): New function.
5042 (aarch64_simd_imm_zero_p): Delete.
5043 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
5044 vector modes. Accept constants in the range of CNT[BHWD].
5045 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
5046 ask for an Advanced SIMD mode.
5047 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
5048 (aarch64_simd_vector_alignment): Handle SVE predicates.
5049 (aarch64_vectorize_preferred_vector_alignment): New function.
5050 (aarch64_simd_vector_alignment_reachable): Use it instead of
5052 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
5053 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
5055 (MAX_VECT_LEN): Delete.
5056 (expand_vec_perm_d): Add a vec_flags field.
5057 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
5058 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
5059 (aarch64_evpc_ext): Don't apply a big-endian lane correction
5061 (aarch64_evpc_rev): Rename to...
5062 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
5063 (aarch64_evpc_rev_global): New function.
5064 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
5065 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
5067 (aarch64_evpc_sve_tbl): New function.
5068 (aarch64_expand_vec_perm_const_1): Update after rename of
5069 aarch64_evpc_rev. Handle SVE permutes too, trying
5070 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
5071 than aarch64_evpc_tbl.
5072 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
5073 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
5074 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
5075 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
5076 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
5077 (aarch64_expand_sve_vcond): New functions.
5078 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
5079 of aarch64_vector_mode_p.
5080 (aarch64_dwarf_poly_indeterminate_value): New function.
5081 (aarch64_compute_pressure_classes): Likewise.
5082 (aarch64_can_change_mode_class): Likewise.
5083 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
5084 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
5085 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
5086 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
5087 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
5088 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
5089 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
5090 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
5092 (Dn, Dl, Dr): Accept const as well as const_vector.
5093 (Dz): Likewise. Compare against CONST0_RTX.
5094 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
5095 of "vector" where appropriate.
5096 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
5097 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
5098 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
5099 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
5100 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
5101 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
5102 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
5103 (v_int_equiv): Extend to SVE modes.
5104 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
5106 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
5107 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
5108 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
5109 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
5110 (SVE_COND_FP_CMP): New int iterators.
5111 (perm_hilo): Handle the new unpack unspecs.
5112 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
5114 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
5115 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
5116 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
5117 (aarch64_equality_operator, aarch64_constant_vector_operand)
5118 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
5119 (aarch64_sve_nonimmediate_operand): Likewise.
5120 (aarch64_sve_general_operand): Likewise.
5121 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
5122 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
5123 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
5124 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
5125 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
5126 (aarch64_sve_float_arith_immediate): Likewise.
5127 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
5128 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
5129 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
5130 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
5131 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
5132 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
5133 (aarch64_sve_float_arith_operand): Likewise.
5134 (aarch64_sve_float_arith_with_sub_operand): Likewise.
5135 (aarch64_sve_float_mul_operand): Likewise.
5136 (aarch64_sve_vec_perm_operand): Likewise.
5137 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
5138 (aarch64_mov_operand): Accept const_poly_int and const_vector.
5139 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
5140 as well as const_vector.
5141 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
5142 in file. Use CONST0_RTX and CONSTM1_RTX.
5143 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
5144 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
5145 Use aarch64_simd_imm_zero.
5146 * config/aarch64/aarch64-sve.md: New file.
5147 * config/aarch64/aarch64.md: Include it.
5148 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
5149 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
5150 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
5151 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
5152 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
5153 (sve): New attribute.
5154 (enabled): Disable instructions with the sve attribute unless
5156 (movqi, movhi): Pass CONST_POLY_INT operaneds through
5157 aarch64_expand_mov_immediate.
5158 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
5159 CNT[BHSD] immediates.
5160 (movti): Split CONST_POLY_INT moves into two halves.
5161 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
5162 Split additions that need a temporary here if the destination
5163 is the stack pointer.
5164 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
5165 (*add<mode>3_poly_1): New instruction.
5166 (set_clobber_cc): New expander.
5168 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5170 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
5171 parameter and use it instead of GET_MODE_SIZE (innermode). Use
5172 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
5173 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
5174 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
5175 Change innermode from fixed_mode_size to machine_mode.
5176 (simplify_subreg): Update call accordingly. Handle a constant-sized
5177 subreg of a variable-length CONST_VECTOR.
5179 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5180 Alan Hayward <alan.hayward@arm.com>
5181 David Sherwood <david.sherwood@arm.com>
5183 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
5184 (add_offset_to_base): New function, split out from...
5185 (create_mem_ref): ...here. When handling a scale other than 1,
5186 check first whether the address is valid without the offset.
5187 Add it into the base if so, leaving the index and scale as-is.
5189 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5192 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
5193 fold_for_warn before checking if arg2 is INTEGER_CST.
5195 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
5197 * config/rs6000/predicates.md (load_multiple_operation): Delete.
5198 (store_multiple_operation): Delete.
5199 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
5200 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
5201 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
5202 guarded by TARGET_STRING.
5203 (rs6000_output_load_multiple): Delete.
5204 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
5205 OPTION_MASK_STRING / TARGET_STRING handling.
5206 (print_operand) <'N', 'O'>: Add comment that these are unused now.
5207 (const rs6000_opt_masks) <"string">: Change mask to 0.
5208 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
5209 (MASK_STRING): Delete.
5210 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
5212 (load_multiple): Delete.
5219 (store_multiple): Delete.
5226 (movmemsi_8reg): Delete.
5227 (corresponding unnamed define_insn): Delete.
5228 (movmemsi_6reg): Delete.
5229 (corresponding unnamed define_insn): Delete.
5230 (movmemsi_4reg): Delete.
5231 (corresponding unnamed define_insn): Delete.
5232 (movmemsi_2reg): Delete.
5233 (corresponding unnamed define_insn): Delete.
5234 (movmemsi_1reg): Delete.
5235 (corresponding unnamed define_insn): Delete.
5236 * config/rs6000/rs6000.opt (mno-string): New.
5237 (mstring): Replace by deprecation warning stub.
5238 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
5240 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5242 * regrename.c (regrename_do_replace): If replacing the same
5243 reg multiple times, try to reuse last created gen_raw_REG.
5246 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
5247 main to workaround a bug in GDB.
5249 2018-01-12 Tom de Vries <tom@codesourcery.com>
5252 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
5254 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
5256 PR rtl-optimization/80481
5257 * ira-color.c (get_cap_member): New function.
5258 (allocnos_conflict_by_live_ranges_p): Use it.
5259 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
5260 (setup_slot_coalesced_allocno_live_ranges): Ditto.
5262 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
5265 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
5266 (*saddl_se_1): Ditto.
5268 (*ssubl_se_1): Ditto.
5270 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5272 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
5273 rather than wi::to_widest for DR_INITs.
5274 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
5275 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
5276 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
5278 (vect_analyze_group_access_1): Note that here.
5280 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5282 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
5283 polynomial type sizes.
5285 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5287 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
5288 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
5289 (gimple_add_tmp_var): Likewise.
5291 2018-01-12 Martin Liska <mliska@suse.cz>
5293 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
5294 (gimple_alloc_sizes): Likewise.
5295 (dump_gimple_statistics): Use PRIu64 in printf format.
5296 * gimple.h: Change uint64_t to int.
5298 2018-01-12 Martin Liska <mliska@suse.cz>
5300 * tree-core.h: Use uint64_t instead of int.
5301 * tree.c (tree_node_counts): Likewise.
5302 (tree_node_sizes): Likewise.
5303 (dump_tree_statistics): Use PRIu64 in printf format.
5305 2018-01-12 Martin Liska <mliska@suse.cz>
5307 * Makefile.in: As qsort_chk is implemented in vec.c, add
5308 vec.o to linkage of gencfn-macros.
5309 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
5310 passing the info to record_node_allocation_statistics.
5311 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
5313 * ggc-common.c (struct ggc_usage): Add operator== and use
5314 it in operator< and compare function.
5315 * mem-stats.h (struct mem_usage): Likewise.
5316 * vec.c (struct vec_usage): Remove operator< and compare
5317 function. Can be simply inherited.
5319 2018-01-12 Martin Jambor <mjambor@suse.cz>
5322 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
5323 * tree-ssa-math-opts.c: Include domwalk.h.
5324 (convert_mult_to_fma_1): New function.
5325 (fma_transformation_info): New type.
5326 (fma_deferring_state): Likewise.
5327 (cancel_fma_deferring): New function.
5328 (result_of_phi): Likewise.
5329 (last_fma_candidate_feeds_initial_phi): Likewise.
5330 (convert_mult_to_fma): Added deferring logic, split actual
5331 transformation to convert_mult_to_fma_1.
5332 (math_opts_dom_walker): New type.
5333 (math_opts_dom_walker::after_dom_children): New method, body moved
5334 here from pass_optimize_widening_mul::execute, added deferring logic
5336 (pass_optimize_widening_mul::execute): Moved most of code to
5337 math_opts_dom_walker::after_dom_children.
5338 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
5339 * config/i386/i386.c (ix86_option_override_internal): Added
5340 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
5342 2018-01-12 Richard Biener <rguenther@suse.de>
5345 * dwarf2out.c (gen_variable_die): Do not reset old_die for
5346 inline instance vars.
5348 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
5351 * config/rx/rx.c (rx_is_restricted_memory_address):
5354 2018-01-12 Richard Biener <rguenther@suse.de>
5356 PR tree-optimization/80846
5357 * target.def (split_reduction): New target hook.
5358 * targhooks.c (default_split_reduction): New function.
5359 * targhooks.h (default_split_reduction): Declare.
5360 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
5361 target requests first reduce vectors by combining low and high
5363 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
5364 (get_vectype_for_scalar_type_and_size): Export.
5365 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
5366 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
5367 * doc/tm.texi: Regenerate.
5368 * config/i386/i386.c (ix86_split_reduction): Implement
5369 TARGET_VECTORIZE_SPLIT_REDUCTION.
5371 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5374 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
5375 in PIC mode except for TARGET_VXWORKS_RTP.
5376 * config/sparc/sparc.c: Include cfgrtl.h.
5377 (TARGET_INIT_PIC_REG): Define.
5378 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
5379 (sparc_pic_register_p): New predicate.
5380 (sparc_legitimate_address_p): Use it.
5381 (sparc_legitimize_pic_address): Likewise.
5382 (sparc_delegitimize_address): Likewise.
5383 (sparc_mode_dependent_address_p): Likewise.
5384 (gen_load_pcrel_sym): Remove 4th parameter.
5385 (load_got_register): Adjust call to above. Remove obsolete stuff.
5386 (sparc_expand_prologue): Do not call load_got_register here.
5387 (sparc_flat_expand_prologue): Likewise.
5388 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
5389 (sparc_use_pseudo_pic_reg): New function.
5390 (sparc_init_pic_reg): Likewise.
5391 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5392 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
5394 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
5396 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
5397 Add item for branch_cost.
5399 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5401 PR rtl-optimization/83565
5402 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
5403 not extend the result to a larger mode for rotate operations.
5404 (num_sign_bit_copies1): Likewise.
5406 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5409 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
5411 Use values-Xc.o for -pedantic.
5412 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
5414 2018-01-12 Martin Liska <mliska@suse.cz>
5417 * ipa-devirt.c (final_warning_record::grow_type_warnings):
5419 (possible_polymorphic_call_targets): Use it.
5420 (ipa_devirt): Likewise.
5422 2018-01-12 Martin Liska <mliska@suse.cz>
5424 * profile-count.h (enum profile_quality): Use 0 as invalid
5425 enum value of profile_quality.
5427 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
5429 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
5430 -mext-string options.
5432 2018-01-12 Richard Biener <rguenther@suse.de>
5434 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
5435 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
5436 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
5438 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
5440 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5442 * configure.ac (--with-long-double-format): Add support for the
5443 configuration option to change the default long double format on
5445 * config.gcc (powerpc*-linux*-*): Likewise.
5446 * configure: Regenerate.
5447 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
5448 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
5449 used without modification.
5451 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5453 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
5454 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
5455 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
5456 MISC_BUILTIN_SPEC_BARRIER.
5457 (rs6000_init_builtins): Likewise.
5458 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
5460 (speculation_barrier): New define_insn.
5461 * doc/extend.texi: Document __builtin_speculation_barrier.
5463 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5466 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
5467 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
5468 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
5470 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
5471 integral modes instead of "ss" and "sd".
5472 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
5473 vectors with 32-bit and 64-bit elements.
5474 (vecdupssescalarmodesuffix): New mode attribute.
5475 (vec_dup<mode>): Use it.
5477 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
5480 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
5481 frame if argument is passed on stack.
5483 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5486 * ree.c (combine_reaching_defs): Optimize also
5487 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
5488 reg2=any_extend(exp); reg1=reg2;, formatting fix.
5490 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5493 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
5495 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5498 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
5499 after they are computed.
5501 2018-01-11 Bin Cheng <bin.cheng@arm.com>
5503 PR tree-optimization/83695
5504 * gimple-loop-linterchange.cc
5505 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
5506 reset cached scev information after interchange.
5507 (pass_linterchange::execute): Remove call to scev_reset_htab.
5509 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5511 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
5512 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
5513 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
5514 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
5515 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
5516 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
5517 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
5518 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
5519 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
5520 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
5521 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
5522 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
5523 (V_lane_reg): Likewise.
5524 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
5526 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
5527 (vfmal_lane_low<mode>_intrinsic,
5528 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
5529 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
5530 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
5531 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
5532 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
5533 vfmsl_lane_high<mode>_intrinsic): New define_insns.
5535 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5537 * config/arm/arm-cpus.in (fp16fml): New feature.
5538 (ALL_SIMD): Add fp16fml.
5539 (armv8.2-a): Add fp16fml as an option.
5540 (armv8.3-a): Likewise.
5541 (armv8.4-a): Add fp16fml as part of fp16.
5542 * config/arm/arm.h (TARGET_FP16FML): Define.
5543 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
5545 * config/arm/arm-modes.def (V2HF): Define.
5546 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
5547 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
5548 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
5549 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
5550 vfmsl_low, vfmsl_high): New set of builtins.
5551 * config/arm/iterators.md (PLUSMINUS): New code iterator.
5552 (vfml_op): New code attribute.
5553 (VFMLHALVES): New int iterator.
5554 (VFML, VFMLSEL): New mode attributes.
5555 (V_reg): Define mapping for V2HF.
5556 (V_hi, V_lo): New mode attributes.
5557 (VF_constraint): Likewise.
5558 (vfml_half, vfml_half_selector): New int attributes.
5559 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
5561 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
5562 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
5564 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
5565 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
5566 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
5567 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
5569 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
5570 Document new effective target and option set.
5572 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5574 * config/arm/arm-cpus.in (armv8_4): New feature.
5575 (ARMv8_4a): New fgroup.
5576 (armv8.4-a): New arch.
5577 * config/arm/arm-tables.opt: Regenerate.
5578 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
5579 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
5580 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
5581 Add matching rules for -march=armv8.4-a and extensions.
5582 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
5584 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
5587 * config/rx/rx.md (BW): New mode attribute.
5588 (sync_lock_test_and_setsi): Add mode suffix to insn output.
5590 2018-01-11 Richard Biener <rguenther@suse.de>
5592 PR tree-optimization/83435
5593 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
5594 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
5595 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
5597 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5598 Alan Hayward <alan.hayward@arm.com>
5599 David Sherwood <david.sherwood@arm.com>
5601 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
5603 (aarch64_classify_address): Initialize it. Track polynomial offsets.
5604 (aarch64_print_address_internal): Use it to check for a zero offset.
5606 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5607 Alan Hayward <alan.hayward@arm.com>
5608 David Sherwood <david.sherwood@arm.com>
5610 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
5611 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
5612 Return a poly_int64 rather than a HOST_WIDE_INT.
5613 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
5614 rather than a HOST_WIDE_INT.
5615 * config/aarch64/aarch64.h (aarch64_frame): Protect with
5616 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
5617 hard_fp_offset, frame_size, initial_adjust, callee_offset and
5618 final_offset from HOST_WIDE_INT to poly_int64.
5619 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
5620 to_constant when getting the number of units in an Advanced SIMD
5622 (aarch64_builtin_vectorized_function): Check for a constant number
5624 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
5626 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
5627 attribute instead of GET_MODE_NUNITS.
5628 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
5629 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
5630 GET_MODE_SIZE for fixed-size registers.
5631 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
5632 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
5633 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
5634 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
5635 (aarch64_print_operand, aarch64_print_address_internal)
5636 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
5637 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
5638 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
5639 Handle polynomial GET_MODE_SIZE.
5640 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
5641 wider than SImode without modification.
5642 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
5643 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
5644 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
5645 passing and returning SVE modes.
5646 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
5647 rather than GEN_INT.
5648 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
5649 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
5650 (aarch64_allocate_and_probe_stack_space): Likewise.
5651 (aarch64_layout_frame): Cope with polynomial offsets.
5652 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
5653 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
5655 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
5656 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
5657 poly_int64 rather than a HOST_WIDE_INT.
5658 (aarch64_get_separate_components, aarch64_process_components)
5659 (aarch64_expand_prologue, aarch64_expand_epilogue)
5660 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
5661 (aarch64_anchor_offset): New function, split out from...
5662 (aarch64_legitimize_address): ...here.
5663 (aarch64_builtin_vectorization_cost): Handle polynomial
5664 TYPE_VECTOR_SUBPARTS.
5665 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
5667 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
5668 number of elements from the PARALLEL rather than the mode.
5669 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
5670 rather than GET_MODE_BITSIZE.
5671 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
5672 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
5673 (aarch64_expand_vec_perm_const_1): Handle polynomial
5674 d->perm.length () and d->perm elements.
5675 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
5676 Apply to_constant to d->perm elements.
5677 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
5678 polynomial CONST_VECTOR_NUNITS.
5679 (aarch64_move_pointer): Take amount as a poly_int64 rather
5681 (aarch64_progress_pointer): Avoid temporary variable.
5682 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
5683 the mode attribute instead of GET_MODE.
5685 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5686 Alan Hayward <alan.hayward@arm.com>
5687 David Sherwood <david.sherwood@arm.com>
5689 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
5690 x exists before using it.
5691 (aarch64_add_constant_internal): Rename to...
5692 (aarch64_add_offset_1): ...this. Replace regnum with separate
5693 src and dest rtxes. Handle the case in which they're different,
5694 including when the offset is zero. Replace scratchreg with an rtx.
5695 Use 2 additions if there is no spare register into which we can
5696 move a 16-bit constant.
5697 (aarch64_add_constant): Delete.
5698 (aarch64_add_offset): Replace reg with separate src and dest
5699 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
5700 Use aarch64_add_offset_1.
5701 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
5702 an rtx rather than an int. Take the delta as a poly_int64
5703 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
5704 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
5705 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
5706 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
5707 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
5709 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
5710 aarch64_add_constant.
5712 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5714 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
5715 Use scalar_float_mode.
5717 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5719 * config/aarch64/aarch64-simd.md
5720 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
5721 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
5722 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
5723 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
5724 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
5725 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
5726 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
5727 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
5728 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
5729 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
5731 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5734 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
5735 targ_options->x_arm_arch_string is non NULL.
5737 2018-01-11 Tamar Christina <tamar.christina@arm.com>
5739 * config/aarch64/aarch64.h
5740 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
5742 2018-01-11 Sudakshina Das <sudi.das@arm.com>
5745 * expmed.c (emit_store_flag_force): Swap if const op0
5746 and change VOIDmode to mode of op0.
5748 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5750 PR rtl-optimization/83761
5751 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
5752 than bytes to mode_for_size.
5754 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5757 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
5758 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
5761 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5764 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
5765 when in layout mode.
5766 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
5767 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
5770 2018-01-10 Michael Collison <michael.collison@arm.com>
5772 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
5773 * config/aarch64/aarch64-option-extension.def: Add
5774 AARCH64_OPT_EXTENSION of 'fp16fml'.
5775 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5776 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
5777 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
5778 * config/aarch64/constraints.md (Ui7): New constraint.
5779 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
5780 (VFMLA_SEL_W): Ditto.
5783 (VFMLA16_LOW): New int iterator.
5784 (VFMLA16_HIGH): Ditto.
5785 (UNSPEC_FMLAL): New unspec.
5786 (UNSPEC_FMLSL): Ditto.
5787 (UNSPEC_FMLAL2): Ditto.
5788 (UNSPEC_FMLSL2): Ditto.
5789 (f16mac): New code attribute.
5790 * config/aarch64/aarch64-simd-builtins.def
5791 (aarch64_fmlal_lowv2sf): Ditto.
5792 (aarch64_fmlsl_lowv2sf): Ditto.
5793 (aarch64_fmlalq_lowv4sf): Ditto.
5794 (aarch64_fmlslq_lowv4sf): Ditto.
5795 (aarch64_fmlal_highv2sf): Ditto.
5796 (aarch64_fmlsl_highv2sf): Ditto.
5797 (aarch64_fmlalq_highv4sf): Ditto.
5798 (aarch64_fmlslq_highv4sf): Ditto.
5799 (aarch64_fmlal_lane_lowv2sf): Ditto.
5800 (aarch64_fmlsl_lane_lowv2sf): Ditto.
5801 (aarch64_fmlal_laneq_lowv2sf): Ditto.
5802 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
5803 (aarch64_fmlalq_lane_lowv4sf): Ditto.
5804 (aarch64_fmlsl_lane_lowv4sf): Ditto.
5805 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
5806 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
5807 (aarch64_fmlal_lane_highv2sf): Ditto.
5808 (aarch64_fmlsl_lane_highv2sf): Ditto.
5809 (aarch64_fmlal_laneq_highv2sf): Ditto.
5810 (aarch64_fmlsl_laneq_highv2sf): Ditto.
5811 (aarch64_fmlalq_lane_highv4sf): Ditto.
5812 (aarch64_fmlsl_lane_highv4sf): Ditto.
5813 (aarch64_fmlalq_laneq_highv4sf): Ditto.
5814 (aarch64_fmlsl_laneq_highv4sf): Ditto.
5815 * config/aarch64/aarch64-simd.md:
5816 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
5817 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5818 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
5819 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5820 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
5821 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
5822 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
5823 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
5824 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
5825 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
5826 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
5827 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
5828 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
5829 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
5830 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
5831 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
5832 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
5833 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
5834 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
5835 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
5836 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
5837 (vfmlsl_low_u32): Ditto.
5838 (vfmlalq_low_u32): Ditto.
5839 (vfmlslq_low_u32): Ditto.
5840 (vfmlal_high_u32): Ditto.
5841 (vfmlsl_high_u32): Ditto.
5842 (vfmlalq_high_u32): Ditto.
5843 (vfmlslq_high_u32): Ditto.
5844 (vfmlal_lane_low_u32): Ditto.
5845 (vfmlsl_lane_low_u32): Ditto.
5846 (vfmlal_laneq_low_u32): Ditto.
5847 (vfmlsl_laneq_low_u32): Ditto.
5848 (vfmlalq_lane_low_u32): Ditto.
5849 (vfmlslq_lane_low_u32): Ditto.
5850 (vfmlalq_laneq_low_u32): Ditto.
5851 (vfmlslq_laneq_low_u32): Ditto.
5852 (vfmlal_lane_high_u32): Ditto.
5853 (vfmlsl_lane_high_u32): Ditto.
5854 (vfmlal_laneq_high_u32): Ditto.
5855 (vfmlsl_laneq_high_u32): Ditto.
5856 (vfmlalq_lane_high_u32): Ditto.
5857 (vfmlslq_lane_high_u32): Ditto.
5858 (vfmlalq_laneq_high_u32): Ditto.
5859 (vfmlslq_laneq_high_u32): Ditto.
5860 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
5861 (AARCH64_FL_FOR_ARCH8_4): New.
5862 (AARCH64_ISA_F16FML): New ISA flag.
5863 (TARGET_F16FML): New feature flag for fp16fml.
5864 (doc/invoke.texi): Document new fp16fml option.
5866 2018-01-10 Michael Collison <michael.collison@arm.com>
5868 * config/aarch64/aarch64-builtins.c:
5869 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
5870 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5871 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
5872 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
5873 (AARCH64_ISA_SHA3): New ISA flag.
5874 (TARGET_SHA3): New feature flag for sha3.
5875 * config/aarch64/iterators.md (sha512_op): New int attribute.
5876 (CRYPTO_SHA512): New int iterator.
5877 (UNSPEC_SHA512H): New unspec.
5878 (UNSPEC_SHA512H2): Ditto.
5879 (UNSPEC_SHA512SU0): Ditto.
5880 (UNSPEC_SHA512SU1): Ditto.
5881 * config/aarch64/aarch64-simd-builtins.def
5882 (aarch64_crypto_sha512hqv2di): New builtin.
5883 (aarch64_crypto_sha512h2qv2di): Ditto.
5884 (aarch64_crypto_sha512su0qv2di): Ditto.
5885 (aarch64_crypto_sha512su1qv2di): Ditto.
5886 (aarch64_eor3qv8hi): Ditto.
5887 (aarch64_rax1qv2di): Ditto.
5888 (aarch64_xarqv2di): Ditto.
5889 (aarch64_bcaxqv8hi): Ditto.
5890 * config/aarch64/aarch64-simd.md:
5891 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
5892 (aarch64_crypto_sha512su0qv2di): Ditto.
5893 (aarch64_crypto_sha512su1qv2di): Ditto.
5894 (aarch64_eor3qv8hi): Ditto.
5895 (aarch64_rax1qv2di): Ditto.
5896 (aarch64_xarqv2di): Ditto.
5897 (aarch64_bcaxqv8hi): Ditto.
5898 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
5899 (vsha512h2q_u64): Ditto.
5900 (vsha512su0q_u64): Ditto.
5901 (vsha512su1q_u64): Ditto.
5902 (veor3q_u16): Ditto.
5903 (vrax1q_u64): Ditto.
5905 (vbcaxq_u16): Ditto.
5906 * config/arm/types.md (crypto_sha512): New type attribute.
5907 (crypto_sha3): Ditto.
5908 (doc/invoke.texi): Document new sha3 option.
5910 2018-01-10 Michael Collison <michael.collison@arm.com>
5912 * config/aarch64/aarch64-builtins.c:
5913 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
5914 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5915 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
5916 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
5917 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
5918 (AARCH64_ISA_SM4): New ISA flag.
5919 (TARGET_SM4): New feature flag for sm4.
5920 * config/aarch64/aarch64-simd-builtins.def
5921 (aarch64_sm3ss1qv4si): Ditto.
5922 (aarch64_sm3tt1aq4si): Ditto.
5923 (aarch64_sm3tt1bq4si): Ditto.
5924 (aarch64_sm3tt2aq4si): Ditto.
5925 (aarch64_sm3tt2bq4si): Ditto.
5926 (aarch64_sm3partw1qv4si): Ditto.
5927 (aarch64_sm3partw2qv4si): Ditto.
5928 (aarch64_sm4eqv4si): Ditto.
5929 (aarch64_sm4ekeyqv4si): Ditto.
5930 * config/aarch64/aarch64-simd.md:
5931 (aarch64_sm3ss1qv4si): Ditto.
5932 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
5933 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
5934 (aarch64_sm4eqv4si): Ditto.
5935 (aarch64_sm4ekeyqv4si): Ditto.
5936 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
5937 (sm3part_op): Ditto.
5938 (CRYPTO_SM3TT): Ditto.
5939 (CRYPTO_SM3PART): Ditto.
5940 (UNSPEC_SM3SS1): New unspec.
5941 (UNSPEC_SM3TT1A): Ditto.
5942 (UNSPEC_SM3TT1B): Ditto.
5943 (UNSPEC_SM3TT2A): Ditto.
5944 (UNSPEC_SM3TT2B): Ditto.
5945 (UNSPEC_SM3PARTW1): Ditto.
5946 (UNSPEC_SM3PARTW2): Ditto.
5947 (UNSPEC_SM4E): Ditto.
5948 (UNSPEC_SM4EKEY): Ditto.
5949 * config/aarch64/constraints.md (Ui2): New constraint.
5950 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
5951 * config/arm/types.md (crypto_sm3): New type attribute.
5952 (crypto_sm4): Ditto.
5953 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
5954 (vsm3tt1aq_u32): Ditto.
5955 (vsm3tt1bq_u32): Ditto.
5956 (vsm3tt2aq_u32): Ditto.
5957 (vsm3tt2bq_u32): Ditto.
5958 (vsm3partw1q_u32): Ditto.
5959 (vsm3partw2q_u32): Ditto.
5960 (vsm4eq_u32): Ditto.
5961 (vsm4ekeyq_u32): Ditto.
5962 (doc/invoke.texi): Document new sm4 option.
5964 2018-01-10 Michael Collison <michael.collison@arm.com>
5966 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
5967 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
5968 (AARCH64_FL_FOR_ARCH8_4): New.
5969 (AARCH64_FL_V8_4): New flag.
5970 (doc/invoke.texi): Document new armv8.4-a option.
5972 2018-01-10 Michael Collison <michael.collison@arm.com>
5974 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5975 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
5976 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
5977 * config/aarch64/aarch64-option-extension.def: Add
5978 AARCH64_OPT_EXTENSION of 'sha2'.
5979 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
5980 (crypto): Disable sha2 and aes if crypto disabled.
5981 (crypto): Enable aes and sha2 if enabled.
5982 (simd): Disable sha2 and aes if simd disabled.
5983 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
5985 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
5986 (TARGET_SHA2): New feature flag for sha2.
5987 (TARGET_AES): New feature flag for aes.
5988 * config/aarch64/aarch64-simd.md:
5989 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
5990 conditional on TARGET_AES.
5991 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
5992 (aarch64_crypto_sha1hsi): Make pattern conditional
5994 (aarch64_crypto_sha1hv4si): Ditto.
5995 (aarch64_be_crypto_sha1hv4si): Ditto.
5996 (aarch64_crypto_sha1su1v4si): Ditto.
5997 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
5998 (aarch64_crypto_sha1su0v4si): Ditto.
5999 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
6000 (aarch64_crypto_sha256su0v4si): Ditto.
6001 (aarch64_crypto_sha256su1v4si): Ditto.
6002 (doc/invoke.texi): Document new aes and sha2 options.
6004 2018-01-10 Martin Sebor <msebor@redhat.com>
6006 PR tree-optimization/83781
6007 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
6010 2018-01-11 Martin Sebor <msebor@gmail.com>
6011 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
6013 PR tree-optimization/83501
6014 PR tree-optimization/81703
6016 * tree-ssa-strlen.c (get_string_cst): Rename...
6017 (get_string_len): ...to this. Handle global constants.
6018 (handle_char_store): Adjust.
6020 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
6021 Jim Wilson <jimw@sifive.com>
6023 * config/riscv/riscv-protos.h (riscv_output_return): New.
6024 * config/riscv/riscv.c (struct machine_function): New naked_p field.
6025 (riscv_attribute_table, riscv_output_return),
6026 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
6027 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
6028 (riscv_compute_frame_info): Only compute frame->mask if not a naked
6030 (riscv_expand_prologue): Add early return for naked function.
6031 (riscv_expand_epilogue): Likewise.
6032 (riscv_function_ok_for_sibcall): Return false for naked function.
6033 (riscv_set_current_function): New.
6034 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
6035 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
6036 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
6037 * doc/extend.texi (RISC-V Function Attributes): New.
6039 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
6041 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
6042 check for 128-bit long double before checking TCmode.
6043 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
6044 128-bit long doubles before checking TFmode or TCmode.
6045 (FLOAT128_IBM_P): Likewise.
6047 2018-01-10 Martin Sebor <msebor@redhat.com>
6049 PR tree-optimization/83671
6050 * builtins.c (c_strlen): Unconditionally return zero for the empty
6052 Use -Warray-bounds for warnings.
6053 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
6054 for non-constant array indices with COMPONENT_REF, arrays of
6055 arrays, and pointers to arrays.
6056 (gimple_fold_builtin_strlen): Determine and set length range for
6057 non-constant character arrays.
6059 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
6062 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
6065 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
6067 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
6069 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6072 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
6073 VECTOR_MEM_ALTIVEC_OR_VSX_P.
6074 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
6075 indexed_or_indirect_operand predicate.
6076 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
6077 (*vsx_le_perm_load_v8hi): Likewise.
6078 (*vsx_le_perm_load_v16qi): Likewise.
6079 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
6080 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
6081 (*vsx_le_perm_store_v8hi): Likewise.
6082 (*vsx_le_perm_store_v16qi): Likewise.
6083 (eight unnamed splitters): Likewise.
6085 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
6087 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
6088 * config/rs6000/emmintrin.h: Likewise.
6089 * config/rs6000/mmintrin.h: Likewise.
6090 * config/rs6000/xmmintrin.h: Likewise.
6092 2018-01-10 David Malcolm <dmalcolm@redhat.com>
6095 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
6097 * tree.c (tree_nop_conversion): Return true for location wrapper
6099 (maybe_wrap_with_location): New function.
6100 (selftest::check_strip_nops): New function.
6101 (selftest::test_location_wrappers): New function.
6102 (selftest::tree_c_tests): Call it.
6103 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
6104 (maybe_wrap_with_location): New decl.
6105 (EXPR_LOCATION_WRAPPER_P): New macro.
6106 (location_wrapper_p): New inline function.
6107 (tree_strip_any_location_wrapper): New inline function.
6109 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
6112 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
6113 stack_realign_offset for the largest alignment of stack slot
6115 (ix86_find_max_used_stack_alignment): New function.
6116 (ix86_finalize_stack_frame_flags): Use it. Set
6117 max_used_stack_alignment if we don't realign stack.
6118 * config/i386/i386.h (machine_function): Add
6119 max_used_stack_alignment.
6121 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
6123 * config/arm/arm.opt (-mbranch-cost): New option.
6124 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
6127 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
6130 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
6131 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
6133 2018-01-10 Richard Biener <rguenther@suse.de>
6136 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
6137 early out so it also covers the case where we have a non-NULL
6140 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6142 PR tree-optimization/83753
6143 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
6144 for non-strided grouped accesses if the number of elements is 1.
6146 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6149 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
6150 * i386.h (TARGET_USE_GATHER): Define.
6151 * x86-tune.def (X86_TUNE_USE_GATHER): New.
6153 2018-01-10 Martin Liska <mliska@suse.cz>
6156 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
6157 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
6159 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
6160 CLEANUP_NO_PARTITIONING is not set.
6162 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6164 * doc/rtl.texi: Remove documentation of (const ...) wrappers
6165 for vectors, as a partial revert of r254296.
6166 * rtl.h (const_vec_p): Delete.
6167 (const_vec_duplicate_p): Don't test for vector CONSTs.
6168 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
6169 * expmed.c (make_tree): Likewise.
6172 * common.md (E, F): Use CONSTANT_P instead of checking for
6174 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
6175 checking for CONST_VECTOR.
6177 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6180 * predict.c (force_edge_cold): Handle in more sane way edges
6183 2018-01-09 Carl Love <cel@us.ibm.com>
6185 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
6187 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
6188 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
6189 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
6190 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
6191 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
6192 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
6193 * config/rs6000/rs6000-protos.h: Add extern defition for
6194 rs6000_generate_float2_double_code.
6195 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
6197 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
6198 (float2_v2df): Add define_expand.
6200 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
6203 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
6204 op_mode in the force_to_mode call.
6206 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6208 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
6209 instead of checking each element individually.
6210 (aarch64_evpc_uzp): Likewise.
6211 (aarch64_evpc_zip): Likewise.
6212 (aarch64_evpc_ext): Likewise.
6213 (aarch64_evpc_rev): Likewise.
6214 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
6215 instead of checking each element individually. Return true without
6217 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
6218 whether all selected elements come from the same input, instead of
6219 checking each element individually. Remove calls to gen_rtx_REG,
6220 start_sequence and end_sequence and instead assert that no rtl is
6223 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6225 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
6226 order of HIGH and CONST checks.
6228 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6230 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
6231 if the destination isn't an SSA_NAME.
6233 2018-01-09 Richard Biener <rguenther@suse.de>
6235 PR tree-optimization/83668
6236 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
6238 (canonicalize_loop_form): ... here, renamed from ...
6239 (canonicalize_loop_closed_ssa_form): ... this and amended to
6240 swap successor edges for loop exit blocks to make us use
6241 the RPO order we need for initial schedule generation.
6243 2018-01-09 Joseph Myers <joseph@codesourcery.com>
6245 PR tree-optimization/64811
6246 * match.pd: When optimizing comparisons with Inf, avoid
6247 introducing or losing exceptions from comparisons with NaN.
6249 2018-01-09 Martin Liska <mliska@suse.cz>
6252 * asan.c (shadow_mem_size): Add gcc_assert.
6254 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
6256 Don't save registers in main().
6259 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
6260 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
6261 * config/avr/avr.c (avr_set_current_function): Don't error if
6262 naked, OS_task or OS_main are specified at the same time.
6263 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
6265 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
6267 * common/config/avr/avr-common.c (avr_option_optimization_table):
6268 Switch on -mmain-is-OS_task for optimizing compilations.
6270 2018-01-09 Richard Biener <rguenther@suse.de>
6272 PR tree-optimization/83572
6273 * graphite.c: Include cfganal.h.
6274 (graphite_transform_loops): Connect infinite loops to exit
6275 and remove fake edges at the end.
6277 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6279 * ipa-inline.c (edge_badness): Revert accidental checkin.
6281 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6284 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
6285 symbols; not inline clones.
6287 2018-01-09 Jakub Jelinek <jakub@redhat.com>
6290 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
6291 hard registers. Formatting fixes.
6293 PR preprocessor/83722
6294 * gcc.c (try_generate_repro): Pass
6295 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
6296 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
6299 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
6300 Kito Cheng <kito.cheng@gmail.com>
6302 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
6303 (riscv_leaf_function_p): Delete.
6304 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
6306 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6308 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
6310 (do_ifelse): New function.
6311 (do_isel): New function.
6312 (do_sub3): New function.
6313 (do_add3): New function.
6314 (do_load_mask_compare): New function.
6315 (do_overlap_load_compare): New function.
6316 (expand_compare_loop): New function.
6317 (expand_block_compare): Call expand_compare_loop() when appropriate.
6318 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
6320 (-mblock-compare-inline-loop-limit): New option.
6322 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6325 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
6326 Reverse order of second and third operands in first alternative.
6327 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
6328 of first and second elements in UNSPEC_VPERMR vector.
6329 (altivec_expand_vec_perm_le): Likewise.
6331 2018-01-08 Jeff Law <law@redhat.com>
6333 PR rtl-optimizatin/81308
6334 * tree-switch-conversion.c (cfg_altered): New file scoped static.
6335 (process_switch): If group_case_labels makes a change, then set
6337 (pass_convert_switch::execute): If a switch is converted, then
6338 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
6340 PR rtl-optimization/81308
6341 * recog.c (split_all_insns): Conditionally cleanup the CFG after
6344 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
6346 PR target/83663 - Revert r255946
6347 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
6348 generation for cases where splatting a value is not useful.
6349 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
6350 across a vec_duplicate and a paradoxical subreg forming a vector
6351 mode to a vec_concat.
6353 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6355 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
6356 -march=armv8.3-a variants.
6357 * config/arm/t-multilib: Likewise.
6358 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
6360 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6362 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
6364 (cceq_ior_compare_complement): Give it a name so I can use it, and
6365 change boolean_or_operator predicate to boolean_operator so it can
6366 be used to generate a crand.
6367 (eqne): New code iterator.
6368 (bd/bd_neg): New code_attrs.
6369 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
6370 a single define_insn.
6371 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
6372 decrement (bdnzt/bdnzf/bdzt/bdzf).
6373 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
6374 with the new names of the branch decrement patterns, and added the
6375 names of the branch decrement conditional patterns.
6377 2018-01-08 Richard Biener <rguenther@suse.de>
6379 PR tree-optimization/83563
6380 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
6383 2018-01-08 Richard Biener <rguenther@suse.de>
6386 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
6388 2018-01-08 Richard Biener <rguenther@suse.de>
6390 PR tree-optimization/83685
6391 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
6392 references to abnormals.
6394 2018-01-08 Richard Biener <rguenther@suse.de>
6397 * dwarf2out.c (output_indirect_strings): Handle empty
6398 skeleton_debug_str_hash.
6399 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
6401 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6403 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
6404 (emit_store_direct): Likewise.
6405 (arc_trampoline_adjust_address): Likewise.
6406 (arc_asm_trampoline_template): New function.
6407 (arc_initialize_trampoline): Use asm_trampoline_template.
6408 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
6409 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
6410 * config/arc/arc.md (flush_icache): Delete pattern.
6412 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6414 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
6415 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
6418 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6421 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
6422 by not USED_FOR_TARGET.
6423 (make_pass_resolve_sw_modes): Likewise.
6425 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6427 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
6430 2018-01-08 Richard Biener <rguenther@suse.de>
6433 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
6435 2018-01-08 Richard Biener <rguenther@suse.de>
6438 * match.pd ((t * 2) / 2) -> t): Add missing :c.
6440 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
6443 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
6444 basic blocks with a small number of successors.
6445 (convert_control_dep_chain_into_preds): Improve handling of
6447 (dump_predicates): Split apart into...
6448 (dump_pred_chain): ...here...
6449 (dump_pred_info): ...and here.
6450 (can_one_predicate_be_invalidated_p): Add debugging printfs.
6451 (can_chain_union_be_invalidated_p): Improve check for invalidation
6453 (uninit_uses_cannot_happen): Avoid unnecessary if
6454 convert_control_dep_chain_into_preds yielded nothing.
6456 2018-01-06 Martin Sebor <msebor@redhat.com>
6458 PR tree-optimization/83640
6459 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
6460 subtracting negative offset from size.
6461 (builtin_access::overlap): Adjust offset bounds of the access to fall
6462 within the size of the object if possible.
6464 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
6466 PR rtl-optimization/83699
6467 * expmed.c (extract_bit_field_1): Restrict the vector usage of
6468 extract_bit_field_as_subreg to cases in which the extracted
6469 value is also a vector.
6471 * lra-constraints.c (process_alt_operands): Test for the equivalence
6472 substitutions when detecting a possible reload cycle.
6474 2018-01-06 Jakub Jelinek <jakub@redhat.com>
6477 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
6478 by default if flag_selective_schedling{,2}. Formatting fixes.
6480 PR rtl-optimization/83682
6481 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
6482 if it has non-VECTOR_MODE element mode.
6483 (vec_duplicate_p): Likewise.
6486 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
6487 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
6489 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6492 * config/i386/i386-builtin.def
6493 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
6494 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
6495 Require also OPTION_MASK_ISA_AVX512F in addition to
6496 OPTION_MASK_ISA_GFNI.
6497 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
6498 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
6499 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
6500 to OPTION_MASK_ISA_GFNI.
6501 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
6502 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
6503 OPTION_MASK_ISA_AVX512BW.
6504 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
6505 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
6506 addition to OPTION_MASK_ISA_GFNI.
6507 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
6508 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
6509 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
6510 to OPTION_MASK_ISA_GFNI.
6511 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
6512 a requirement for all ISAs rather than any of them with a few
6514 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
6516 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
6517 bitmasks to be enabled with 3 exceptions, instead of requiring any
6518 enabled ISA with lots of exceptions.
6519 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
6520 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
6521 Change avx512bw in isa attribute to avx512f.
6522 * config/i386/sgxintrin.h: Add license boilerplate.
6523 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
6524 to __AVX512F__ and __AVX512VL to __AVX512VL__.
6525 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
6526 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
6528 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
6529 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
6530 temporarily sse2 rather than sse if not enabled already.
6533 * config/i386/sse.md (VI248_VLBW): Rename to ...
6534 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
6535 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
6536 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
6537 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
6538 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
6539 mode iterator instead of VI248_VLBW.
6541 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
6543 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
6544 (record_modified): Skip clobbers; add debug output.
6545 (param_change_prob): Use sreal frequencies.
6547 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6549 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
6550 punt for user-aligned variables.
6552 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6554 * tree-chrec.c (chrec_contains_symbols): Return true for
6557 2018-01-05 Sudakshina Das <sudi.das@arm.com>
6560 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
6561 of (x|y) == x for BICS pattern.
6563 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6565 PR tree-optimization/83605
6566 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
6567 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
6570 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
6572 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
6573 * config/epiphany/rtems.h: New file.
6575 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6576 Uros Bizjak <ubizjak@gmail.com>
6579 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
6580 QIreg_operand instead of register_operand predicate.
6581 * config/i386/i386.c (ix86_rop_should_change_byte_p,
6582 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
6583 comments instead of -fmitigate[-_]rop.
6585 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6588 * cgraphunit.c (symbol_table::compile): Switch to text_section
6589 before calling assembly_start debug hook.
6590 * run-rtl-passes.c (run_rtl_passes): Likewise.
6593 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6595 * tree-vrp.c (extract_range_from_binary_expr_1): Check
6596 range_int_cst_p rather than !symbolic_range_p before calling
6597 extract_range_from_multiplicative_op_1.
6599 2018-01-04 Jeff Law <law@redhat.com>
6601 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
6602 redundant test in assertion.
6604 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6606 * doc/rtl.texi: Document machine_mode wrapper classes.
6608 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6610 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
6613 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6615 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
6616 the VEC_PERM_EXPR fold to fail.
6618 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6621 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
6622 to switched_sections.
6624 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6627 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
6630 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
6633 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
6634 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
6636 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6639 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
6640 is BLKmode and bitpos not zero or mode change is needed.
6642 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6645 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
6648 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
6651 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
6652 instead of MULT rtx. Update all corresponding splitters.
6654 (*ssub<modesuffix>): Ditto.
6656 (*cmp_sadd_di): Update split patterns.
6657 (*cmp_sadd_si): Ditto.
6658 (*cmp_sadd_sidi): Ditto.
6659 (*cmp_ssub_di): Ditto.
6660 (*cmp_ssub_si): Ditto.
6661 (*cmp_ssub_sidi): Ditto.
6662 * config/alpha/predicates.md (const23_operand): New predicate.
6663 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
6664 Look for ASHIFT, not MULT inner operand.
6665 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
6667 2018-01-04 Martin Liska <mliska@suse.cz>
6669 PR gcov-profile/83669
6670 * gcov.c (output_intermediate_file): Add version to intermediate
6672 * doc/gcov.texi: Document new field 'version' in intermediate
6673 file format. Fix location of '-k' option of gcov command.
6675 2018-01-04 Martin Liska <mliska@suse.cz>
6678 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
6680 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6682 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
6684 2018-01-03 Martin Sebor <msebor@redhat.com>
6686 PR tree-optimization/83655
6687 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
6688 checking calls with invalid arguments.
6690 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6692 * tree-vect-stmts.c (vect_get_store_rhs): New function.
6693 (vectorizable_mask_load_store): Delete.
6694 (vectorizable_call): Return false for masked loads and stores.
6695 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
6696 instead of gimple_assign_rhs1.
6697 (vectorizable_load): Handle IFN_MASK_LOAD.
6698 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
6700 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6702 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
6704 (vectorizable_mask_load_store): ...here.
6705 (vectorizable_load): ...and here.
6707 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6709 * tree-vect-stmts.c (vect_build_all_ones_mask)
6710 (vect_build_zero_merge_argument): New functions, split out from...
6711 (vectorizable_load): ...here.
6713 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6715 * tree-vect-stmts.c (vect_check_store_rhs): New function,
6717 (vectorizable_mask_load_store): ...here.
6718 (vectorizable_store): ...and here.
6720 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6722 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
6724 (vectorizable_mask_load_store): ...here.
6726 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6728 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
6729 (vect_model_store_cost): Take a vec_load_store_type instead of a
6731 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
6732 (vect_model_store_cost): Take a vec_load_store_type instead of a
6734 (vectorizable_mask_load_store): Update accordingly.
6735 (vectorizable_store): Likewise.
6736 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
6738 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6740 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
6741 IFN_MASK_LOAD calls here rather than...
6742 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
6744 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6745 Alan Hayward <alan.hayward@arm.com>
6746 David Sherwood <david.sherwood@arm.com>
6748 * expmed.c (extract_bit_field_1): For vector extracts,
6749 fall back to extract_bit_field_as_subreg if vec_extract
6752 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6753 Alan Hayward <alan.hayward@arm.com>
6754 David Sherwood <david.sherwood@arm.com>
6756 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
6757 they are variable or constant sized.
6758 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
6759 slots for constant-sized data.
6761 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6762 Alan Hayward <alan.hayward@arm.com>
6763 David Sherwood <david.sherwood@arm.com>
6765 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
6766 handling COND_EXPRs with boolean comparisons, try to find a better
6767 basis for the mask type than the boolean itself.
6769 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6771 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
6772 is calculated and how it can be overridden.
6773 * genmodes.c (max_bitsize_mode_any_mode): New variable.
6774 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
6776 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
6779 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6780 Alan Hayward <alan.hayward@arm.com>
6781 David Sherwood <david.sherwood@arm.com>
6783 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
6784 Remove the mode argument.
6785 (aarch64_simd_valid_immediate): Remove the mode and inverse
6787 * config/aarch64/iterators.md (bitsize): New iterator.
6788 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
6789 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
6790 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
6791 aarch64_simd_valid_immediate.
6792 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
6793 (aarch64_reg_or_bic_imm): Likewise.
6794 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
6795 with an insn_type enum and msl with a modifier_type enum.
6796 Replace element_width with a scalar_mode. Change the shift
6797 to unsigned int. Add constructors for scalar_float_mode and
6798 scalar_int_mode elements.
6799 (aarch64_vect_float_const_representable_p): Delete.
6800 (aarch64_can_const_movi_rtx_p)
6801 (aarch64_simd_scalar_immediate_valid_for_move)
6802 (aarch64_simd_make_constant): Update call to
6803 aarch64_simd_valid_immediate.
6804 (aarch64_advsimd_valid_immediate_hs): New function.
6805 (aarch64_advsimd_valid_immediate): Likewise.
6806 (aarch64_simd_valid_immediate): Remove mode and inverse
6807 arguments. Rewrite to use the above. Use const_vec_duplicate_p
6808 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
6809 and aarch64_float_const_representable_p on the result.
6810 (aarch64_output_simd_mov_immediate): Remove mode argument.
6811 Update call to aarch64_simd_valid_immediate and use of
6812 simd_immediate_info.
6813 (aarch64_output_scalar_simd_mov_immediate): Update call
6816 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6817 Alan Hayward <alan.hayward@arm.com>
6818 David Sherwood <david.sherwood@arm.com>
6820 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
6821 (mode_nunits): Likewise CONST_MODE_NUNITS.
6822 * machmode.def (ADJUST_NUNITS): Document.
6823 * genmodes.c (mode_data::need_nunits_adj): New field.
6824 (blank_mode): Update accordingly.
6825 (adj_nunits): New variable.
6826 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
6828 (emit_mode_size_inline): Set need_bytesize_adj for all modes
6829 listed in adj_nunits.
6830 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
6831 listed in adj_nunits. Don't emit case statements for such modes.
6832 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
6833 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
6834 nothing if adj_nunits is nonnull.
6835 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
6836 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
6837 (emit_mode_fbit): Update use of print_maybe_const_decl.
6838 (emit_move_size): Likewise. Treat the array as non-const
6840 (emit_mode_adjustments): Handle adj_nunits.
6842 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6844 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
6845 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
6846 (VECTOR_MODES): Use it.
6847 (make_vector_modes): Take the prefix as an argument.
6849 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6850 Alan Hayward <alan.hayward@arm.com>
6851 David Sherwood <david.sherwood@arm.com>
6853 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
6854 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
6855 for MODE_VECTOR_BOOL.
6856 * machmode.def (VECTOR_BOOL_MODE): Document.
6857 * genmodes.c (VECTOR_BOOL_MODE): New macro.
6858 (make_vector_bool_mode): New function.
6859 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
6861 * lto-streamer-in.c (lto_input_mode_table): Likewise.
6862 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
6864 * stor-layout.c (int_mode_for_mode): Likewise.
6865 * tree.c (build_vector_type_for_mode): Likewise.
6866 * varasm.c (output_constant_pool_2): Likewise.
6867 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
6868 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
6869 for MODE_VECTOR_BOOL.
6870 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
6871 of mode class checks.
6872 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
6873 instead of a list of mode class checks.
6874 (expand_vector_scalar_condition): Likewise.
6875 (type_for_widest_vector_mode): Handle BImode as an inner mode.
6877 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6878 Alan Hayward <alan.hayward@arm.com>
6879 David Sherwood <david.sherwood@arm.com>
6881 * machmode.h (mode_size): Change from unsigned short to
6883 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
6884 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6885 or if measurement_type is not polynomial.
6886 (fixed_size_mode::includes_p): Check for constant-sized modes.
6887 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
6888 return a poly_uint16 rather than an unsigned short.
6889 (emit_mode_size): Change the type of mode_size from unsigned short
6890 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
6891 (emit_mode_adjustments): Cope with polynomial vector sizes.
6892 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6894 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6896 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
6897 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
6898 * caller-save.c (setup_save_areas): Likewise.
6899 (replace_reg_with_saved_mem): Likewise.
6900 * calls.c (emit_library_call_value_1): Likewise.
6901 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
6902 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
6903 (gen_lowpart_for_combine): Likewise.
6904 * convert.c (convert_to_integer_1): Likewise.
6905 * cse.c (equiv_constant, cse_insn): Likewise.
6906 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
6907 (cselib_subst_to_values): Likewise.
6908 * dce.c (word_dce_process_block): Likewise.
6909 * df-problems.c (df_word_lr_mark_ref): Likewise.
6910 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
6911 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
6912 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
6913 (rtl_for_decl_location): Likewise.
6914 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
6915 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
6916 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
6917 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
6918 (expand_expr_real_1): Likewise.
6919 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
6920 (pad_below): Likewise.
6921 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6922 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
6923 * ira.c (get_subreg_tracking_sizes): Likewise.
6924 * ira-build.c (ira_create_allocno_objects): Likewise.
6925 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
6926 (ira_sort_regnos_for_alter_reg): Likewise.
6927 * ira-costs.c (record_operand_costs): Likewise.
6928 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
6929 (resolve_simple_move): Likewise.
6930 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
6931 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
6932 (lra_constraints): Likewise.
6933 (CONST_POOL_OK_P): Reject variable-sized modes.
6934 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
6935 (add_pseudo_to_slot, lra_spill): Likewise.
6936 * omp-low.c (omp_clause_aligned_alignment): Likewise.
6937 * optabs-query.c (get_best_extraction_insn): Likewise.
6938 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
6939 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
6940 (expand_mult_highpart, valid_multiword_target_p): Likewise.
6941 * recog.c (offsettable_address_addr_space_p): Likewise.
6942 * regcprop.c (maybe_mode_change): Likewise.
6943 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
6944 * regrename.c (build_def_use): Likewise.
6945 * regstat.c (dump_reg_info): Likewise.
6946 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
6947 (find_reloads, find_reloads_subreg_address): Likewise.
6948 * reload1.c (eliminate_regs_1): Likewise.
6949 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
6950 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
6951 (simplify_binary_operation_1, simplify_subreg): Likewise.
6952 * targhooks.c (default_function_arg_padding): Likewise.
6953 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
6954 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
6955 (verify_gimple_assign_ternary): Likewise.
6956 * tree-inline.c (estimate_move_cost): Likewise.
6957 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
6958 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
6959 (get_address_cost_ainc): Likewise.
6960 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
6961 (vect_supportable_dr_alignment): Likewise.
6962 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6963 (vectorizable_reduction): Likewise.
6964 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
6965 (vectorizable_operation, vectorizable_load): Likewise.
6966 * tree.c (build_same_sized_truth_vector_type): Likewise.
6967 * valtrack.c (cleanup_auto_inc_dec): Likewise.
6968 * var-tracking.c (emit_note_insn_var_location): Likewise.
6969 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
6970 (ADDR_VEC_ALIGN): Likewise.
6972 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6973 Alan Hayward <alan.hayward@arm.com>
6974 David Sherwood <david.sherwood@arm.com>
6976 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
6978 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6979 or if measurement_type is polynomial.
6980 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
6981 * combine.c (make_extraction): Likewise.
6982 * dse.c (find_shift_sequence): Likewise.
6983 * dwarf2out.c (mem_loc_descriptor): Likewise.
6984 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
6985 (extract_bit_field, extract_low_bits): Likewise.
6986 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
6987 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
6988 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
6989 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
6990 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6991 * reload.c (find_reloads): Likewise.
6992 * reload1.c (alter_reg): Likewise.
6993 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
6994 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
6995 * tree-if-conv.c (predicate_mem_writes): Likewise.
6996 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
6997 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
6998 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
6999 * valtrack.c (dead_debug_insert_temp): Likewise.
7000 * varasm.c (mergeable_constant_section): Likewise.
7001 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
7003 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7004 Alan Hayward <alan.hayward@arm.com>
7005 David Sherwood <david.sherwood@arm.com>
7007 * expr.c (expand_assignment): Cope with polynomial mode sizes
7008 when assigning to a CONCAT.
7010 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7011 Alan Hayward <alan.hayward@arm.com>
7012 David Sherwood <david.sherwood@arm.com>
7014 * machmode.h (mode_precision): Change from unsigned short to
7016 (mode_to_precision): Return a poly_uint16 rather than an unsigned
7018 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
7019 or if measurement_type is not polynomial.
7020 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
7021 in which the mode is already known to be a scalar_int_mode.
7022 * genmodes.c (emit_mode_precision): Change the type of mode_precision
7023 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
7025 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7026 for GET_MODE_PRECISION.
7027 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7028 for GET_MODE_PRECISION.
7029 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
7031 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
7032 (expand_field_assignment, make_extraction): Likewise.
7033 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
7034 (get_last_value): Likewise.
7035 * convert.c (convert_to_integer_1): Likewise.
7036 * cse.c (cse_insn): Likewise.
7037 * expr.c (expand_expr_real_1): Likewise.
7038 * lra-constraints.c (simplify_operand_subreg): Likewise.
7039 * optabs-query.c (can_atomic_load_p): Likewise.
7040 * optabs.c (expand_atomic_load): Likewise.
7041 (expand_atomic_store): Likewise.
7042 * ree.c (combine_reaching_defs): Likewise.
7043 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
7044 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
7045 * tree.h (type_has_mode_precision_p): Likewise.
7046 * ubsan.c (instrument_si_overflow): Likewise.
7048 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7049 Alan Hayward <alan.hayward@arm.com>
7050 David Sherwood <david.sherwood@arm.com>
7052 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
7053 polynomial numbers of units.
7054 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
7055 (valid_vector_subparts_p): New function.
7056 (build_vector_type): Remove temporary shim and take the number
7057 of units as a poly_uint64 rather than an int.
7058 (build_opaque_vector_type): Take the number of units as a
7059 poly_uint64 rather than an int.
7060 * tree.c (build_vector_from_ctor): Handle polynomial
7061 TYPE_VECTOR_SUBPARTS.
7062 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
7063 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
7064 (build_vector_from_val): If the number of units is variable,
7065 use build_vec_duplicate_cst for constant operands and
7066 VEC_DUPLICATE_EXPR otherwise.
7067 (make_vector_type): Remove temporary is_constant ().
7068 (build_vector_type, build_opaque_vector_type): Take the number of
7069 units as a poly_uint64 rather than an int.
7070 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
7072 * cfgexpand.c (expand_debug_expr): Likewise.
7073 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
7074 (store_constructor, expand_expr_real_1): Likewise.
7075 (const_scalar_mask_from_tree): Likewise.
7076 * fold-const-call.c (fold_const_reduction): Likewise.
7077 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
7078 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
7079 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
7080 (fold_relational_const): Likewise.
7081 (native_interpret_vector): Likewise. Change the size from an
7082 int to an unsigned int.
7083 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
7084 TYPE_VECTOR_SUBPARTS.
7085 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
7086 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
7087 duplicating a non-constant operand into a variable-length vector.
7088 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
7089 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
7090 * ipa-icf.c (sem_variable::equals): Likewise.
7091 * match.pd: Likewise.
7092 * omp-simd-clone.c (simd_clone_subparts): Likewise.
7093 * print-tree.c (print_node): Likewise.
7094 * stor-layout.c (layout_type): Likewise.
7095 * targhooks.c (default_builtin_vectorization_cost): Likewise.
7096 * tree-cfg.c (verify_gimple_comparison): Likewise.
7097 (verify_gimple_assign_binary): Likewise.
7098 (verify_gimple_assign_ternary): Likewise.
7099 (verify_gimple_assign_single): Likewise.
7100 * tree-pretty-print.c (dump_generic_node): Likewise.
7101 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7102 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
7103 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
7104 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
7105 (vect_shift_permute_load_chain): Likewise.
7106 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
7107 (expand_vector_condition, optimize_vector_constructor): Likewise.
7108 (lower_vec_perm, get_compute_type): Likewise.
7109 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
7110 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
7111 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
7112 (vect_recog_mask_conversion_pattern): Likewise.
7113 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
7114 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
7115 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7116 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
7117 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
7118 (vectorizable_shift, vectorizable_operation, vectorizable_store)
7119 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
7120 (supportable_widening_operation): Likewise.
7121 (supportable_narrowing_operation): Likewise.
7122 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
7124 * varasm.c (output_constant): Likewise.
7126 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7127 Alan Hayward <alan.hayward@arm.com>
7128 David Sherwood <david.sherwood@arm.com>
7130 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
7131 so that both the length == 3 and length != 3 cases set up their
7132 own permute vectors. Add comments explaining why we know the
7133 number of elements is constant.
7134 (vect_permute_load_chain): Likewise.
7136 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7137 Alan Hayward <alan.hayward@arm.com>
7138 David Sherwood <david.sherwood@arm.com>
7140 * machmode.h (mode_nunits): Change from unsigned char to
7142 (ONLY_FIXED_SIZE_MODES): New macro.
7143 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
7144 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
7145 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
7147 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
7148 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
7149 or if measurement_type is not polynomial.
7150 * genmodes.c (ZERO_COEFFS): New macro.
7151 (emit_mode_nunits_inline): Make mode_nunits_inline return a
7153 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
7154 Use ZERO_COEFFS when emitting initializers.
7155 * data-streamer.h (bp_pack_poly_value): New function.
7156 (bp_unpack_poly_value): Likewise.
7157 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7158 for GET_MODE_NUNITS.
7159 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7160 for GET_MODE_NUNITS.
7161 * tree.c (make_vector_type): Remove temporary shim and make
7162 the real function take the number of units as a poly_uint64
7164 (build_vector_type_for_mode): Handle polynomial nunits.
7165 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
7166 * emit-rtl.c (const_vec_series_p_1): Likewise.
7167 (gen_rtx_CONST_VECTOR): Likewise.
7168 * fold-const.c (test_vec_duplicate_folding): Likewise.
7169 * genrecog.c (validate_pattern): Likewise.
7170 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
7171 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7172 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
7173 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
7174 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
7175 * rtlanal.c (subreg_get_info): Likewise.
7176 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7177 (vect_grouped_load_supported): Likewise.
7178 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
7179 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
7180 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
7181 (simplify_const_unary_operation, simplify_binary_operation_1)
7182 (simplify_const_binary_operation, simplify_ternary_operation)
7183 (test_vector_ops_duplicate, test_vector_ops): Likewise.
7184 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
7185 instead of CONST_VECTOR_NUNITS.
7186 * varasm.c (output_constant_pool_2): Likewise.
7187 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
7188 explicit-encoded elements in the XVEC for variable-length vectors.
7190 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7192 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
7194 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7195 Alan Hayward <alan.hayward@arm.com>
7196 David Sherwood <david.sherwood@arm.com>
7198 * coretypes.h (fixed_size_mode): Declare.
7199 (fixed_size_mode_pod): New typedef.
7200 * builtins.h (target_builtins::x_apply_args_mode)
7201 (target_builtins::x_apply_result_mode): Change type to
7202 fixed_size_mode_pod.
7203 * builtins.c (apply_args_size, apply_result_size, result_vector)
7204 (expand_builtin_apply_args_1, expand_builtin_apply)
7205 (expand_builtin_return): Update accordingly.
7207 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7209 * cse.c (hash_rtx_cb): Hash only the encoded elements.
7210 * cselib.c (cselib_hash_rtx): Likewise.
7211 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
7212 CONST_VECTOR encoding.
7214 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7215 Jeff Law <law@redhat.com>
7218 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
7219 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
7220 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
7221 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
7224 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
7225 explicitly probe *sp in a noreturn function if there were any callee
7226 register saves or frame pointer is needed.
7228 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7231 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
7232 BLKmode for ternary, binary or unary expressions.
7235 * var-tracking.c (delete_vta_debug_insn): New inline function.
7236 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
7237 insns from get_insns () to NULL instead of each bb separately.
7238 Use delete_vta_debug_insn. No longer static.
7239 (vt_debug_insns_local, variable_tracking_main_1): Adjust
7240 delete_vta_debug_insns callers.
7241 * rtl.h (delete_vta_debug_insns): Declare.
7242 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
7243 instead of variable_tracking_main.
7245 2018-01-03 Martin Sebor <msebor@redhat.com>
7247 PR tree-optimization/83603
7248 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
7249 arguments past the endof the argument list in functions declared
7250 without a prototype.
7251 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
7252 Avoid checking when arguments are null.
7254 2018-01-03 Martin Sebor <msebor@redhat.com>
7257 * doc/extend.texi (attribute const): Fix a typo.
7258 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
7259 issuing -Wsuggest-attribute for void functions.
7261 2018-01-03 Martin Sebor <msebor@redhat.com>
7263 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
7264 offset_int::from instead of wide_int::to_shwi.
7265 (maybe_diag_overlap): Remove assertion.
7266 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
7267 * gimple-ssa-sprintf.c (format_directive): Same.
7268 (parse_directive): Same.
7269 (sprintf_dom_walker::compute_format_length): Same.
7270 (try_substitute_return_value): Same.
7272 2018-01-03 Jeff Law <law@redhat.com>
7275 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
7276 non-constant residual for zero at runtime and avoid probing in
7277 that case. Reorganize code for trailing problem to mirror handling
7280 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7282 PR tree-optimization/83501
7283 * tree-ssa-strlen.c (get_string_cst): New.
7284 (handle_char_store): Call get_string_cst.
7286 2018-01-03 Martin Liska <mliska@suse.cz>
7288 PR tree-optimization/83593
7289 * tree-ssa-strlen.c: Include tree-cfg.h.
7290 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
7291 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
7292 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
7294 (strlen_dom_walker::before_dom_children): Call
7295 gimple_purge_dead_eh_edges. Dump tranformation with details
7297 (strlen_dom_walker::before_dom_children): Update call by adding
7298 new argument cleanup_eh.
7299 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
7301 2018-01-03 Martin Liska <mliska@suse.cz>
7304 * cif-code.def (VARIADIC_THUNK): New enum value.
7305 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
7308 2018-01-03 Jan Beulich <jbeulich@suse.com>
7310 * sse.md (mov<mode>_internal): Tighten condition for when to use
7311 vmovdqu<ssescalarsize> for TI and OI modes.
7313 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7315 Update copyright years.
7317 2018-01-03 Martin Liska <mliska@suse.cz>
7320 * ipa-visibility.c (function_and_variable_visibility): Skip
7321 functions with noipa attribure.
7323 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7325 * gcc.c (process_command): Update copyright notice dates.
7326 * gcov-dump.c (print_version): Ditto.
7327 * gcov.c (print_version): Ditto.
7328 * gcov-tool.c (print_version): Ditto.
7329 * gengtype.c (create_file): Ditto.
7330 * doc/cpp.texi: Bump @copying's copyright year.
7331 * doc/cppinternals.texi: Ditto.
7332 * doc/gcc.texi: Ditto.
7333 * doc/gccint.texi: Ditto.
7334 * doc/gcov.texi: Ditto.
7335 * doc/install.texi: Ditto.
7336 * doc/invoke.texi: Ditto.
7338 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7340 * vector-builder.h (vector_builder::m_full_nelts): Change from
7341 unsigned int to poly_uint64.
7342 (vector_builder::full_nelts): Update prototype accordingly.
7343 (vector_builder::new_vector): Likewise.
7344 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
7345 (vector_builder::operator ==): Likewise.
7346 (vector_builder::finalize): Likewise.
7347 * int-vector-builder.h (int_vector_builder::int_vector_builder):
7348 Take the number of elements as a poly_uint64 rather than an
7350 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
7351 from unsigned int to poly_uint64.
7352 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
7353 (vec_perm_indices::new_vector): Likewise.
7354 (vec_perm_indices::length): Likewise.
7355 (vec_perm_indices::nelts_per_input): Likewise.
7356 (vec_perm_indices::input_nelts): Likewise.
7357 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
7358 number of elements per input as a poly_uint64 rather than an
7359 unsigned int. Use the original encoding for variable-length
7360 vectors, rather than clamping each individual element.
7361 For the second and subsequent elements in each pattern,
7362 clamp the step and base before clamping their sum.
7363 (vec_perm_indices::series_p): Handle polynomial element counts.
7364 (vec_perm_indices::all_in_range_p): Likewise.
7365 (vec_perm_indices_to_tree): Likewise.
7366 (vec_perm_indices_to_rtx): Likewise.
7367 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
7368 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
7369 (tree_vector_builder::new_binary_operation): Handle polynomial
7370 element counts. Return false if we need to know the number
7371 of elements at compile time.
7372 * fold-const.c (fold_vec_perm): Punt if the number of elements
7373 isn't known at compile time.
7375 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7377 * vec-perm-indices.h (vec_perm_builder): Change element type
7378 from HOST_WIDE_INT to poly_int64.
7379 (vec_perm_indices::element_type): Update accordingly.
7380 (vec_perm_indices::clamp): Handle polynomial element_types.
7381 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7382 (vec_perm_indices::all_in_range_p): Likewise.
7383 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
7385 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
7386 polynomial vec_perm_indices element types.
7387 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
7388 * fold-const.c (fold_vec_perm): Likewise.
7389 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
7390 * tree-vect-generic.c (lower_vec_perm): Likewise.
7391 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7392 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
7393 element type to HOST_WIDE_INT.
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 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
7400 rather than an int. Use plus_constant.
7401 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
7402 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
7404 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7405 Alan Hayward <alan.hayward@arm.com>
7406 David Sherwood <david.sherwood@arm.com>
7408 * calls.c (emit_call_1, expand_call): Change struct_value_size from
7409 a HOST_WIDE_INT to a poly_int64.
7411 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7412 Alan Hayward <alan.hayward@arm.com>
7413 David Sherwood <david.sherwood@arm.com>
7415 * calls.c (load_register_parameters): Cope with polynomial
7416 mode sizes. Require a constant size for BLKmode parameters
7417 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
7418 forces a parameter to be padded at the lsb end in order to
7419 fill a complete number of words, require the parameter size
7420 to be ordered wrt UNITS_PER_WORD.
7422 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7423 Alan Hayward <alan.hayward@arm.com>
7424 David Sherwood <david.sherwood@arm.com>
7426 * reload1.c (spill_stack_slot_width): Change element type
7427 from unsigned int to poly_uint64_pod.
7428 (alter_reg): Treat mode sizes as polynomial.
7430 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7431 Alan Hayward <alan.hayward@arm.com>
7432 David Sherwood <david.sherwood@arm.com>
7434 * reload.c (complex_word_subreg_p): New function.
7435 (reload_inner_reg_of_subreg, push_reload): Use it.
7437 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7438 Alan Hayward <alan.hayward@arm.com>
7439 David Sherwood <david.sherwood@arm.com>
7441 * lra-constraints.c (process_alt_operands): Reject matched
7442 operands whose sizes aren't ordered.
7443 (match_reload): Refer to this check here.
7445 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7446 Alan Hayward <alan.hayward@arm.com>
7447 David Sherwood <david.sherwood@arm.com>
7449 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
7450 that the mode size is in the set {1, 2, 4, 8, 16}.
7452 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7453 Alan Hayward <alan.hayward@arm.com>
7454 David Sherwood <david.sherwood@arm.com>
7456 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
7457 Use plus_constant instead of gen_rtx_PLUS.
7459 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7460 Alan Hayward <alan.hayward@arm.com>
7461 David Sherwood <david.sherwood@arm.com>
7463 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
7464 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
7465 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
7466 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
7467 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
7468 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
7469 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
7470 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
7471 * config/i386/i386.c (ix86_push_rounding): ...this new function.
7472 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
7474 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
7475 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
7476 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
7477 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
7478 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
7479 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
7480 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
7481 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
7482 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
7483 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
7485 * expr.c (emit_move_resolve_push): Treat the input and result
7486 of PUSH_ROUNDING as a poly_int64.
7487 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
7488 (emit_push_insn): Likewise.
7489 * lra-eliminations.c (mark_not_eliminable): Likewise.
7490 * recog.c (push_operand): Likewise.
7491 * reload1.c (elimination_effects): Likewise.
7492 * rtlanal.c (nonzero_bits1): Likewise.
7493 * calls.c (store_one_arg): Likewise. Require the padding to be
7494 known at compile time.
7496 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7497 Alan Hayward <alan.hayward@arm.com>
7498 David Sherwood <david.sherwood@arm.com>
7500 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
7501 Use plus_constant instead of gen_rtx_PLUS.
7503 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7504 Alan Hayward <alan.hayward@arm.com>
7505 David Sherwood <david.sherwood@arm.com>
7507 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
7510 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7511 Alan Hayward <alan.hayward@arm.com>
7512 David Sherwood <david.sherwood@arm.com>
7514 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
7515 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
7516 via stack temporaries. Treat the mode size as polynomial too.
7518 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7519 Alan Hayward <alan.hayward@arm.com>
7520 David Sherwood <david.sherwood@arm.com>
7522 * expr.c (expand_expr_real_2): When handling conversions involving
7523 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
7524 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
7525 as a poly_uint64 too.
7527 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7528 Alan Hayward <alan.hayward@arm.com>
7529 David Sherwood <david.sherwood@arm.com>
7531 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
7533 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7534 Alan Hayward <alan.hayward@arm.com>
7535 David Sherwood <david.sherwood@arm.com>
7537 * combine.c (can_change_dest_mode): Handle polynomial
7538 REGMODE_NATURAL_SIZE.
7539 * expmed.c (store_bit_field_1): Likewise.
7540 * expr.c (store_constructor): Likewise.
7541 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
7542 and polynomial REGMODE_NATURAL_SIZE.
7543 (gen_lowpart_common): Likewise.
7544 * reginfo.c (record_subregs_of_mode): Likewise.
7545 * rtlanal.c (read_modify_subreg_p): Likewise.
7547 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7548 Alan Hayward <alan.hayward@arm.com>
7549 David Sherwood <david.sherwood@arm.com>
7551 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
7552 numbers of elements.
7554 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7555 Alan Hayward <alan.hayward@arm.com>
7556 David Sherwood <david.sherwood@arm.com>
7558 * match.pd: Cope with polynomial numbers of vector elements.
7560 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7561 Alan Hayward <alan.hayward@arm.com>
7562 David Sherwood <david.sherwood@arm.com>
7564 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
7565 in a POINTER_PLUS_EXPR.
7567 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7568 Alan Hayward <alan.hayward@arm.com>
7569 David Sherwood <david.sherwood@arm.com>
7571 * omp-simd-clone.c (simd_clone_subparts): New function.
7572 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
7573 (ipa_simd_modify_function_body): Likewise.
7575 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7576 Alan Hayward <alan.hayward@arm.com>
7577 David Sherwood <david.sherwood@arm.com>
7579 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
7580 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
7581 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
7582 (expand_vector_condition, vector_element): Likewise.
7583 (subparts_gt): New function.
7584 (get_compute_type): Use subparts_gt.
7585 (count_type_subparts): Delete.
7586 (expand_vector_operations_1): Use subparts_gt instead of
7587 count_type_subparts.
7589 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7590 Alan Hayward <alan.hayward@arm.com>
7591 David Sherwood <david.sherwood@arm.com>
7593 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
7594 (vect_compile_time_alias): ...this new function. Do the calculation
7595 on poly_ints rather than trees.
7596 (vect_prune_runtime_alias_test_list): Update call accordingly.
7598 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7599 Alan Hayward <alan.hayward@arm.com>
7600 David Sherwood <david.sherwood@arm.com>
7602 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
7604 (vect_schedule_slp_instance): Likewise.
7606 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7607 Alan Hayward <alan.hayward@arm.com>
7608 David Sherwood <david.sherwood@arm.com>
7610 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
7611 constant and extern definitions for variable-length vectors.
7612 (vect_get_constant_vectors): Note that the number of units
7613 is known to be constant.
7615 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7616 Alan Hayward <alan.hayward@arm.com>
7617 David Sherwood <david.sherwood@arm.com>
7619 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
7620 of units as polynomial. Choose between WIDE and NARROW based
7623 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7624 Alan Hayward <alan.hayward@arm.com>
7625 David Sherwood <david.sherwood@arm.com>
7627 * tree-vect-stmts.c (simd_clone_subparts): New function.
7628 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
7630 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7631 Alan Hayward <alan.hayward@arm.com>
7632 David Sherwood <david.sherwood@arm.com>
7634 * tree-vect-stmts.c (vectorizable_call): Treat the number of
7635 vectors as polynomial. Use build_index_vector for
7638 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7639 Alan Hayward <alan.hayward@arm.com>
7640 David Sherwood <david.sherwood@arm.com>
7642 * tree-vect-stmts.c (get_load_store_type): Treat the number of
7643 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
7644 for variable-length vectors.
7645 (vectorizable_mask_load_store): Treat the number of units as
7646 polynomial, asserting that it is constant if the condition has
7647 already been enforced.
7648 (vectorizable_store, vectorizable_load): Likewise.
7650 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7651 Alan Hayward <alan.hayward@arm.com>
7652 David Sherwood <david.sherwood@arm.com>
7654 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
7655 of units as polynomial. Punt if we can't tell at compile time
7656 which vector contains the final result.
7658 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7659 Alan Hayward <alan.hayward@arm.com>
7660 David Sherwood <david.sherwood@arm.com>
7662 * tree-vect-loop.c (vectorizable_induction): Treat the number
7663 of units as polynomial. Punt on SLP inductions. Use an integer
7664 VEC_SERIES_EXPR for variable-length integer reductions. Use a
7665 cast of such a series for variable-length floating-point
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.h (build_index_vector): Declare.
7673 * tree.c (build_index_vector): New function.
7674 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
7675 of units as polynomial, forcibly converting it to a constant if
7676 vectorizable_reduction has already enforced the condition.
7677 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
7678 to create a {1,2,3,...} vector.
7679 (vectorizable_reduction): Treat the number of units as polynomial.
7680 Choose vectype_in based on the largest scalar element size rather
7681 than the smallest number of units. Enforce the restrictions
7684 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7685 Alan Hayward <alan.hayward@arm.com>
7686 David Sherwood <david.sherwood@arm.com>
7688 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
7689 number of units as polynomial.
7691 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7692 Alan Hayward <alan.hayward@arm.com>
7693 David Sherwood <david.sherwood@arm.com>
7695 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
7696 * target.def (autovectorize_vector_sizes): Return the vector sizes
7697 by pointer, using vector_sizes rather than a bitmask.
7698 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
7699 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
7700 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
7702 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
7703 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
7704 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
7705 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
7706 * omp-general.c (omp_max_vf): Likewise.
7707 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7708 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
7709 * tree-vect-loop.c (vect_analyze_loop): Likewise.
7710 * tree-vect-slp.c (vect_slp_bb): Likewise.
7711 * doc/tm.texi: Regenerate.
7712 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
7714 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
7715 the vector size as a poly_uint64 rather than an unsigned int.
7716 (current_vector_size): Change from an unsigned int to a poly_uint64.
7717 (get_vectype_for_scalar_type): Update accordingly.
7718 * tree.h (build_truth_vector_type): Take the size and number of
7719 units as a poly_uint64 rather than an unsigned int.
7720 (build_vector_type): Add a temporary overload that takes
7721 the number of units as a poly_uint64 rather than an unsigned int.
7722 * tree.c (make_vector_type): Likewise.
7723 (build_truth_vector_type): Take the number of units as a poly_uint64
7724 rather than an unsigned int.
7726 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7727 Alan Hayward <alan.hayward@arm.com>
7728 David Sherwood <david.sherwood@arm.com>
7730 * target.def (get_mask_mode): Take the number of units and length
7731 as poly_uint64s rather than unsigned ints.
7732 * targhooks.h (default_get_mask_mode): Update accordingly.
7733 * targhooks.c (default_get_mask_mode): Likewise.
7734 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
7735 * doc/tm.texi: Regenerate.
7737 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7738 Alan Hayward <alan.hayward@arm.com>
7739 David Sherwood <david.sherwood@arm.com>
7741 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
7742 * omp-general.c (omp_max_vf): Likewise.
7743 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
7744 (expand_omp_simd): Handle polynomial safelen.
7745 * omp-low.c (omplow_simd_context): Add a default constructor.
7746 (omplow_simd_context::max_vf): Change from int to poly_uint64.
7747 (lower_rec_simd_input_clauses): Update accordingly.
7748 (lower_rec_input_clauses): Likewise.
7750 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7751 Alan Hayward <alan.hayward@arm.com>
7752 David Sherwood <david.sherwood@arm.com>
7754 * tree-vectorizer.h (vect_nunits_for_cost): New function.
7755 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
7756 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
7757 (vect_analyze_slp_cost): Likewise.
7758 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
7759 (vect_model_load_cost): Likewise.
7761 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7762 Alan Hayward <alan.hayward@arm.com>
7763 David Sherwood <david.sherwood@arm.com>
7765 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
7766 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
7767 from an unsigned int * to a poly_uint64_pod *.
7768 (calculate_unrolling_factor): New function.
7769 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
7771 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7772 Alan Hayward <alan.hayward@arm.com>
7773 David Sherwood <david.sherwood@arm.com>
7775 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
7776 from an unsigned int to a poly_uint64.
7777 (_loop_vec_info::slp_unrolling_factor): Likewise.
7778 (_loop_vec_info::vectorization_factor): Change from an int
7780 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
7781 (vect_get_num_vectors): New function.
7782 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
7783 (vect_get_num_copies): Use vect_get_num_vectors.
7784 (vect_analyze_data_ref_dependences): Change max_vf from an int *
7785 to an unsigned int *.
7786 (vect_analyze_data_refs): Change min_vf from an int * to a
7788 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7789 than an unsigned HOST_WIDE_INT.
7790 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
7791 (vect_analyze_data_ref_dependence): Change max_vf from an int *
7792 to an unsigned int *.
7793 (vect_analyze_data_ref_dependences): Likewise.
7794 (vect_compute_data_ref_alignment): Handle polynomial vf.
7795 (vect_enhance_data_refs_alignment): Likewise.
7796 (vect_prune_runtime_alias_test_list): Likewise.
7797 (vect_shift_permute_load_chain): Likewise.
7798 (vect_supportable_dr_alignment): Likewise.
7799 (dependence_distance_ge_vf): Take the vectorization factor as a
7800 poly_uint64 rather than an unsigned HOST_WIDE_INT.
7801 (vect_analyze_data_refs): Change min_vf from an int * to a
7803 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
7804 vfm1 as a poly_uint64 rather than an int. Make the same change
7805 for the returned bound_scalar.
7806 (vect_gen_vector_loop_niters): Handle polynomial vf.
7807 (vect_do_peeling): Likewise. Update call to
7808 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
7809 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
7811 * tree-vect-loop.c (vect_determine_vectorization_factor)
7812 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
7813 (vect_get_known_peeling_cost): Likewise.
7814 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
7815 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
7816 (vect_transform_loop): Likewise. Use the lowest possible VF when
7817 updating the upper bounds of the loop.
7818 (vect_min_worthwhile_factor): Make static. Return an unsigned int
7820 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
7821 polynomial unroll factors.
7822 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
7823 (vect_make_slp_decision): Likewise.
7824 (vect_supported_load_permutation_p): Likewise, and polynomial
7826 (vect_analyze_slp_cost): Handle polynomial vf.
7827 (vect_slp_analyze_node_operations): Likewise.
7828 (vect_slp_analyze_bb_1): Likewise.
7829 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7830 than an unsigned HOST_WIDE_INT.
7831 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
7832 (vectorizable_load): Handle polynomial vf.
7833 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
7835 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
7837 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7838 Alan Hayward <alan.hayward@arm.com>
7839 David Sherwood <david.sherwood@arm.com>
7841 * match.pd: Handle bit operations involving three constants
7842 and try to fold one pair.
7844 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7846 * tree-vect-loop-manip.c: Include gimple-fold.h.
7847 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
7848 niters_maybe_zero parameters. Handle other cases besides a step of 1.
7849 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
7850 Add a path that uses a step of VF instead of 1, but disable it
7852 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
7853 and niters_no_overflow parameters. Update calls to
7854 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
7855 Create a new SSA name if the latter choses to use a ste other
7856 than zero, and return it via niters_vector_mult_vf_var.
7857 * tree-vect-loop.c (vect_transform_loop): Update calls to
7858 vect_do_peeling, vect_gen_vector_loop_niters and
7859 slpeel_make_loop_iterate_ntimes.
7860 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
7861 (vect_gen_vector_loop_niters): Update declarations after above changes.
7863 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
7865 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
7866 128-bit round to integer instructions.
7867 (ceil<mode>2): Likewise.
7868 (btrunc<mode>2): Likewise.
7869 (round<mode>2): Likewise.
7871 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7873 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
7874 unaligned VSX load/store on P8/P9.
7875 (expand_block_clear): Allow the use of unaligned VSX
7876 load/store on P8/P9.
7878 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7880 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
7882 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
7883 swap associated with both a load and a store.
7885 2018-01-02 Andrew Waterman <andrew@sifive.com>
7887 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
7888 * config/riscv/riscv.md (clear_cache): Use it.
7890 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
7892 * web.c: Remove out-of-date comment.
7894 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7896 * expr.c (fixup_args_size_notes): Check that any existing
7897 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
7898 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
7899 (emit_single_push_insn): ...here.
7901 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7903 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
7904 (const_vector_encoded_nelts): New function.
7905 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
7906 (const_vector_int_elt, const_vector_elt): Declare.
7907 * emit-rtl.c (const_vector_int_elt_1): New function.
7908 (const_vector_elt): Likewise.
7909 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
7910 of CONST_VECTOR_ELT.
7912 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7914 * expr.c: Include rtx-vector-builder.h.
7915 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
7916 directly on the tree encoding.
7917 (const_vector_from_tree): Likewise.
7918 * optabs.c: Include rtx-vector-builder.h.
7919 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
7920 sequence of "u" values.
7921 * vec-perm-indices.c: Include rtx-vector-builder.h.
7922 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
7923 directly on the vec_perm_indices encoding.
7925 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7927 * doc/rtl.texi (const_vector): Describe new encoding scheme.
7928 * Makefile.in (OBJS): Add rtx-vector-builder.o.
7929 * rtx-vector-builder.h: New file.
7930 * rtx-vector-builder.c: Likewise.
7931 * rtl.h (rtx_def::u2): Add a const_vector field.
7932 (CONST_VECTOR_NPATTERNS): New macro.
7933 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
7934 (CONST_VECTOR_DUPLICATE_P): Likewise.
7935 (CONST_VECTOR_STEPPED_P): Likewise.
7936 (CONST_VECTOR_ENCODED_ELT): Likewise.
7937 (const_vec_duplicate_p): Check for a duplicated vector encoding.
7938 (unwrap_const_vec_duplicate): Likewise.
7939 (const_vec_series_p): Check for a non-duplicated vector encoding.
7940 Say that the function only returns true for integer vectors.
7941 * emit-rtl.c: Include rtx-vector-builder.h.
7942 (gen_const_vec_duplicate_1): Delete.
7943 (gen_const_vector): Call gen_const_vec_duplicate instead of
7944 gen_const_vec_duplicate_1.
7945 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
7946 (gen_const_vec_duplicate): Use rtx_vector_builder.
7947 (gen_const_vec_series): Likewise.
7948 (gen_rtx_CONST_VECTOR): Likewise.
7949 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
7950 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7951 Build a new vector rather than modifying a CONST_VECTOR in-place.
7952 (handle_special_swappables): Update call accordingly.
7953 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
7954 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7955 Build a new vector rather than modifying a CONST_VECTOR in-place.
7956 (handle_special_swappables): Update call accordingly.
7958 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7960 * simplify-rtx.c (simplify_const_binary_operation): Use
7961 CONST_VECTOR_ELT instead of XVECEXP.
7963 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7965 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
7966 the selector elements to be different from the data elements
7967 if the selector is a VECTOR_CST.
7968 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
7969 ssizetype for the selector.
7971 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7973 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
7974 before testing each element individually.
7975 * tree-vect-generic.c (lower_vec_perm): Likewise.
7977 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7979 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
7980 * selftest-run-tests.c (selftest::run_tests): Call it.
7981 * vector-builder.h (vector_builder::operator ==): New function.
7982 (vector_builder::operator !=): Likewise.
7983 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
7984 (vec_perm_indices::all_from_input_p): New function.
7985 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7986 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
7987 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
7988 instead of reading the VECTOR_CST directly. Detect whether both
7989 vector inputs are the same before constructing the vec_perm_indices,
7990 and update the number of inputs argument accordingly. Use the
7991 utility functions added above. Only construct sel2 if we need to.
7993 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7995 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
7996 the broadcast of the low byte.
7997 (expand_mult_highpart): Use an explicit encoding for the permutes.
7998 * optabs-query.c (can_mult_highpart_p): Likewise.
7999 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
8000 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8001 (vectorizable_bswap): Likewise.
8002 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
8003 explicit encoding for the power-of-2 permutes.
8004 (vect_permute_store_chain): Likewise.
8005 (vect_grouped_load_supported): Likewise.
8006 (vect_permute_load_chain): Likewise.
8008 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8010 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
8011 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
8012 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
8013 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
8014 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
8015 (vect_gen_perm_mask_any): Likewise.
8017 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8019 * int-vector-builder.h: New file.
8020 * vec-perm-indices.h: Include int-vector-builder.h.
8021 (vec_perm_indices): Redefine as an int_vector_builder.
8022 (auto_vec_perm_indices): Delete.
8023 (vec_perm_builder): Redefine as a stand-alone class.
8024 (vec_perm_indices::vec_perm_indices): New function.
8025 (vec_perm_indices::clamp): Likewise.
8026 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
8027 (vec_perm_indices::new_vector): New function.
8028 (vec_perm_indices::new_expanded_vector): Update for new
8029 vec_perm_indices class.
8030 (vec_perm_indices::rotate_inputs): New function.
8031 (vec_perm_indices::all_in_range_p): Operate directly on the
8032 encoded form, without computing elided elements.
8033 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
8034 encoding. Update for new vec_perm_indices class.
8035 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
8036 the given vec_perm_builder.
8037 (expand_vec_perm_var): Update vec_perm_builder constructor.
8038 (expand_mult_highpart): Use vec_perm_builder instead of
8039 auto_vec_perm_indices.
8040 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
8041 vec_perm_indices instead of auto_vec_perm_indices. Use a single
8042 or double series encoding as appropriate.
8043 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
8044 vec_perm_indices instead of auto_vec_perm_indices.
8045 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8046 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8047 (vect_permute_store_chain): Likewise.
8048 (vect_grouped_load_supported): Likewise.
8049 (vect_permute_load_chain): Likewise.
8050 (vect_shift_permute_load_chain): Likewise.
8051 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8052 (vect_transform_slp_perm_load): Likewise.
8053 (vect_schedule_slp_instance): Likewise.
8054 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8055 (vectorizable_mask_load_store): Likewise.
8056 (vectorizable_bswap): Likewise.
8057 (vectorizable_store): Likewise.
8058 (vectorizable_load): Likewise.
8059 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
8060 vec_perm_indices instead of auto_vec_perm_indices. Use
8061 tree_to_vec_perm_builder to read the vector from a tree.
8062 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
8063 vec_perm_builder instead of a vec_perm_indices.
8064 (have_whole_vector_shift): Use vec_perm_builder and
8065 vec_perm_indices instead of auto_vec_perm_indices. Leave the
8066 truncation to calc_vec_perm_mask_for_shift.
8067 (vect_create_epilog_for_reduction): Likewise.
8068 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
8069 from auto_vec_perm_indices to vec_perm_indices.
8070 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8071 instead of changing individual elements.
8072 (aarch64_vectorize_vec_perm_const): Use new_vector to install
8073 the vector in d.perm.
8074 * config/arm/arm.c (expand_vec_perm_d::perm): Change
8075 from auto_vec_perm_indices to vec_perm_indices.
8076 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
8077 instead of changing individual elements.
8078 (arm_vectorize_vec_perm_const): Use new_vector to install
8079 the vector in d.perm.
8080 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
8081 Update vec_perm_builder constructor.
8082 (rs6000_expand_interleave): Likewise.
8083 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
8084 (rs6000_expand_interleave): Likewise.
8086 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8088 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
8089 to qimode could truncate the indices.
8090 * optabs.c (expand_vec_perm_var): Likewise.
8092 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8094 * Makefile.in (OBJS): Add vec-perm-indices.o.
8095 * vec-perm-indices.h: New file.
8096 * vec-perm-indices.c: Likewise.
8097 * target.h (vec_perm_indices): Replace with a forward class
8099 (auto_vec_perm_indices): Move to vec-perm-indices.h.
8100 * optabs.h: Include vec-perm-indices.h.
8101 (expand_vec_perm): Delete.
8102 (selector_fits_mode_p, expand_vec_perm_var): Declare.
8103 (expand_vec_perm_const): Declare.
8104 * target.def (vec_perm_const_ok): Replace with...
8105 (vec_perm_const): ...this new hook.
8106 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
8107 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
8108 * doc/tm.texi: Regenerate.
8109 * optabs.def (vec_perm_const): Delete.
8110 * doc/md.texi (vec_perm_const): Likewise.
8111 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
8112 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
8113 expand_vec_perm for constant permutation vectors. Assert that
8114 the mode of variable permutation vectors is the integer equivalent
8115 of the mode that is being permuted.
8116 * optabs-query.h (selector_fits_mode_p): Declare.
8117 * optabs-query.c: Include vec-perm-indices.h.
8118 (selector_fits_mode_p): New function.
8119 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
8120 is defined, instead of checking whether the vec_perm_const_optab
8121 exists. Use targetm.vectorize.vec_perm_const instead of
8122 targetm.vectorize.vec_perm_const_ok. Check whether the indices
8123 fit in the vector mode before using a variable permute.
8124 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
8125 vec_perm_indices instead of an rtx.
8126 (expand_vec_perm): Replace with...
8127 (expand_vec_perm_const): ...this new function. Take the selector
8128 as a vec_perm_indices rather than an rtx. Also take the mode of
8129 the selector. Update call to shift_amt_for_vec_perm_mask.
8130 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
8131 Use vec_perm_indices::new_expanded_vector to expand the original
8132 selector into bytes. Check whether the indices fit in the vector
8133 mode before using a variable permute.
8134 (expand_vec_perm_var): Make global.
8135 (expand_mult_highpart): Use expand_vec_perm_const.
8136 * fold-const.c: Includes vec-perm-indices.h.
8137 * tree-ssa-forwprop.c: Likewise.
8138 * tree-vect-data-refs.c: Likewise.
8139 * tree-vect-generic.c: Likewise.
8140 * tree-vect-loop.c: Likewise.
8141 * tree-vect-slp.c: Likewise.
8142 * tree-vect-stmts.c: Likewise.
8143 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
8145 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
8146 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
8147 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
8148 (aarch64_vectorize_vec_perm_const): ...this new function.
8149 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8150 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8151 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
8152 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
8153 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8154 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8155 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
8157 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
8158 check for NEON modes.
8159 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
8160 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
8161 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
8162 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
8164 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
8165 the old VEC_PERM_CONST conditions.
8166 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
8167 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
8168 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
8169 (ia64_vectorize_vec_perm_const_ok): Merge into...
8170 (ia64_vectorize_vec_perm_const): ...this new function.
8171 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
8172 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
8173 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
8174 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
8175 * config/mips/mips.c (mips_expand_vec_perm_const)
8176 (mips_vectorize_vec_perm_const_ok): Merge into...
8177 (mips_vectorize_vec_perm_const): ...this new function.
8178 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
8179 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
8180 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
8181 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
8182 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
8183 (rs6000_expand_vec_perm_const): Delete.
8184 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
8186 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8187 (altivec_expand_vec_perm_const_le): Take each operand individually.
8188 Operate on constant selectors rather than rtxes.
8189 (altivec_expand_vec_perm_const): Likewise. Update call to
8190 altivec_expand_vec_perm_const_le.
8191 (rs6000_expand_vec_perm_const): Delete.
8192 (rs6000_vectorize_vec_perm_const_ok): Delete.
8193 (rs6000_vectorize_vec_perm_const): New function.
8194 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8195 an element count and rtx array.
8196 (rs6000_expand_extract_even): Update call accordingly.
8197 (rs6000_expand_interleave): Likewise.
8198 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
8199 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
8200 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
8201 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
8202 (rs6000_expand_vec_perm_const): Delete.
8203 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8204 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8205 (altivec_expand_vec_perm_const_le): Take each operand individually.
8206 Operate on constant selectors rather than rtxes.
8207 (altivec_expand_vec_perm_const): Likewise. Update call to
8208 altivec_expand_vec_perm_const_le.
8209 (rs6000_expand_vec_perm_const): Delete.
8210 (rs6000_vectorize_vec_perm_const_ok): Delete.
8211 (rs6000_vectorize_vec_perm_const): New function. Remove stray
8212 reference to the SPE evmerge intructions.
8213 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8214 an element count and rtx array.
8215 (rs6000_expand_extract_even): Update call accordingly.
8216 (rs6000_expand_interleave): Likewise.
8217 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
8218 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
8220 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8222 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8224 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
8225 vector mode and that that mode matches the mode of the data
8227 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
8228 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
8229 directly using expand_vec_perm_1 when forcing selectors into
8231 (expand_vec_perm_var): New function, split out from expand_vec_perm.
8233 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8235 * optabs-query.h (can_vec_perm_p): Delete.
8236 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
8237 * optabs-query.c (can_vec_perm_p): Split into...
8238 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
8239 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
8240 particular selector is valid.
8241 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8242 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8243 (vect_grouped_load_supported): Likewise.
8244 (vect_shift_permute_load_chain): Likewise.
8245 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8246 (vect_transform_slp_perm_load): Likewise.
8247 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8248 (vectorizable_bswap): Likewise.
8249 (vect_gen_perm_mask_checked): Likewise.
8250 * fold-const.c (fold_ternary_loc): Likewise. Don't take
8251 implementations of variable permutation vectors into account
8252 when deciding which selector to use.
8253 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
8254 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
8255 with a false third argument.
8256 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
8257 to test whether the constant selector is valid and can_vec_perm_var_p
8258 to test whether a variable selector is valid.
8260 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8262 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
8263 * optabs-query.c (can_vec_perm_p): Likewise.
8264 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
8265 instead of vec_perm_indices.
8266 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
8267 (vect_gen_perm_mask_checked): Likewise,
8268 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
8269 (vect_gen_perm_mask_checked): Likewise,
8271 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8273 * optabs-query.h (qimode_for_vec_perm): Declare.
8274 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
8275 (qimode_for_vec_perm): ...this new function.
8276 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
8278 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8280 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
8281 does not have a conditional at the top.
8283 2018-01-02 Richard Biener <rguenther@suse.de>
8285 * ipa-inline.c (big_speedup_p): Fix expression.
8287 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8290 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
8293 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8297 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
8298 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
8299 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
8300 cond_taken_branch_cost 3->4.
8302 2018-01-01 Jakub Jelinek <jakub@redhat.com>
8304 PR tree-optimization/83581
8305 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
8306 TODO_cleanup_cfg if any changes have been made.
8309 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
8310 convert_modes if target mode has the right side, but different mode
8314 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
8315 last argument when extracting from CONCAT. If either from_real or
8316 from_imag is NULL, use expansion through memory. If result is not
8317 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
8318 the parts directly to inner mode, if even that fails, use expansion
8322 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
8323 check for bswap in mode rather than HImode and use that in expand_unop
8326 Copyright (C) 2018 Free Software Foundation, Inc.
8328 Copying and distribution of this file, with or without modification,
8329 are permitted in any medium without royalty provided the copyright
8330 notice and this notice are preserved.