1 2018-03-06 Jakub Jelinek <jakub@redhat.com>
3 PR tree-optimization/84687
4 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
6 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
8 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
11 Rename to ppc_speculation_barrier.
12 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
13 __builtin_ppc_speculation_barrier.
15 2018-03-05 Jakub Jelinek <jakub@redhat.com>
18 * combine.c (combine_simplify_rtx): Don't try to simplify if
19 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
22 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
24 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
25 to 32 bytes when compiling for POWER9.
27 2018-03-05 Jakub Jelinek <jakub@redhat.com>
30 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
31 regparm >= 3 with no arg reg available also for calls with
32 flag_force_indirect_call. Pass decl to ix86_function_regparm.
35 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
37 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
39 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
42 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
44 2018-03-05 Richard Biener <rguenther@suse.de>
46 PR tree-optimization/84486
47 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
48 When inserting a __builtin_assume_aligned call set the LHS
49 SSA name alignment info accordingly.
51 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
53 PR tree-optimization/84114
54 * config/aarch64/aarch64.c (aarch64_reassociation_width)
55 Avoid reassociation of FLOAT_MODE addition.
57 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
59 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
60 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
61 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
62 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
63 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
64 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
65 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
67 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
68 __builtin_ia32_wbinvd): New builtins.
70 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
72 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
73 (ix86_valid_target_attribute_inner_p): Ditto.
74 (ix86_init_mmx_sse_builtins): Add special_args2.
75 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
76 TARGET_WBNOINVD_P): New.
77 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
78 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
79 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
80 * config/i386/immintrin.h (_wbinvd): New intrinsic.
81 * config/i386/pconfigintrin.h: New file.
82 * config/i386/wbnoinvdintrin.h: Ditto.
83 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
84 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
86 2018-03-05 Richard Biener <rguenther@suse.de>
88 PR tree-optimization/84670
89 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
91 (BB_VISITED_WITH_VISITED_SUCCS): New define.
92 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
93 (compute_antic_aux): Only assert the number of values in ANTIC_IN
94 doesn't grow if all successors (recursively) were visited at least
97 2018-03-05 Richard Biener <rguenther@suse.de>
99 PR tree-optimization/84650
100 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
101 if executed in the loop pipeline.
103 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
105 * doc/configfiles.texi (Configuration Files): Move info about
106 conditionalizing $target-protos.h to...
107 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
108 differs from $target-protos.h.
110 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
111 Chung-Ju Wu <jasonwucj@gmail.com>
113 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
114 * config/nds32/nds32-multiple.md (setmemsi): Define.
115 * config/nds32/nds32-memory-manipulation.c
116 (nds32_gen_dup_4_byte_to_word_value): New.
117 (emit_setmem_word_loop): New.
118 (emit_setmem_byte_loop): New.
119 (nds32_expand_setmem_loop): New.
120 (nds32_expand_setmem_loop_v3m): New.
121 (nds32_expand_setmem_unroll): New.
122 (nds32_expand_setmem): New.
124 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
125 Chung-Ju Wu <jasonwucj@gmail.com>
127 * config/nds32/nds32-memory-manipulation.c
128 (nds32_emit_load_store): New.
129 (nds32_emit_post_inc_load_store): New.
130 (nds32_emit_mem_move): New.
131 (nds32_emit_mem_move_block): New.
132 (nds32_expand_movmemsi_loop_unknown_size): New.
133 (nds32_expand_movmemsi_loop_known_size): New.
134 (nds32_expand_movmemsi_loop): New.
135 (nds32_expand_movmemsi_unroll): New.
136 (nds32_expand_movmemqi): Rename ...
137 (nds32_expand_movmemsi): ... to this.
138 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
139 (movmemsi): ... to this.
140 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
141 (nds32_expand_movmemsi): ... to this.
143 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
144 Monk Chiang <sh.chiang04@gmail.com>
145 Chung-Ju Wu <jasonwucj@gmail.com>
147 * config/nds32/nds32-protos.h
148 (nds32_expand_load_multiple): New arguments.
149 (nds32_expand_store_multiple): Ditto.
150 (nds32_valid_multiple_load_store): Rename ...
151 (nds32_valid_multiple_load_store_p): ... to this.
152 * config/nds32/nds32-memory-manipulation.c
153 (nds32_expand_load_multiple): Refine implementation.
154 (nds32_expand_store_multiple): Ditto.
155 * config/nds32/nds32-multiple.md
156 (load_multiple): Update nds32_expand_load_multiple interface.
157 (store_multiple): Update nds32_expand_store_multiple interface.
158 * config/nds32/nds32-predicates.c
159 (nds32_valid_multiple_load_store): Rename ...
160 (nds32_valid_multiple_load_store_p): ... to this and refine
162 * config/nds32/predicates.md
163 (nds32_load_multiple_and_update_address_operation): New predicate.
164 (nds32_store_multiple_and_update_address_operation): New predicate.
166 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
167 Chung-Ju Wu <jasonwucj@gmail.com>
169 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
170 (combo): New attribute.
171 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
173 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
175 * config/nds32/nds32.opt: Change -mcmodel= default value.
177 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
178 Monk Chiang <sh.chiang04@gmail.com>
179 Chung-Ju Wu <jasonwucj@gmail.com>
181 * config/nds32/constants.md (unspec_element): New enum.
182 * config/nds32/constraints.md (Umw): New constraint.
183 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
184 * config/nds32/nds32-intrinsic.md: Likewise.
185 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
186 (nds32_valid_smw_lwm_base_p): New.
187 (nds32_output_smw_single_word): New.
188 (nds32_output_lmw_single_word): New.
189 (nds32_expand_unaligned_load): New.
190 (nds32_expand_unaligned_store): New.
191 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
192 (nds32_output_smw_single_word): Declare.
193 (nds32_output_lmw_single_word): Declare.
194 (nds32_expand_unaligned_load): Declare.
195 (nds32_expand_unaligned_store): Declare.
196 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
197 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
198 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
199 NDS32_BUILTIN_UASTORE_DW.
200 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
203 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
204 Kito Cheng <kito.cheng@gmail.com>
205 Chung-Ju Wu <jasonwucj@gmail.com>
207 * config/nds32/nds32-intrinsic.c
208 (nds32_expand_builtin_null_ftype_reg): Delete.
209 (nds32_expand_builtin_reg_ftype_imm): Ditto.
210 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
211 (nds32_read_argument): New.
212 (nds32_legitimize_target): Ditto.
213 (nds32_legitimize_argument): Ditto.
214 (nds32_check_constant_argument): Ditto.
215 (nds32_expand_unop_builtin): Ditto.
216 (nds32_expand_unopimm_builtin): Ditto.
217 (nds32_expand_binop_builtin): Ditto.
218 (nds32_builtin_decl_impl): Ditto.
219 (builtin_description): Ditto.
220 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
221 (nds32_init_builtins_impl): Ditto.
222 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
223 (nds32_builtin_decl): New.
224 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
225 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
227 2018-03-02 Jeff Law <law@redhat.com>
229 * reorg.c (stop_search_p): Handle DEBUG_INSN.
230 (redundant_insn, fill_simple_delay_slots): Likewise.
231 (fill_slots_from_thread): Likewise.
232 * resource.c (mark_referenced_resources): Likewise.
233 (mark_set_resources, find_dead_or_set_registers): Likewise.
235 2018-03-02 Jakub Jelinek <jakub@redhat.com>
237 * substring-locations.h (format_warning_va): Formatting fix for
239 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
241 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
242 * substring-locations.c: Include intl.h.
243 (format_warning_va): Turned into small wrapper around
244 format_warning_n_va, renamed to ...
245 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
246 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
248 (format_warning_at_substring_n): New function.
249 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
250 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
251 format_warning_at_substring with just a shorter name instead of
252 const function pointer.
253 (fmtwarn_n): New function.
254 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
255 appropriate, get rid of all the fmtstr temporaries, move conditionals
256 with G_() wrapped string literals directly into fmtwarn arguments,
257 cast dir.len to (int), formatting fixes.
259 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
261 * doc/invoke.texi: Remove "Cilk Plus" references.
263 2018-03-02 Jakub Jelinek <jakub@redhat.com>
264 Richard Biener <rguenther@suse.de>
267 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
268 for error or warning attributes if CALL_FROM_THUNK_P is set.
271 2018-03-02 Jakub Jelinek <jakub@redhat.com>
274 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
275 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
278 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
279 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
281 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
282 instead of -1U in last predictors element's probability member.
284 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
287 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
288 arguments if they are comparable.
290 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
292 PR tree-optimization/84634
293 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
294 masks and masked_loop_p with a single loop_masks, making sure it's
295 null for bb vectorization.
297 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
299 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
300 (vect_analyze_data_ref_access): Use loop->safe_len rather than
301 loop->force_vectorize to check whether there is no alias.
303 2018-03-02 Jakub Jelinek <jakub@redhat.com>
306 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
308 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
310 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
311 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
312 instead of a loop around prev_real_insn.
313 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
317 * config/i386/i386.c (ix86_print_operand): Use conditional
318 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
321 2018-03-02 Richard Biener <rguenther@suse.de>
323 PR tree-optimization/84427
324 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
325 (bitmap_set_subtract_values): Rewrite to handle multiple
328 (prune_clobbered_mems): Likewise.
329 (phi_translate): Take edge instead of pred/phiblock.
330 (phi_translate_1): Likewise.
331 (phi_translate_set): Likewise. Insert all translated
332 exprs for a value into the set, keeping possibly multiple
333 expressions per value.
334 (compute_antic_aux): Adjust for phi_translate changes.
335 When intersecting union the expressions and prune those
336 not in the final value set, keeping possibly multiple
337 expressions per value. Do not use value-insertion
338 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
339 all expressions. Add verification that the value-sets
340 only shrink during iteration.
341 (compute_partial_antic_aux): Adjust for the phi_translate changes.
342 (do_pre_regular_insertion): Likewise.
343 (do_pre_partial_partial_insertion): Likewise.
345 2018-03-02 Richard Biener <rguenther@suse.de>
348 * config/darwin.c (saved_debug_info_level): New static global.
349 (darwin_asm_lto_start): Disable debug info generation for LTO out.
350 (darwin_asm_lto_end): Restore debug info generation settings.
352 2018-03-01 Martin Liska <mliska@suse.cz>
355 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
358 2018-03-01 Richard Biener <rguenther@suse.de>
361 * dwarf2out.c (gen_variable_die): Properly handle late VLA
362 type annotation with LTO when debug was disabled at compile-time.
364 2018-03-01 Matthew Fortune <mfortune@gmail.com>
366 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
368 (mips_final_postscan_insn): Likewise.
370 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
372 PR rtl-optimization/84528
373 * alias.c (init_alias_target): Add commentary.
374 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
375 a unique base value if the frame pointer is not eliminated
376 to the stack pointer.
378 2018-03-01 Tom de Vries <tom@codesourcery.com>
380 PR rtl-optimization/83327
381 * lra-int.h (hard_regs_spilled_into): Declare.
382 * lra.c (hard_regs_spilled_into): Define.
383 (init_reg_info): Init hard_regs_spilled_into.
384 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
385 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
386 (process_bb_lives): Handle hard_regs_spilled_into.
387 (lra_create_live_ranges_1): Before doing liveness propagation, clear
388 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
390 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
392 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
393 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
394 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
395 * config/rs6000/aix72.h: New file.
397 2018-02-28 Jakub Jelinek <jakub@redhat.com>
399 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
400 instead of warning_at with conditional singular and plural messages
404 * stor-layout.c (update_alignment_for_field): For
405 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
406 && !DECL_PACKED (field), do the alignment update, just use
407 only desired_align instead of MAX (type_align, desired_align)
409 (place_field): Don't do known_align < desired_align handling
410 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
411 is non-NULL, instead do it after rli->prev_field handling and
412 only if not within a bitfield word. For DECL_PACKED (field)
413 use type_align of BITS_PER_UNIT.
415 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
417 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
418 superfluous parentheses and trailing spaces.
420 2018-02-28 Richard Biener <rguenther@suse.de>
422 PR tree-optimization/84584
423 * graphite-scop-detection.c (scop_detection::add_scop): Discard
424 SCoPs with fake exit edge.
426 2018-02-28 Martin Liska <mliska@suse.cz>
429 * timevar.c (timer::print): Fix format to properly print 100%
432 2018-02-28 Richard Biener <rguenther@suse.de>
435 * genmatch.c (capture_info::walk_match): Do not mark
436 captured expressions without operands as expr_p given
437 they act more like predicates and should be subject to
438 "lost tail" side-effect preserving.
440 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
442 PR rtl-optimization/81611
443 * auto-inc-dec.c (attempt_change): Move dead note from
444 mem_insn if it's the next use of regno
445 (find_address): Take address use of reg holding
446 non-incremented value. Add parm to limit search to the named
448 (merge_in_block): Attempt to use a mem insn that is the next
449 use of the original regno.
451 2018-02-27 Martin Sebor <msebor@redhat.com>
454 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
455 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
457 2018-02-27 Martin Sebor <msebor@redhat.com>
460 * diagnostic-core.h (warning_n, error_n, inform_n): Change
461 n argument to unsigned HOST_WIDE_INT.
462 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
463 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
464 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
465 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
467 2018-02-27 Richard Biener <rguenther@suse.de>
469 PR tree-optimization/84512
470 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
471 Do not use the estimate returned from record_stmt_cost for
472 the scalar iteration cost but sum properly using add_stmt_cost.
474 2018-02-27 Richard Biener <rguenther@suse.de>
476 PR tree-optimization/84466
477 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
478 Adjust last change to less strictly validate use operands.
480 2018-02-27 Martin Liska <mliska@suse.cz>
482 PR gcov-profile/84548
483 * gcov.c (process_file): Allow partial overlap and consider it
484 also as group functions.
485 (output_lines): Properly calculate range of lines for a group.
487 2018-02-27 Martin Liska <mliska@suse.cz>
489 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
490 'ggc' suffixes. Change first column width.
491 (timer::print): Fix formatting of the column.
493 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
495 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
496 preserve inline entry blocks for the sake of debug inline
497 entry point markers alone.
498 (remove_unused_locals): Suggest in comments a better place to
499 force the preservation of inline entry blocks that are
500 otherwise unused, but do not preserve them.
502 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
504 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
506 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
509 * config/i386/constraints.md (Bs): Replace
510 ix86_indirect_branch_register with
511 TARGET_INDIRECT_BRANCH_REGISTER.
513 * config/i386/i386.md (indirect_jump): Likewise.
514 (tablejump): Likewise.
515 (*sibcall_memory): Likewise.
516 (*sibcall_value_memory): Likewise.
517 Peepholes of indirect call and jump via memory: Likewise.
518 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
519 (*sibcall_value_GOT_32): Likewise.
520 * config/i386/predicates.md (indirect_branch_operand): Likewise.
521 (GOT_memory_operand): Likewise.
522 (call_insn_operand): Likewise.
523 (sibcall_insn_operand): Likewise.
524 (GOT32_symbol_operand): Likewise.
525 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
527 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
529 PR rtl-optimization/83496
530 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
531 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
532 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
533 redundant insn, if any.
534 (relax_delay_slots): Likewise.
535 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
537 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
539 PR tree-optimization/83965
540 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
541 that grouped statements are part of a reduction chain. Return
542 true if the statement is not marked as a reduction itself but
544 (vect_recog_dot_prod_pattern): Don't check whether the statement
545 is part of a group here.
546 (vect_recog_sad_pattern): Likewise.
547 (vect_recog_widen_sum_pattern): Likewise.
549 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
552 * final.c (rest_of_clean_state): Also look for calls inside sequences.
554 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
557 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
559 (ix86_output_indirect_function_return): New prototype.
560 (ix86_split_simple_return_pop_internal): Likewise.
561 * config/i386/i386.c (indirect_return_via_cx): New.
562 (indirect_return_via_cx_bnd): Likewise.
563 (indirect_thunk_name): Handle return va CX_REG.
564 (output_indirect_thunk_function): Create alias for
565 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
566 (ix86_output_indirect_jmp): Remove the bool argument.
567 (ix86_output_indirect_function_return): New function.
568 (ix86_split_simple_return_pop_internal): Likewise.
569 * config/i386/i386.md (*indirect_jump): Don't pass false
570 to ix86_output_indirect_jmp.
571 (*tablejump_1): Likewise.
572 (simple_return_pop_internal): Change it to define_insn_and_split.
573 Call ix86_split_simple_return_pop_internal to split it for
575 (simple_return_indirect_internal): Call
576 ix86_output_indirect_function_return instead of
577 ix86_output_indirect_jmp.
579 2018-02-26 Jakub Jelinek <jakub@redhat.com>
582 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
583 memset and value initialization afterwards.
585 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
587 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
589 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
592 * common/config/aarch64/aarch64-common.c
593 (aarch_option_optimization_table[]): Switch
594 off fomit-frame-pointer
596 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
597 Chung-Ju Wu <jasonwucj@gmail.com>
599 * config/nds32/nds32-multiple.md (load_multiple): Disallow
601 (store_multiple): Ditto.
603 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
605 * config.gcc: Add --with-cpu support for nds32 target.
606 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
607 * config/nds32/nds32.opt: Add -mcpu= option.
609 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
611 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
612 isel=yes): Warn for these deprecated options.
614 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
616 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
617 ISA_2_5_MASKS_EMBEDDED.
619 2018-02-23 Jakub Jelinek <jakub@redhat.com>
621 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
622 p->max as pointers rather than using iterative_hash_expr.
624 2018-02-23 Carl Love <cel@us.ibm.com>
626 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
627 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
629 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
630 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
631 P8V_BUILTIN_VEC_VUNSIGNED2.
633 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
636 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
637 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
639 * lra-constraints.c (curr_insn_transform): Set up
640 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
642 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
643 * lra-eliminations.c (spill_pseudos): Ditto.
644 (process_insn_for_elimination): Ditto.
645 * lra-lives.c (reg_early_clobber_p): Use the new macros.
646 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
647 LRA_NON_CLOBBERED_ALT.
649 2018-02-22 Martin Sebor <msebor@redhat.com>
651 PR tree-optimization/84480
652 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
653 to maybe_diag_stxncpy_trunc. Call it.
654 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
655 from gimple_fold_builtin_strcpy. Print inlining stack.
656 (handle_builtin_stxncpy): Print inlining stack.
657 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
659 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
662 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
663 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
664 and -fcheck-pointer-bounds are used together.
665 (indirect_thunk_prefix): New enum.
666 (indirect_thunk_need_prefix): New function.
667 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
668 "_nt" instead of "_bnd" for NOTRACK prefix.
669 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
670 (output_indirect_thunk_function): Likewise.
672 (ix86_code_end): Update output_indirect_thunk_function calls.
673 (ix86_output_indirect_branch_via_reg): Replace
674 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
675 (ix86_output_indirect_branch_via_push): Likewise.
676 (ix86_output_function_return): Likewise.
677 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
678 incompatible with -fcf-protection=branch and
679 -fcheck-pointer-bounds.
681 2018-02-22 Steve Ellcey <sellcey@cavium.com>
684 * config/aarch64/aarch64.c (aarch64_print_address_internal):
685 Change gcc_assert call to output_operand_lossage.
687 2018-02-22 Steve Ellcey <sellcey@cavium.com>
689 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
691 2018-02-22 DJ Delorie <dj@redhat.com>
692 Sebastian Perta <sebastian.perta@renesas.com>
693 Oleg Endo <olegendo@gcc.gnu.org>
695 * config/rx/rx.c (rx_rtx_costs): New function.
696 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
698 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
700 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
702 2018-02-22 Martin Liska <mliska@suse.cz>
705 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
706 Add "native" as a possible value.
708 2018-02-22 Martin Liska <mliska@suse.cz>
711 * config/i386/i386.c (ix86_option_override_internal):
712 Add "native" as a possible value for -march and -mtune.
714 2018-02-22 Jakub Jelinek <jakub@redhat.com>
717 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
718 to all type variants.
720 PR tree-optimization/84503
721 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
722 width as info->bitpos + info->bitsize - start.
723 (merged_store_group::merge_overlapping): Simplify width computation.
724 (check_no_overlap): New function.
725 (imm_store_chain_info::try_coalesce_bswap): Compute expected
726 start + width and last_order of the group, fail if check_no_overlap
728 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
729 to group if check_no_overlap fails.
731 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
733 * config/rs6000/altivec.md: Delete contraint arguments to
734 define_expand, define_split, and define_peephole2, and in
735 define_insn_and_split if always unused.
736 * config/rs6000/darwin.md: Ditto.
737 * config/rs6000/dfp.md: Ditto.
738 * config/rs6000/rs6000.md: Ditto.
739 * config/rs6000/sync.md: Ditto.
740 * config/rs6000/vector.md: Ditto.
741 * config/rs6000/vsx.md: Ditto.
743 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
745 * config/rs6000/altivec.md: Write output control strings as braced
746 blocks instead of double-quoted strings.
747 * config/rs6000/darwin.md: Ditto.
748 * config/rs6000/rs6000.md: Ditto.
749 * config/rs6000/vector.md: Ditto.
750 * config/rs6000/vsx.md: Ditto.
752 2018-02-21 Jason Merrill <jason@redhat.com>
754 PR c++/84314 - ICE with templates and fastcall attribute.
755 * attribs.c (build_type_attribute_qual_variant): Remove assert.
757 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
759 * ipa-cp.c (determine_versionability): Fix comment typos.
761 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
764 * ipa-cp.c (determine_versionability): Do not version functions caling
767 2018-02-21 Martin Liska <mliska@suse.cz>
770 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
771 Add "native" as a possible value.
772 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
773 the macro when native cpu detection is available.
775 2018-02-21 Martin Liska <mliska@suse.cz>
778 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
779 Add "native" as a possible value.
780 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
781 when native cpu detection is available.
783 2018-02-21 Jakub Jelinek <jakub@redhat.com>
784 Martin Sebor <msebor@redhat.com>
786 PR tree-optimization/84478
787 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
789 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
790 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
791 support which is conservatively correct, for 2 only stay conservative
792 for maxlen. Formatting and comment capitalization fixes. Add STRICT
793 argument to the 2 argument get_range_strlen, adjust 6 arg
794 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
796 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
797 (gimple_fold_builtin_strlen): Pass true as last argument to
800 2018-02-20 Martin Sebor <msebor@redhat.com>
803 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
804 (builtin_memref::set_base_and_offset): Same. Handle inner references.
805 (builtin_memref::builtin_memref): Factor out parts into
806 set_base_and_offset and call it.
808 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
811 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
812 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
813 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
814 search at the associated MODE_INT.
816 2018-02-20 Jeff Law <law@redhat.com>
819 PR tree-optimization/81592
821 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
822 for range data rather than using global data.
823 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
824 range data rather than using global data.
825 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
826 pass it to children as needed.
827 (struct directive::fmtresult): Similarly.
828 (struct directive::set_width): Similarly.
829 (struct directive::set_precision): Similarly.
830 (format_integer, format_directive, parse_directive): Similarly.
831 (format_none): Accept unnamed vr_values parameter.
832 (format_percent, format_floating, format_character): Similarly.
833 (format_string, format_plain): Similarly.
834 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
835 the EVRP range analyzer for range data rather than using global data.
836 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
837 gimple-ssa-evrp-analyze.h
838 (class sprintf_dom_walker): Add after_dom_children member function.
839 Add evrp_range_analyzer member.
840 (sprintf_dom_walker::before_dom_children): Call into the EVRP
841 range analyzer as needed.
842 (sprintf_dom_walker::after_dom_children): New member function.
843 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
845 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
846 (evrp_range_analyzer::pop_to_marker): Likewise.
848 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
850 PR tree-optimization/84419
851 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
852 with the required type if its current type is compatible but
855 2018-02-20 Jakub Jelinek <jakub@redhat.com>
858 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
861 2018-02-20 Martin Liska <mliska@suse.cz>
864 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
865 possible values if we don't have a hint.
867 2018-02-20 Martin Liska <mliska@suse.cz>
871 * final.c (shorten_branches): Build align_tab array with one
873 * opts.c (finish_options): Add alignment option limit check.
874 (MAX_CODE_ALIGN): Likewise.
875 (MAX_CODE_ALIGN_VALUE): Likewise.
876 * doc/invoke.texi: Document maximum allowed option value for
877 all -falign-* options.
879 2018-02-19 Jakub Jelinek <jakub@redhat.com>
882 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
883 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
884 * var-tracking.c (emit_note_insn_var_location): Remove all references
885 to NOTE_INSN_CALL_ARG_LOCATION.
886 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
887 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
888 Use copy_rtx_if_shared.
889 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
890 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
891 (dwarf2out_var_location): Remove handling of
892 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
894 * final.c (final_scan_insn): Remove all references to
895 NOTE_INSN_CALL_ARG_LOCATION.
896 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
897 before dumping final insns.
898 * except.c (emit_note_eh_region_end): Remove all references to
899 NOTE_INSN_CALL_ARG_LOCATION.
900 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
901 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
902 * config/arc/arc.c (hwloop_optimize): Likewise.
903 * config/arm/arm.c (create_fix_barrier): Likewise.
904 * config/s390/s390.c (s390_chunkify_start): Likewise.
905 * config/sh/sh.c (find_barrier): Likewise.
906 * config/i386/i386.c (rest_of_insert_endbranch,
907 ix86_seh_fixup_eh_fallthru): Likewise.
908 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
909 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
910 * config/frv/frv.c (frv_function_prologue): Likewise.
911 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
913 (note_outside_basic_block_p): Remove all references to
914 NOTE_INSN_CALL_ARG_LOCATION.
915 * gengtype.c (adjust_field_rtx_def): Likewise.
916 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
918 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
919 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
922 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
925 PR tree-optimization/84452
926 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
927 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
930 2018-02-19 Martin Liska <mliska@suse.cz>
933 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
935 2018-02-19 Martin Liska <mliska@suse.cz>
936 Richard Sandiford <richard.sandiford@linaro.org>
938 PR tree-optimization/82491
939 * gimple-fold.c (get_base_constructor): Make earlier bail out
942 2018-02-19 Carl Love <cel@us.ibm.com>
944 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
945 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
947 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
950 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
952 * config/rl78/rl78.md (movdf): New define expand.
954 2018-02-19 Martin Liska <mliska@suse.cz>
957 * doc/invoke.texi: Fix typo.
958 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
960 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
962 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
963 handle rs6000_single_float and rs6000_double_float specially for
966 2018-02-16 Jeff Law <law@redhat.com>
968 * config/rx/rx.c (add_pop_cfi_notes): New function.;
971 2018-02-16 Jakub Jelinek <jakub@redhat.com>
974 * ipa-inline.c (inline_small_functions): Fix a typo.
976 2018-02-16 Nathan Sidwell <nathan@acm.org>
978 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
980 2018-02-16 Carl Love <cel@us.ibm.com>
982 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
983 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
984 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
985 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
986 expansion to P8V_BUILTIN_VEC_FLOAT2.
988 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
990 PR rtl-optimization/70023
991 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
992 src_regno into account.
994 2018-02-16 Carl Love <cel@us.ibm.com>
996 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
997 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
998 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
999 * config/rs6000/rs6000.c: Remove case statements for
1000 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
1001 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
1002 and P9V_BUILTIN_VEC_VINSERT4B.
1003 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
1004 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
1005 * config/rs6000/vsx.md:
1006 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
1009 2018-02-16 Carl Love <cel@us.ibm.com>
1011 * config/rs6000/altivec.h: Add builtin names vec_extract4b
1013 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
1015 * config/rs6000/rs6000-c.c: Add the definitions for
1016 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
1017 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
1018 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
1019 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
1020 definition for insert4b and define insn *insert3b_internal.
1021 * doc/extend.texi: Add documentation for vec_extract4b.
1023 2018-02-16 Nathan Sidwell <nathan@acm.org>
1025 * doc/extend.texi (Backwards Compatibility): Mention friend
1026 injection. Note for-scope is deprecated.
1027 * doc/invoke.texi (-ffriend-injection): Deprecate.
1029 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
1031 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
1032 that moved to I2, also allow destinations that are a paradoxical
1033 subreg (instead of a normal reg).
1035 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
1038 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
1041 2018-02-16 Richard Biener <rguenther@suse.de>
1043 PR tree-optimization/84037
1044 PR tree-optimization/84016
1046 * config/i386/i386.c (ix86_builtin_vectorization_cost):
1047 Adjust vec_construct for the fact we need additional higher latency
1048 128bit inserts for AVX256 and AVX512 vector builds.
1049 (ix86_add_stmt_cost): Scale vector construction cost for
1052 2018-02-16 Richard Biener <rguenther@suse.de>
1054 PR tree-optimization/84417
1055 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
1056 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
1057 (non_rewritable_lvalue_p): Likewise, use poly-ints.
1059 2018-02-16 Martin Liska <mliska@suse.cz>
1062 * internal-fn.def (ASAN_CHECK): Set proper flags.
1063 (ASAN_MARK): Likewise.
1065 2018-02-16 Julia Koval <julia.koval@intel.com>
1067 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
1068 from PTA_CANNONLAKE.
1070 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1073 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
1074 Use ++iter rather than iter++ for std::list iterators.
1075 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
1076 defer deleting them until all nodes in the forest are processed. Do
1077 free even leaf nodes. Change to_process into auto_vec.
1080 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
1081 * vec.h (vec_default_construct): Use memset instead of placement new
1082 if BROKEN_VALUE_INITIALIZATION is defined.
1083 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
1084 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
1087 PR rtl-optimization/83723
1088 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
1089 * lra.c (lra_substitute_pseudo): Likewise. If true, use
1090 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
1092 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
1094 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
1096 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
1098 PR rtl-optimization/81443
1099 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
1100 from inner REGs to paradoxical SUBREGs.
1102 2018-02-16 Richard Biener <rguenther@suse.de>
1104 PR tree-optimization/84399
1105 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1106 For operands we can analyze at their definition make sure we can
1107 analyze them at each use as well.
1109 2018-02-16 Richard Biener <rguenther@suse.de>
1111 PR tree-optimization/84190
1112 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
1113 volatile accesses if the decl isn't volatile.
1115 2018-02-15 Jason Merrill <jason@redhat.com>
1117 PR c++/84314 - ICE with templates and fastcall attribute.
1118 * attribs.c (build_type_attribute_qual_variant): Don't clobber
1119 TYPE_CANONICAL on an existing type.
1121 2018-02-15 Jakub Jelinek <jakub@redhat.com>
1123 PR tree-optimization/84383
1124 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
1125 dstoff nor call operand_equal_p if dstbase is NULL.
1127 PR tree-optimization/84334
1128 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
1129 also a CONSTANT_CLASS_P, punt.
1131 2018-02-14 Jim Wilson <jimw@sifive.com>
1133 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
1134 first SMALL_OPERAND check. New local min_second_step. Move assert
1135 to where locals are set. Add TARGET_RVC support.
1136 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
1138 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
1140 * doc/invoke.texi: Correct -Wformat-overflow code sample.
1142 2018-02-14 Martin Sebor <msebor@redhat.com>
1144 PR tree-optimization/83698
1145 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
1146 arrays constrain the offset range to their bounds.
1147 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
1148 (builtin_access::overlap): Avoid setting the size of overlap if it's
1150 (maybe_diag_overlap): Also consider arrays when deciding what values
1151 of offsets to include in diagnostics.
1153 2018-02-14 Martin Sebor <msebor@redhat.com>
1156 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
1157 that correspond to the kind of a declaration.
1159 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
1162 * config/pa/pa.md: Load address of PIC label using the linkage table
1163 if the label is nonlocal.
1165 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
1167 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
1168 warning message if user requests -maltivec=be.
1169 * doc/invoke.texi: Document deprecation of -maltivec=be.
1171 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
1174 * config/rs6000/rs6000-c.c: Update definitions for
1175 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
1176 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
1178 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1181 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
1182 add _get_ssp intrinsics. Remove argument from
1183 __builtin_ia32_rdssp[d|q].
1184 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
1185 * config/i386/i386-builtin.def: Remove argument from
1186 __builtin_ia32_rdssp[d|q].
1187 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
1188 ix86_expand_special_args_builtin for _rdssp[d|q].
1189 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
1190 Clear register before usage.
1191 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
1192 Add documentation for new _get_ssp and _inc_ssp intrinsics.
1194 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
1196 PR tree-optimization/84357
1197 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
1198 operand 1 of an ARRAY_REF too.
1200 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
1203 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
1204 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1206 (set_of_reg): New struct.
1207 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
1208 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
1209 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1211 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
1212 Split into bitclr, bitset, bitinvert patterns if appropriate.
1213 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
1214 use rx_fuse_in_memory_bitop.
1215 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
1216 to named insn, correct maximum insn length.
1218 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
1221 * machmode.def: Define a complex mode for PARTIAL_INT.
1222 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
1224 * doc/rtl.texi: Document CSPImode.
1225 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
1227 (msp430_hard_regno_nregs_with_padding): Likewise.
1229 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
1232 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
1234 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
1236 PR rtl-optimization/84169
1237 * combine.c (try_combine): New variable split_i2i3. Set it to true if
1238 we generated a parallel as new i3 and we split that to new i2 and i3
1239 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
1240 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
1241 those to i2, not i1. Partially rewrite this scan code.
1243 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1246 * stor-layout.c (place_field): For variable length fields, adjust
1247 offset_align afterwards not just based on the field's alignment,
1248 but also on the size.
1251 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
1252 of exps and logs in the use_exp2 case.
1254 2018-02-13 Jeff Law <law@redhat.com>
1256 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
1259 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
1262 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
1266 * common.opt (gas-loc-support, gas-locview-support): New.
1267 (ginline-points, ginternal-reset-location-views): New.
1268 * doc/invoke.texi: Document them. Use @itemx where intended.
1269 (gvariable-location-views): Adjust.
1270 * target.def (reset_location_view): New.
1271 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
1272 (TARGET_RESET_LOCATION_VIEW): New.
1273 * doc/tm.texi: Rebuilt.
1274 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
1275 (dwarf2out_default_as_locview_support): New.
1276 (output_asm_line_debug_info): Use option variables.
1277 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
1278 (output_loc_list): Likewise.
1279 (add_high_low_attributes): Check option variables.
1280 Don't output entry view attribute in strict mode.
1281 (gen_inlined_subroutine_die): Check option variables.
1282 (dwarf2out_inline_entry): Likewise.
1283 (init_sections_and_labels): Likewise.
1284 (dwarf2out_early_finish): Likewise.
1285 (maybe_reset_location_view): New, from...
1286 (dwarf2out_var_location): ... here. Call it.
1287 * debug.h (dwarf2out_default_as_loc_support): Declare.
1288 (dwarf2out_default_as_locview_support): Declare.
1289 * hooks.c (hook_int_rtx_insn_0): New.
1290 * hooks.h (hook_int_rtx_insn_0): Declare.
1291 * toplev.c (process_options): Take -gas-loc-support and
1292 -gas-locview-support from dwarf2out. Enable
1293 -gvariable-location-views by default only with locview
1294 assembler support. Enable -ginternal-reset-location-views by
1295 default only if the target defines the corresponding hook.
1296 Enable -ginline-points by default if location views are
1297 enabled; force it disabled if statement frontiers are
1299 * tree-inline.c (expand_call_inline): Check option variables.
1300 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
1302 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
1304 PR tree-optimization/84321
1305 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
1306 handling. Also check whether the anti-range contains any values
1307 that satisfy the mask; switch to a VR_RANGE if not.
1309 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
1312 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
1314 2018-02-13 Martin Jambor <mjambor@suse.cz>
1317 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
1318 of call statements, also set location of a load to a temporary.
1320 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
1322 * config/rl78/rl78.c (add_vector_labels): New function.
1323 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
1324 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
1325 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
1326 which checks that no arguments are passed.
1327 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
1328 * doc/extend.texi: Documentation for the new attribute.
1330 2018-02-13 Andreas Schwab <schwab@suse.de>
1332 * config/riscv/linux.h (CPP_SPEC): Define.
1334 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1337 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
1338 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
1339 OPTION_MASK_ISA_AES as first argument to def_builtin_const
1340 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
1341 instead of OPTION_MASK_ISA_PCLMUL as first argument to
1342 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
1343 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
1344 temporarily for AES and PCLMUL builtins.
1346 PR tree-optimization/84339
1347 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
1348 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
1352 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
1353 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
1354 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
1356 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
1358 * omp-simd-clone.h: New file.
1359 * omp-simd-clone.c: Include omp-simd-clone.h.
1360 (expand_simd_clones): No longer static.
1361 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
1362 cgraph.h and omp-simd-clone.h.
1363 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
1364 (vect_recog_widen_shift_pattern): Formatting fix.
1365 (vect_pattern_recog_1): Don't check optab for calls.
1368 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
1369 operands[2] into a REG before using gen_lowpart on it.
1371 2018-02-12 Jeff Law <law@redhat.com>
1374 * config/sh/sh.c (find_barrier): Consider a sibling call
1377 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
1378 successfully back substituting a reg.
1380 2018-02-12 Richard Biener <rguenther@suse.de>
1382 PR tree-optimization/84037
1383 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
1384 parameter, move visited init to caller.
1385 (vect_slp_analyze_operations): Separate cost from validity
1386 check, initialize visited once for all instances.
1387 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
1389 * tree-vect-stmts.c (vect_model_simple_cost): Make early
1391 (vect_model_promotion_demotion_cost): Likewise.
1392 (vectorizable_bswap): Guard cost modeling with !slp_node
1393 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
1395 (vectorizable_call): Likewise.
1396 (vectorizable_conversion): Likewise.
1397 (vectorizable_assignment): Likewise.
1398 (vectorizable_shift): Likewise.
1399 (vectorizable_operation): Likewise.
1400 (vectorizable_store): Likewise.
1401 (vectorizable_load): Likewise.
1402 (vectorizable_condition): Likewise.
1403 (vectorizable_comparison): Likewise.
1405 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
1408 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
1409 (ASAN_MARK): Fix fnspec to account for return value, change pointer
1410 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
1412 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1415 * params.def (inline-min-speedup): Increase from 8 to 15.
1416 (max-inline-insns-auto): Decrease from 40 to 30.
1417 * ipa-split.c (consider_split): Add some buffer for function to
1418 be considered inlining candidate.
1419 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
1422 2018-02-12 Richard Biener <rguenther@suse.de>
1424 PR tree-optimization/84037
1425 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
1426 matched stmts if we cannot swap the non-matched ones.
1428 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1430 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
1431 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
1432 _mm_maskz_scalef_round_ss): New intrinsics.
1433 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
1434 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
1435 __builtin_ia32_scalefss_round): Remove.
1436 (__builtin_ia32_scalefsd_mask_round,
1437 __builtin_ia32_scalefss_mask_round): New intrinsics.
1438 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
1439 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
1440 ((match_operand:VF_128 2 "<round_nimm_predicate>"
1441 "<round_constraint>")): Changed to ...
1442 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
1443 "<round_scalar_constraint>")): ... this.
1444 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
1445 %0, %1, %2<round_op3>}"): Changed to ...
1446 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
1447 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
1448 %2<round_scalar_mask_op3>}"): ... this.
1449 * config/i386/subst.md (round_scalar_nimm_predicate): New.
1451 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1453 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
1454 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
1455 (_mm_maskz_sqrt_round_ss): New intrinsics.
1456 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
1457 (__builtin_ia32_sqrtsd_mask_round)
1458 (__builtin_ia32_sqrtss_mask_round): New builtins.
1459 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
1460 (__builtin_ia32_sqrtss_round): Remove.
1461 (__builtin_ia32_sqrtsd_mask_round)
1462 (__builtin_ia32_sqrtss_mask_round): New builtins.
1463 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
1464 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
1465 ((match_operand:VF_128 1 "vector_operand"
1466 "xBm,<round_constraint>")): Changed to ...
1467 ((match_operand:VF_128 1 "vector_operand"
1468 "xBm,<round_scalar_constraint>")): ... this.
1469 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
1470 %0, %2, %<iptr>1<round_op3>}): Changed to ...
1471 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
1472 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
1473 %<iptr>1<round_scalar_mask_op3>}): ... this.
1474 ((set_attr "prefix" "<round_prefix>")): Changed to ...
1475 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
1477 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
1480 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
1481 Cast vec_cmpeq result to correct type.
1482 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
1483 Cast vec_cmpgt result to correct type.
1485 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
1487 * final.c (final_scan_insn_1): Renamed from...
1488 (final_scan_insn): ... this. New wrapper, to recover
1489 seen from the outermost call in recursive ones.
1490 * config/sparc/sparc.c (output_return): Drop seen from call.
1491 (output_sibcall): Likewise.
1492 * config/visium/visium.c (output_branch): Likewise.
1494 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
1496 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
1499 2018-02-10 Alan Modra <amodra@gmail.com>
1502 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
1503 Specify LR as an input.
1505 2018-02-10 Jakub Jelinek <jakub@redhat.com>
1508 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
1509 remove_member_access_dummy_vars): New functions.
1510 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
1511 lower_omp_1, execute_lower_omp): Use them.
1513 PR rtl-optimization/84308
1514 * shrink-wrap.c (spread_components): Release todo vector.
1516 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
1518 PR rtl-optimization/57193
1519 * ira-color.c (struct allocno_color_data): Add member
1520 conflict_allocno_hard_prefs.
1521 (update_conflict_allocno_hard_prefs): New.
1522 (bucket_allocno_compare_func): Add a preference based on
1523 conflict_allocno_hard_prefs.
1524 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
1525 (color_allocnos): Remove a dead code. Initiate
1526 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
1528 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1531 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
1532 constraint from =wa to wa. Avoid a subreg on the output operand,
1533 instead use a pseudo and subreg it in a move.
1534 (p9_xxbrd_<mode>): Changed to ...
1535 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
1536 (p9_xxbrd_v2df): New expander.
1537 (p9_xxbrw_<mode>): Changed to ...
1538 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
1539 (p9_xxbrw_v4sf): New expander.
1541 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1543 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
1545 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
1548 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
1549 multiply in 32-bit mode.
1550 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
1551 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
1554 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1556 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
1557 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
1558 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
1559 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
1561 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
1564 * dwarf2out.c (is_trivial_indirect_ref): New function.
1565 (dwarf2out_late_global_decl): Do not generate a location
1566 attribute for variables that have a non-trivial DECL_VALUE_EXPR
1567 and that are not defined in the current unit.
1569 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1571 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
1572 instead of a libcall for UNORDERED.
1574 2018-02-09 Tamar Christina <tamar.christina@arm.com>
1577 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
1578 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
1580 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1583 * config/s390/s390.c (s390_set_current_function): Invoke
1584 s390_indirect_branch_settings also if fndecl didn't change.
1586 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1588 * config/rs6000/rs6000.md (blockage): Set length to zero.
1590 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1592 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
1594 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1597 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
1598 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
1602 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
1603 PARALLEL incoming that failed vt_get_decl_and_offset check.
1606 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
1607 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
1609 (get_variable_section): For decls in named .bss* sections pass true as
1610 second argument to bss_initializer_p.
1612 2018-02-09 Marek Polacek <polacek@redhat.com>
1613 Jakub Jelinek <jakub@redhat.com>
1616 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
1617 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
1618 Sync some changes from cxx_fold_indirect_ref.
1620 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1622 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
1624 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
1625 (BLOCK_INLINE_ENTRY_LABEL): New.
1626 (dwarf2out_var_location): Disregard inline entry markers.
1627 (inline_entry_data): New struct.
1628 (inline_entry_data_hasher): New hashtable type.
1629 (inline_entry_data_hasher::hash): New.
1630 (inline_entry_data_hasher::equal): New.
1631 (inline_entry_data_table): New variable.
1632 (add_high_low_attributes): Add DW_AT_entry_pc and
1633 DW_AT_GNU_entry_view attributes if a pending entry is found
1634 in inline_entry_data_table. Add old entry_pc attribute only
1635 if debug nonbinding markers are disabled.
1636 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
1637 markers are enabled.
1638 (block_within_block_p, dwarf2out_inline_entry): New.
1639 (dwarf2out_finish): Check that no entries remained in
1640 inline_entry_data_table.
1641 * final.c (reemit_insn_block_notes): Handle inline entry notes.
1642 (final_scan_insn, notice_source_line): Likewise.
1643 (rest_of_clean_state): Skip inline entry markers.
1644 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
1646 * gimple.c (gimple_build_debug_inline_entry): New.
1647 * gimple.h (enum gimple_debug_subcode): Add
1648 GIMPLE_DEBUG_INLINE_ENTRY.
1649 (gimple_build_debug_inline_entry): Declare.
1650 (gimple_debug_inline_entry_p): New.
1651 (gimple_debug_nonbind_marker_p): Adjust.
1652 * insn-notes.def (INLINE_ENTRY): New.
1653 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
1654 inline entry marker notes.
1655 (print_insn): Likewise.
1656 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
1657 (INSN_DEBUG_MARKER_KIND): Likewise.
1658 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
1659 * tree-inline.c (expand_call_inline): Build and insert
1660 debug_inline_entry stmt.
1661 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
1662 inline entry blocks early, if nonbind markers are enabled.
1663 (dump_scope_block): Dump fragment info.
1664 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
1665 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
1666 (gimple_build_debug_inline_entry): New.
1667 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
1668 Enable/disable inline entry points too.
1669 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
1670 (DEBUG_INSN): Describe inline entry markers.
1672 * common.opt (gvariable-location-views): New.
1673 (gvariable-location-views=incompat5): New.
1674 * config.in: Rebuilt.
1675 * configure: Rebuilt.
1676 * configure.ac: Test assembler for view support.
1677 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
1678 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
1679 * dwarf2out.c (var_loc_view): New typedef.
1680 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
1681 (dwarf2out_locviews_in_attribute): New.
1682 (dwarf2out_locviews_in_loclist): New.
1683 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
1684 (enum dw_line_info_opcode): Add LI_adv_address.
1685 (struct dw_line_info_table): Add view.
1686 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
1687 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
1688 (zero_view_p): New variable.
1689 (ZERO_VIEW_P): New macro.
1690 (output_asm_line_debug_info): New.
1691 (struct var_loc_node): Add view.
1692 (add_AT_view_list, AT_loc_list): New.
1693 (add_var_loc_to_decl): Add view param. Test it against last.
1694 (new_loc_list): Add view params. Record them.
1695 (AT_loc_list_ptr): Handle loc and view lists.
1696 (view_list_to_loc_list_val_node): New.
1697 (print_dw_val): Handle dw_val_class_view_list.
1698 (size_of_die): Likewise.
1699 (value_format): Likewise.
1700 (loc_list_has_views): New.
1701 (gen_llsym): Set vl_symbol too.
1702 (maybe_gen_llsym, skip_loc_list_entry): New.
1703 (dwarf2out_maybe_output_loclist_view_pair): New.
1704 (output_loc_list): Output view list or entries too.
1705 (output_view_list_offset): New.
1706 (output_die): Handle dw_val_class_view_list.
1707 (output_dwarf_version): New.
1708 (output_compilation_unit_header): Use it.
1709 (output_skeleton_debug_sections): Likewise.
1710 (output_rnglists, output_line_info): Likewise.
1711 (output_pubnames, output_aranges): Update version comments.
1712 (output_one_line_info_table): Output view numbers in asm comments.
1713 (dw_loc_list): Determine current endview, pass it to new_loc_list.
1714 Call maybe_gen_llsym.
1715 (loc_list_from_tree_1): Adjust.
1716 (add_AT_location_description): Create view list attribute if
1717 needed, check it's absent otherwise.
1718 (convert_cfa_to_fb_loc_list): Adjust.
1719 (maybe_emit_file): Call output_asm_line_debug_info for test.
1720 (dwarf2out_var_location): Reset views as needed. Precompute
1721 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
1722 attribute. Set view.
1723 (new_line_info_table): Reset next view.
1724 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
1725 (dwarf2out_source_line): Likewise. Output view resets and labels to
1726 the assembler, or select appropriate line info opcodes.
1727 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
1728 (optimize_string_length): Catch it. Adjust.
1729 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
1730 dw_val_class_view_list, and remove it if no longer needed.
1731 (hash_loc_list): Hash view numbers.
1732 (loc_list_hasher::equal): Compare them.
1733 (optimize_location_lists): Check whether a view list symbol is
1734 needed, and whether the locview attribute is present, and
1735 whether they match. Remove the locview attribute if no longer
1737 (index_location_lists): Call skip_loc_list_entry for test.
1738 (dwarf2out_finish): Call output_asm_line_debug_info for test.
1739 Use output_dwarf_version.
1740 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
1741 (struct dw_val_node): Add val_view_list.
1742 * final.c (SEEN_NEXT_VIEW): New.
1743 (set_next_view_needed): New.
1744 (clear_next_view_needed): New.
1745 (maybe_output_next_view): New.
1746 (final_start_function): Rename to...
1747 (final_start_function_1): ... this. Take pointer to FIRST,
1748 add SEEN parameter. Emit param bindings in the initial view.
1749 (final_start_function): Reintroduce SEEN-less interface.
1750 (final): Rename to...
1751 (final_1): ... this. Take SEEN parameter. Output final pending
1752 next view at the end.
1753 (final): Reintroduce seen-less interface.
1754 (final_scan_insn): Output pending next view before switching
1755 sections or ending a block. Mark the next view as needed when
1756 outputting variable locations. Notify debug backend of section
1757 changes, and of location view changes.
1758 (rest_of_handle_final): Adjust.
1759 * toplev.c (process_options): Autodetect value for debug variable
1760 location views option. Warn on incompat5 without -gdwarf-5.
1761 * doc/invoke.texi (gvariable-location-views): New.
1762 (gvariable-location-views=incompat5): New.
1763 (gno-variable-location-views): New.
1765 2018-02-08 David Malcolm <dmalcolm@redhat.com>
1767 PR tree-optimization/84136
1768 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
1769 that the result of find_edge is non-NULL.
1771 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1774 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
1775 storing integer register in SImode. Fix cost of 256 and 512
1776 byte aligned SSE register store.
1778 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1780 * config/i386/i386.c (ix86_multiplication_cost): Fix
1781 multiplication cost for TARGET_AVX512DQ.
1783 2018-02-08 Marek Polacek <polacek@redhat.com>
1785 PR tree-optimization/84238
1786 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
1789 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1791 PR tree-optimization/84265
1792 * tree-vect-stmts.c (vectorizable_store): Don't treat
1793 VMAT_CONTIGUOUS accesses as grouped.
1794 (vectorizable_load): Likewise.
1796 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1798 PR tree-optimization/81635
1799 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
1800 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
1801 (test_round_for_mask): New functions.
1802 (wide_int_cc_tests): Call test_round_for_mask.
1803 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
1804 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
1805 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
1806 range returned by get_range_info.
1808 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1811 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
1812 * symtab.c: Include builtins.h
1813 (symtab_node::output_to_lto_symbol_table_p): Move here
1814 from lto-streamer-out.c:output_symbol_p.
1815 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
1816 (output_symbol_p): Move all logic to symtab.c
1817 (produce_symtab): Update.
1819 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1821 * config/s390/s390-opts.h (enum indirect_branch): Define.
1822 * config/s390/s390-protos.h (s390_return_addr_from_memory)
1823 (s390_indirect_branch_via_thunk)
1824 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
1825 (enum s390_indirect_branch_type): Define.
1826 * config/s390/s390.c (struct s390_frame_layout, struct
1827 machine_function): Remove.
1828 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
1829 (indirect_branch_table_label_no, indirect_branch_table_name):
1831 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
1832 (enum s390_indirect_branch_option): Define.
1833 (s390_return_addr_from_memory): New function.
1834 (s390_handle_string_attribute): New function.
1835 (s390_attribute_table): Add new attribute handler.
1836 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
1837 (s390_indirect_branch_via_thunk): New function.
1838 (s390_indirect_branch_via_inline_thunk): New function.
1839 (s390_function_ok_for_sibcall): When jumping via thunk disallow
1840 sibling call optimization for non z10 compiles.
1841 (s390_emit_call): Force indirect branch target to be a single
1842 register. Add r1 clobber for non-z10 compiles.
1843 (s390_emit_epilogue): Emit return jump via return_use expander.
1844 (s390_reorg): Handle JUMP_INSNs as execute targets.
1845 (s390_option_override_internal): Perform validity checks for the
1846 new command line options.
1847 (s390_indirect_branch_attrvalue): New function.
1848 (s390_indirect_branch_settings): New function.
1849 (s390_set_current_function): Invoke s390_indirect_branch_settings.
1850 (s390_output_indirect_thunk_function): New function.
1851 (s390_code_end): Implement target hook.
1852 (s390_case_values_threshold): Implement target hook.
1853 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
1855 * config/s390/s390.h (struct s390_frame_layout)
1856 (struct machine_function): Move here from s390.c.
1857 (TARGET_INDIRECT_BRANCH_NOBP_RET)
1858 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
1859 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
1860 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
1861 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
1862 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
1863 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
1864 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
1865 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
1866 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
1867 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
1868 (mnemonic attribute): Add values which aren't recognized
1870 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
1871 pattern for branch conversion. Fix mnemonic attribute.
1872 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
1873 indirect branch via thunk if requested.
1874 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
1875 ("*indirect_jump"): Disable for branch conversion using out of
1877 ("indirect_jump_via_thunk<mode>_z10")
1878 ("indirect_jump_via_thunk<mode>")
1879 ("indirect_jump_via_inlinethunk<mode>_z10")
1880 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
1881 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
1882 ("casesi_jump_via_inlinethunk<mode>_z10")
1883 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
1884 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
1885 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
1886 ("*indirect2_jump"): Disable for branch conversion.
1887 ("casesi_jump"): Turn into expander and expand patterns for branch
1889 ("return_use"): New expander.
1890 ("*return"): Emit return via thunk and rename it to ...
1891 ("*return<mode>"): ... this one.
1892 * config/s390/s390.opt: Add new options and and enum for the
1895 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1897 * lra-constraints.c (match_reload): Unconditionally use
1898 gen_lowpart_SUBREG, rather than selecting between that
1899 and equivalent gen_rtx_SUBREG code.
1901 2018-02-08 Richard Biener <rguenther@suse.de>
1903 PR tree-optimization/84233
1904 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
1905 changed flag instead of boguously re-using phi_inserted.
1907 2018-02-08 Martin Jambor <mjambor@suse.cz>
1909 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
1910 static local variables.
1912 2018-02-08 Richard Biener <rguenther@suse.de>
1914 PR tree-optimization/84278
1915 * tree-vect-stmts.c (vectorizable_store): When looking for
1916 smaller vector types to perform grouped strided loads/stores
1917 make sure the mode is supported by the target.
1918 (vectorizable_load): Likewise.
1920 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
1922 * config/aarch64/aarch64.c (aarch64_components_for_bb):
1923 Increase LDP/STP opportunities by adding adjacent callee-saves.
1925 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
1927 PR rtl-optimization/84068
1928 PR rtl-optimization/83459
1929 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
1931 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
1933 PR tree-optimization/84224
1934 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
1935 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
1938 2018-02-07 Iain Sandoe <iain@codesourcery.com>
1941 * config/rs6000/altivec.md (*restore_world): Remove LR use.
1942 * config/rs6000/predicates.md (restore_world_operation): Adjust op
1943 count, remove one USE.
1945 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
1947 * doc/install.texi (Configuration): Document the
1948 --with-long-double-format={ibm,ieee} PowerPC configuration
1952 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
1953 Convert from define_expand to be define_insn_and_split. Rework
1954 float/double/_Float128 conversions to QI/HI/SImode to work with
1955 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
1956 conversions to QI/HImode types did a store and then a load to
1957 truncate the value. For conversions to VSX registers, don't split
1958 the insn, instead emit the code directly. Use the code iterator
1959 any_fix to combine signed and unsigned conversions.
1960 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
1961 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
1962 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
1963 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
1964 (fix_<mode>di2_hw): Likewise.
1965 (fixuns_<mode>di2_hw): Likewise.
1966 (fix_<mode>si2_hw): Likewise.
1967 (fixuns_<mode>si2_hw): Likewise.
1968 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
1969 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
1970 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
1971 fix<uns>_trunc<SFDF:mode>si2_p8.
1972 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
1974 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
1975 (fix<uns>_<mode>_mem): Likewise.
1976 (fctiw<u>z_<mode>_mem): Likewise.
1977 (fix<uns>_<mode>_mem): Likewise.
1978 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
1979 the register allocator from doing a direct move to the GPRs to do
1980 a store, and instead use the ISA 3.0 store byte/half-word from
1981 vector register instruction. For IEEE 128-bit floating point,
1982 also optimize stores of 32-bit ints.
1983 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
1985 2018-02-07 Alan Hayward <alan.hayward@arm.com>
1987 * genextract.c (push_pathstr_operand): New function to support
1989 (walk_rtx): Call push_pathstr_operand.
1990 (print_path): Support [a-zA-Z].
1992 2018-02-07 Richard Biener <rguenther@suse.de>
1994 PR tree-optimization/84037
1995 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
1996 (cse_and_gimplify_to_preheader): Declare.
1997 (vect_get_place_in_interleaving_chain): Likewise.
1998 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
2000 (_loop_vec_info::~_loop_vec_info): Delete it.
2001 (cse_and_gimplify_to_preheader): New function.
2002 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
2003 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
2004 (vectorizable_load): Likewise. For grouped stores always base
2005 the IV on the first element.
2006 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
2007 condition before gimplifying.
2009 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2011 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
2012 *DIV_EXPR and *MOD_EXPR.
2014 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
2017 * config/i386/i386.c (ix86_option_override_internal): Mask out
2018 the CF_SET bit when checking -fcf-protection.
2020 2018-02-07 Tom de Vries <tom@codesourcery.com>
2023 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
2026 2018-02-07 Richard Biener <rguenther@suse.de>
2028 PR tree-optimization/84204
2029 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
2032 PR tree-optimization/84205
2033 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
2034 special-case isl_ast_op_zdiv_r.
2036 PR tree-optimization/84223
2037 * graphite-scop-detection.c (gather_bbs::before_dom_children):
2038 Only add conditions from within the region.
2039 (gather_bbs::after_dom_children): Adjust.
2041 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
2044 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
2045 * config/avr/avr.md: Only post-reload split REG-REG moves if
2046 either register is REGERAL_REG_P.
2048 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2050 PR tree-optimization/84235
2051 * tree-ssa-scopedtables.c
2052 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
2053 if the subtraction is performed in floating point type where NaNs are
2054 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
2055 build 1. Formatting fix.
2057 2018-02-06 Jakub Jelinek <jakub@redhat.com>
2060 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
2061 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
2062 and skip it regardless of bb boundaries. Use CALL_P macro,
2063 don't test INSN_P (insn) together with CALL_P or JUMP_P check
2064 unnecessarily, formatting fix.
2066 2018-02-06 Michael Collison <michael.collison@arm.com>
2068 * config/arm/thumb2.md:
2069 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
2070 (*thumb_mov_notscc): Ditto.
2072 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
2075 * config/rs6000/rs6000.md (su code attribute): Use "u" for
2076 unsigned_fix, not "s".
2078 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2080 * configure.ac (gcc_fn_eh_frame_ro): New function.
2081 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
2082 correct .eh_frame permissions.
2083 * configure: Regenerate.
2085 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
2087 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
2090 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2092 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2093 Display warning message for -mno-speculate-indirect-jumps.
2095 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
2097 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
2099 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
2101 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
2103 PR tree-optimization/84225
2104 * tree-eh.c (find_trapping_overflow): Only call
2105 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
2107 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2110 * config/i386/i386.c: Reimplement the check of possible options
2111 -mibt/-mshstk conbination. Change error messages.
2112 * doc/invoke.texi: Fix a typo: remove extra '='.
2114 2018-02-06 Marek Polacek <polacek@redhat.com>
2116 PR tree-optimization/84228
2117 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
2119 2018-02-06 Tamar Christina <tamar.christina@arm.com>
2122 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
2123 emitted arch directives.
2124 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
2125 __ARM_FEATURE_COPROC before changing architectures.
2127 2018-02-06 Richard Biener <rguenther@suse.de>
2129 * config/i386/i386.c (print_reg): Fix typo.
2130 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
2132 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
2134 * configure: Regenerate.
2136 2018-02-05 Martin Sebor <msebor@redhat.com>
2138 PR tree-optimization/83369
2139 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
2142 2018-02-05 Martin Liska <mliska@suse.cz>
2144 * doc/invoke.texi: Cherry-pick upstream r323995.
2146 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
2148 * ira.c (ira_init_register_move_cost): Adjust comment.
2150 2018-02-05 Martin Liska <mliska@suse.cz>
2152 PR gcov-profile/84137
2153 * doc/gcov.texi: Fix typo in documentation.
2155 2018-02-05 Martin Liska <mliska@suse.cz>
2157 PR gcov-profile/83879
2158 * doc/gcov.texi: Document necessity of --dynamic-list-data when
2159 using dlopen functionality.
2161 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
2163 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
2164 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
2165 _mm_maskz_range_ss, _mm_mask_range_round_ss,
2166 _mm_maskz_range_round_ss): New intrinsics.
2167 (__builtin_ia32_rangesd128_round)
2168 (__builtin_ia32_rangess128_round): Remove.
2169 (__builtin_ia32_rangesd128_mask_round,
2170 __builtin_ia32_rangess128_mask_round): New builtins.
2171 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
2172 __builtin_ia32_rangess128_round): Remove.
2173 (__builtin_ia32_rangesd128_mask_round,
2174 __builtin_ia32_rangess128_mask_round): New builtins.
2175 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
2176 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
2177 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
2178 "<round_saeonly_constraint>")): Changed to ...
2179 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
2180 "<round_saeonly_scalar_constraint>")): ... this.
2181 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
2182 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
2183 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
2184 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
2185 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
2187 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
2189 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
2191 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
2192 Remove all values except native, 8540 and 8548.
2194 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
2196 * config/i386/i386.c (ix86_output_function_return): Pass
2197 INVALID_REGNUM, instead of -1, as invalid register number to
2198 indirect_thunk_name and output_indirect_thunk.
2200 2018-02-02 Julia Koval <julia.koval@intel.com>
2202 * config.gcc: Add -march=icelake.
2203 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
2204 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
2205 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
2206 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
2207 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
2208 (processor_target_table): Add icelake.
2209 (ix86_option_override_internal): Handle new PTAs.
2210 (get_builtin_code_for_version): Handle icelake.
2211 (M_INTEL_COREI7_ICELAKE): New.
2212 (fold_builtin_cpu): Handle icelake.
2213 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
2214 * doc/invoke.texi: Add -march=icelake.
2216 2018-02-02 Julia Koval <julia.koval@intel.com>
2218 * config/i386/i386.c (ix86_option_override_internal): Change flags type
2219 to wide_int_bitmask.
2220 * wide-int-bitmask.h: New.
2222 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2225 * config/i386/i386.md: Replace Pmode with word_mode in
2226 builtin_setjmp_setup and builtin_longjmp to support x32.
2228 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
2232 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
2234 (rs6000_supported_cpu_names): New static variable.
2235 (linux_cpu_translation_table): Likewise.
2236 (elf_platform) <cpu>: Define new static variable and use it.
2237 Translate kernel AT_PLATFORM name to canonical name if needed.
2238 Error if platform name is unknown.
2240 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
2243 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
2245 2018-02-01 Jeff Law <law@redhat.com>
2248 * config/i386/i386.c (release_scratch_register_on_entry): Add new
2249 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
2250 the scratch if RELEASE_VIA_POP is false.
2251 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
2252 If we have to save a temporary register, decrement SIZE appropriately.
2253 Pass new arguments to release_scratch_register_on_entry.
2254 (ix86_adjust_stack_and_probe): Likewise.
2255 (ix86_emit_probe_stack_range): Pass new arguments to
2256 release_scratch_register_on_entry.
2258 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
2260 PR rtl-optimization/84157
2261 * combine.c (change_zero_ext): Use REG_P predicate in
2262 front of HARD_REGISTER_P predicate.
2264 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
2266 * config/avr/avr.c (avr_option_override): Move disabling of
2267 -fdelete-null-pointer-checks to...
2268 * common/config/avr/avr-common.c (avr_option_optimization_table):
2271 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2273 PR tree-optimization/81635
2274 * tree-data-ref.c (split_constant_offset_1): For types that
2275 wrap on overflow, try to use range info to prove that wrapping
2278 2018-02-01 Renlin Li <renlin.li@arm.com>
2281 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
2283 (aarch64_register_move_cost): Likewise.
2284 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
2286 (REG_CLASS_NAMES): Likewise.
2287 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
2288 TAILCALL_ADDR_REGS. Remove IP registers.
2289 * config/aarch64/aarch64.md (Ucs): Update register constraint.
2291 2018-02-01 Richard Biener <rguenther@suse.de>
2293 * domwalk.h (dom_walker::dom_walker): Add additional constructor
2294 for specifying RPO order and allow NULL for that.
2295 * domwalk.c (dom_walker::dom_walker): Likewise.
2296 (dom_walker::walk): Handle NULL RPO order.
2297 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
2299 (rewrite_update_dom_walker): Likewise.
2300 (mark_def_dom_walker): Likewise.
2302 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2304 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
2305 (aarch64_maybe_expand_sve_subreg_move): Declare.
2306 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
2307 * config/aarch64/predicates.md (aarch64_any_register_operand): New
2309 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
2310 that are semantically a reverse operation.
2311 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
2312 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
2313 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
2315 (aarch64_can_change_mode_class): For big-endian, forbid changes
2316 between two SVE modes if they have different element sizes.
2318 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2320 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
2321 the TImode handling for big-endian targets.
2323 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2325 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
2326 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
2328 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
2329 Remove BSWAP handing for big-endian targets and use the form of
2330 LD1RQ appropariate for the mode.
2332 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2334 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
2335 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
2338 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2341 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
2342 check for operands that need to go through aarch64_sve_reload_be.
2344 2018-02-01 Jakub Jelinek <jakub@redhat.com>
2346 PR tree-optimization/81661
2347 PR tree-optimization/84117
2348 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
2349 * tree-eh.c: Include gimplify.h.
2350 (find_trapping_overflow, replace_trapping_overflow,
2351 rewrite_to_non_trapping_overflow): New functions.
2352 * tree-vect-loop.c: Include tree-eh.h.
2353 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
2354 * tree-data-ref.c: Include tree-eh.h.
2355 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
2357 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
2359 PR rtl-optimization/84123
2360 * combine.c (change_zero_ext): Check if hard register satisfies
2361 can_change_dest_mode before calling gen_lowpart_SUBREG.
2363 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
2366 * ira.c (ira_init_register_move_cost): Remove assert.
2368 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2370 PR rtl-optimization/84071
2371 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
2372 * doc/tm.texi: Regenerate.
2374 2018-01-31 Richard Biener <rguenther@suse.de>
2376 PR tree-optimization/84132
2377 * tree-data-ref.c (analyze_miv_subscript): Properly
2378 check whether evolution_function_is_affine_multivariate_p
2379 before calling gcd_of_steps_may_divide_p.
2381 2018-01-31 Julia Koval <julia.koval@intel.com>
2384 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
2385 * config/i386/i386.md (rdpid_rex64) New.
2386 (rdpid): Make 32bit only.
2388 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
2391 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
2392 an IDENTIFIER_NODE for FUNCTION_TYPE's.
2394 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2397 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
2399 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
2401 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2403 PR rtl-optimization/84071
2404 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
2405 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
2407 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2409 * config/arc/arc.c (arc_handle_aux_attribute): New function.
2410 (arc_attribute_table): Add 'aux' attribute.
2411 (arc_in_small_data_p): Consider aux like variables.
2412 (arc_is_aux_reg_p): New function.
2413 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
2414 (arc_get_aux_arg): New function.
2415 (prepare_move_operands): Handle aux-register access.
2416 (arc_handle_aux_attribute): New function.
2417 * doc/extend.texi (ARC Variable attributes): Add subsection.
2419 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2421 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
2422 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
2423 (arc_attribute_table): Add 'uncached' attribute.
2424 (arc_print_operand): Print '.di' flag for uncached memory
2426 (arc_in_small_data_p): Do not consider for small data the uncached
2428 (arc_is_uncached_mem_p): New function.
2429 * config/arc/predicates.md (compact_store_memory_operand): Check
2430 for uncached memory accesses.
2431 (nonvol_nonimm_operand): Likewise.
2432 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
2434 2018-01-31 Jakub Jelinek <jakub@redhat.com>
2437 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
2438 falign-loops=): Add Optimization flag.
2440 2018-01-30 Jeff Law <law@redhat.com>
2443 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
2444 INT_REGISTERS_SAVED. Check it prior to calling
2445 get_scratch_register_on_entry.
2446 (ix86_adjust_stack_and_probe): Similarly.
2447 (ix86_emit_probe_stack_range): Similarly.
2448 (ix86_expand_prologue): Corresponding changes.
2450 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2453 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
2454 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
2456 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
2459 * lra-constraints.c (curr_insn_transform): Process AND in the
2462 2018-01-30 Jakub Jelinek <jakub@redhat.com>
2464 PR rtl-optimization/83986
2465 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
2466 dependence against last_pending_memory_flush in addition to
2469 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
2471 PR tree-optimization/81611
2472 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
2475 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2478 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
2481 2018-01-30 Richard Biener <rguenther@suse.de>
2482 Jakub Jelinek <jakub@redhat.com>
2484 PR tree-optimization/84111
2485 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
2486 inner loops added during recursion, as they don't have up-to-date
2489 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2492 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
2493 (can_inline_edge_by_limits_p): ... here.
2494 (can_early_inline_edge_p, check_callers,
2495 update_caller_keys, update_callee_keys, recursive_inlining,
2496 add_new_edges_to_heap, speculation_useful_p,
2497 inline_small_functions,
2498 inline_small_functions, flatten_function,
2499 inline_to_all_callers_1): Update.
2501 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2503 * profile-count.c (profile_count::combine_with_ipa_count): Handle
2506 2018-01-30 Richard Biener <rguenther@suse.de>
2508 PR tree-optimization/83008
2509 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
2510 invariant and constant vector uses in stmts when they need
2513 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2516 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
2517 * configure: Regenerate.
2519 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2521 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
2523 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
2524 Use gen_rtx_REG rather than gen_lowpart.
2526 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2528 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
2529 rather than 0 when creating partial subregs.
2531 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2533 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
2536 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
2539 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
2540 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
2541 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
2542 flags. This restores the settings used before the 2017-07-24.
2543 Turning off pre increment/decrement/modify allows IVOPTS to
2544 optimize DF/SF loops where the index is an int.
2546 2018-01-29 Richard Biener <rguenther@suse.de>
2547 Kelvin Nilsen <kelvin@gcc.gnu.org>
2550 * tree-vect-stmts.c (vectorizable_call): Don't call
2551 targetm.vectorize_builtin_md_vectorized_function if callee is
2554 2018-01-22 Carl Love <cel@us.ibm.com>
2556 * doc/extend.tex: Fix typo in second arg in
2557 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
2559 2018-01-29 Richard Biener <rguenther@suse.de>
2561 PR tree-optimization/84086
2562 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
2563 (flush_ssaname_freelist): When SSA names were released reset
2564 the SCEV hash table.
2566 2018-01-29 Richard Biener <rguenther@suse.de>
2568 PR tree-optimization/84057
2569 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
2570 removed paths when removing edges.
2572 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
2574 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
2575 -mfunction-return=@var{choice}.
2577 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
2580 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
2582 (layout::print_source_line): Likewise.
2583 (test_get_line_width_without_trailing_whitespace): Add test cases.
2585 2018-01-27 Jakub Jelinek <jakub@redhat.com>
2588 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
2591 2018-01-26 Jim Wilson <jimw@sifive.com>
2593 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
2595 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
2598 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2600 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
2601 and CMP + SUB-immediate -> SUBS.
2603 2018-01-26 Martin Sebor <msebor@redhat.com>
2605 PR tree-optimization/83896
2606 * tree-ssa-strlen.c (get_string_len): Rename...
2607 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
2608 Avoid assuming length is constant.
2609 (handle_char_store): Use HOST_WIDE_INT for string length.
2611 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
2614 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
2615 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
2617 2018-01-26 Richard Biener <rguenther@suse.de>
2619 PR rtl-optimization/84003
2620 * dse.c (record_store): Only record redundant stores when
2621 the earlier store aliases at least all accesses the later one does.
2623 2018-01-26 Jakub Jelinek <jakub@redhat.com>
2625 PR rtl-optimization/83985
2626 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
2627 REG_CFA_RESTORE insns.
2628 (delete_unmarked_insns): Don't ignore separate shrink wrapping
2629 REG_CFA_RESTORE insns here.
2632 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
2633 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
2635 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2637 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
2638 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
2639 (arc_init): Likewise.
2640 (arc_override_options): Likewise.
2641 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
2643 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
2645 * config/arc/arc.h (TARGET_DBNZ): Define.
2646 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
2647 properly set the tune attribute.
2648 (dbnz): Use TARGET_DBNZ guard.
2649 * config/arc/arc.opt (mtune): Add core3 option.
2651 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2653 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
2654 recognize new pic like addresses.
2655 (arc_delegitimize_address): Clean up.
2657 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2659 * config/arc/arc-arches.def: Option mrf16 valid for all
2661 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
2662 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
2663 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
2664 * config/arc/arc-tables.opt: Regenerate.
2665 * config/arc/arc.c (arc_conditional_register_usage): Handle
2666 reduced register file case.
2667 (arc_file_start): Set must have build attributes.
2668 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
2670 * config/arc/arc.opt (mrf16): Add new option.
2671 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
2672 * config/arc/genmultilib.awk: Handle new mrf16 option.
2673 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
2674 * config/arc/t-multilib: Regenerate.
2675 * doc/invoke.texi (ARC Options): Document mrf16 option.
2677 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2679 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
2680 * config/arc/arc.c (arc_handle_secure_attribute): New function.
2681 (arc_attribute_table): Add 'secure_call' attribute.
2682 (arc_print_operand): Print secure call operand.
2683 (arc_function_ok_for_sibcall): Don't optimize tail calls when
2685 (arc_is_secure_call_p): New function. * config/arc/arc.md
2686 (call_i): Add support for sjli instruction.
2687 (call_value_i): Likewise.
2688 * config/arc/constraints.md (Csc): New constraint.
2690 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2691 John Eric Martin <John.Martin@emmicro-us.com>
2693 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
2694 * config/arc/arc.c (_arc_jli_section): New struct.
2695 (arc_jli_section): New type.
2696 (rc_jli_sections): New static variable.
2697 (arc_handle_jli_attribute): New function.
2698 (arc_attribute_table): Add jli_always and jli_fixed attribute.
2699 (arc_file_end): New function.
2700 (TARGET_ASM_FILE_END): Define.
2701 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
2702 (arc_add_jli_section): New function.
2703 (jli_call_scan): Likewise.
2704 (arc_reorg): Call jli_call_scan.
2705 (arc_output_addsi): Remove 'S' from printing asm operand.
2706 (arc_is_jli_call_p): New function.
2707 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
2709 (movhi_insn): Likewise.
2710 (movsi_insn): Likewise.
2711 (movsi_set_cc_insn): Likewise.
2712 (loadqi_update): Likewise.
2713 (load_zeroextendqisi_update): Likewise.
2714 (load_signextendqisi_update): Likewise.
2715 (loadhi_update): Likewise.
2716 (load_zeroextendhisi_update): Likewise.
2717 (load_signextendhisi_update): Likewise.
2718 (loadsi_update): Likewise.
2719 (loadsf_update): Likewise.
2720 (movsicc_insn): Likewise.
2721 (bset_insn): Likewise.
2722 (bxor_insn): Likewise.
2723 (bclr_insn): Likewise.
2724 (bmsk_insn): Likewise.
2725 (bicsi3_insn): Likewise.
2726 (cmpsi_cc_c_insn): Likewise.
2727 (movsi_ne): Likewise.
2728 (movsi_cond_exec): Likewise.
2729 (clrsbsi2): Likewise.
2740 (call_i): Remove 'S' asm letter, add jli instruction.
2741 (call_value_i): Likewise.
2742 * config/arc/arc.op (mjli-always): New option.
2743 * config/arc/constraints.md (Cji): New constraint.
2744 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
2746 (subsf3_fpx): Likewise.
2747 (mulsf3_fpx): Likewise.
2748 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
2750 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
2752 * doc/invoke.texi (ARC): Document mjli-always option.
2754 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
2756 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
2757 avoid addition with 0 and use incw and decw where possible.
2759 2018-01-26 Richard Biener <rguenther@suse.de>
2761 PR tree-optimization/81082
2762 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
2763 association if it requires casting to unsigned.
2764 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
2765 from fold_plusminus_mult_expr to catch important cases late when
2766 range info is available.
2768 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2770 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
2771 * configure.ac (hidden_linkonce): New test.
2772 * configure: Regenerate.
2773 * config.in: Regenerate.
2775 2018-01-26 Julia Koval <julia.koval@intel.com>
2777 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
2778 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
2779 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
2780 _mm_mask_bitshuffle_epi64_mask): Fix type.
2781 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
2782 USI_FTYPE_V4DI_V4DI_USI): Remove.
2783 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
2784 __builtin_ia32_vpshufbitqmb256_mask,
2785 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
2786 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
2787 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
2789 2018-01-26 Alan Modra <amodra@gmail.com>
2792 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
2793 UNSPEC_VBPERMQ. Sort other unspecs.
2795 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
2797 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
2799 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
2802 * predict.c (drop_profile): Do not push/pop cfun; update also
2804 (handle_missing_profiles): Fix logic looking for zero profiles.
2806 2018-01-25 Jakub Jelinek <jakub@redhat.com>
2809 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
2810 on functions with #pragma omp declare simd or functions with simd
2812 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
2813 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
2814 Remove trailing \n from warning_at calls.
2816 2018-01-25 Tom de Vries <tom@codesourcery.com>
2819 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
2820 for neutered workers.
2822 2018-01-24 Joseph Myers <joseph@codesourcery.com>
2825 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
2826 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
2828 2018-01-24 Jeff Law <law@redhat.com>
2831 * i386.c (get_probe_interval): Move to earlier point.
2832 (ix86_compute_frame_layout): If -fstack-clash-protection and
2833 the frame is larger than the probe interval, then use pushes
2834 to save registers rather than reg->mem moves.
2835 (ix86_expand_prologue): Remove conditional for int_registers_saved
2838 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
2841 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
2842 min/max for never referenced object.
2844 2018-01-24 Jakub Jelinek <jakub@redhat.com>
2847 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
2849 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
2850 attributes from DECL_ATTRIBUTES (decl) without affecting
2851 DECL_ATTRIBUTES (current_function_decl).
2852 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
2853 functions with non-NULL DECL_ABSTRACT_ORIGIN.
2855 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
2857 PR tree-optimization/83979
2858 * fold-const.c (fold_comparison): Use constant_boolean_node
2859 instead of boolean_{true,false}_node.
2861 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
2863 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
2866 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2868 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
2869 Simplify the clause that sets the length attribute.
2870 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
2871 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
2872 clause that sets the length attribute.
2873 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
2875 2018-01-24 Tom de Vries <tom@codesourcery.com>
2878 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
2879 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
2880 Add strict parameter.
2881 (prevent_branch_around_nothing): Insert dummy insn between branch to
2882 label and label with no ptx insn inbetween.
2883 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
2885 2018-01-24 Tom de Vries <tom@codesourcery.com>
2888 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
2889 for neutered threads in warp.
2890 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
2892 2018-01-24 Richard Biener <rguenther@suse.de>
2894 PR tree-optimization/83176
2895 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
2898 2018-01-24 Richard Biener <rguenther@suse.de>
2900 PR tree-optimization/82819
2901 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
2902 code generating pluses that are no-ops in the target precision.
2904 2018-01-24 Richard Biener <rguenther@suse.de>
2907 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
2909 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2911 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
2912 to merge probabilities.
2913 * predict.c (probably_never_executed): Also mark as cold functions
2914 with global 0 profile and guessed local profile.
2915 * profile-count.c (profile_probability::combine_with_count): New
2917 * profile-count.h (profile_probability::operator*,
2918 profile_probability::operator*=, profile_probability::operator/,
2919 profile_probability::operator/=): Reduce precision to adjusted
2920 and set value to guessed on contradictory divisions.
2921 (profile_probability::combine_with_freq): Remove.
2922 (profile_probability::combine_wiht_count): Declare.
2923 (profile_count::force_nonzero):: Set to adjusted.
2924 (profile_count::probability_in):: Set quality to adjusted.
2925 * tree-ssa-tail-merge.c (replace_block_by): Use
2928 2018-01-23 Andrew Waterman <andrew@sifive.com>
2929 Jim Wilson <jimw@sifive.com>
2931 * config/riscv/riscv.c (riscv_stack_boundary): New.
2932 (riscv_option_override): Set riscv_stack_boundary. Handle
2933 riscv_preferred_stack_boundary_arg.
2934 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
2935 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
2936 (STACK_BOUNDARY): Set to riscv_stack_boundary.
2937 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
2938 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
2939 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
2941 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
2944 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
2945 of struct ix86_frame.
2946 (ix86_expand_epilogue): Likewise. Add a local variable for
2947 the reg_save_offset field in struct ix86_frame.
2949 2018-01-23 Bin Cheng <bin.cheng@arm.com>
2951 PR tree-optimization/82604
2952 * tree-loop-distribution.c (enum partition_kind): New enum item
2953 PKIND_PARTIAL_MEMSET.
2954 (partition_builtin_p): Support above new enum item.
2955 (generate_code_for_partition): Ditto.
2956 (compute_access_range): Differentiate cases that equality can be
2957 proven at all loops, the innermost loops or no loops.
2958 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
2959 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
2960 (finalize_partitions, distribute_loop): Don't fuse partition of
2961 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
2962 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
2965 2018-01-23 Martin Liska <mliska@suse.cz>
2967 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
2968 order to ignore the predictor.
2969 (PRED_POLYMORPHIC_CALL): Likewise.
2970 (PRED_RECURSIVE_CALL): Likewise.
2972 2018-01-23 Martin Liska <mliska@suse.cz>
2974 * tree-profile.c (tree_profiling): Print function header to
2975 aware reader which function we are working on.
2976 * value-prof.c (gimple_find_values_to_profile): Do not print
2977 not interesting value histograms.
2979 2018-01-23 Martin Liska <mliska@suse.cz>
2981 * profile-count.h (enum profile_quality): Add
2982 profile_uninitialized as the first value. Do not number values
2983 as they are zero based.
2984 (profile_count::verify): Update sanity check.
2985 (profile_probability::verify): Likewise.
2987 2018-01-23 Nathan Sidwell <nathan@acm.org>
2989 * doc/invoke.texi (ffor-scope): Deprecate.
2991 2018-01-23 David Malcolm <dmalcolm@redhat.com>
2993 PR tree-optimization/83510
2994 * domwalk.c (set_all_edges_as_executable): New function.
2995 (dom_walker::dom_walker): Convert bool param
2996 "skip_unreachable_blocks" to enum reachability. Move setup of
2997 edge flags to set_all_edges_as_executable and only do it when
2998 reachability is REACHABLE_BLOCKS.
2999 * domwalk.h (enum dom_walker::reachability): New enum.
3000 (dom_walker::dom_walker): Convert bool param
3001 "skip_unreachable_blocks" to enum reachability.
3002 (set_all_edges_as_executable): New decl.
3003 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
3004 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
3006 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
3007 but converting true to REACHABLE_BLOCKS.
3008 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
3010 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
3011 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
3012 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
3014 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
3015 if check_all_array_refs will be called.
3017 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3019 * tree.c (selftest::test_location_wrappers): Add more test
3022 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3024 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
3025 (selftest::test_bit_in_range): Likewise.
3027 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3030 * doc/sourcebuild.texi (vect_float): Say that the selector
3031 only describes the situation when -funsafe-math-optimizations is on.
3032 (vect_float_strict): Document.
3034 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3036 PR tree-optimization/83965
3037 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
3038 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
3039 instead of checking only for a reduction.
3040 (vect_recog_widen_sum_pattern): Likewise.
3042 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3044 * predict.c (probably_never_executed): Only use precise profile info.
3045 (compute_function_frequency): Skip after inlining hack since we now
3046 have quality checking.
3048 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3050 * profile-count.h (profile_probability::very_unlikely,
3051 profile_probability::unlikely, profile_probability::even): Set
3052 precision to guessed.
3054 2018-01-23 Richard Biener <rguenther@suse.de>
3056 PR tree-optimization/83963
3057 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
3058 Properly terminate dominator walk when crossing the exit edge not
3059 when visiting its source block.
3061 2018-01-23 Jakub Jelinek <jakub@redhat.com>
3064 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
3065 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
3067 2018-01-22 Jakub Jelinek <jakub@redhat.com>
3069 PR tree-optimization/83957
3070 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
3071 semicolon after for body surrounded by braces.
3073 PR tree-optimization/83081
3074 * profile-count.h (profile_probability::split): New method.
3075 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
3076 Use profile_probability::split.
3077 (do_compare_rtx_and_jump): Fix adjustment of probabilities
3078 when splitting a single conditional jump into 2.
3080 2018-01-22 David Malcolm <dmalcolm@redhat.com>
3082 PR tree-optimization/69452
3083 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
3086 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3088 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
3089 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
3090 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
3092 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3094 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
3095 * config/rl78/rl78.md (movdi): New define_expand.
3096 * config/rl78/rl78.c (rl78_split_movdi): New function.
3098 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
3101 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
3103 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
3104 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
3105 128-bit to produce an UNSPEC move to get the double word with the
3106 signbit and then a shift directly to do signbit.
3107 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
3108 implementation with a new version that just does either a direct
3109 move or a regular move. Move memory interface to separate insns.
3110 Move insns so they are next to the expander.
3111 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
3112 with signbit move. Split big and little endian case.
3113 (signbit<mode>2_dm_mem_le): Likewise.
3114 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
3115 (signbit<mode>2_dm2): Likewise.
3117 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3119 * config/rl78/rl78.md (anddi3): New define_expand.
3121 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3123 * config/rl78/rl78.md (umindi3): New define_expand.
3125 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3127 * config/rl78/rl78.md (smindi3): New define_expand.
3129 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3131 * config/rl78/rl78.md (smaxdi3): New define_expand.
3133 2018-01-22 Carl Love <cel@us.ibm.com>
3135 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
3136 LVX_V1TI): Add macro expansion.
3137 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
3138 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
3139 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
3140 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
3141 Change check to determine if the instruction is a byte reversing
3142 entry. Fix typo in comment.
3143 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
3144 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
3145 Add def_builtin calls for new builtins.
3146 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
3147 Add define_insn expansion.
3149 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3151 * config/rl78/rl78.md (umaxdi3): New define_expand.
3153 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3155 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
3156 for non-QImode registers.
3158 2018-01-22 Richard Biener <rguenther@suse.de>
3160 PR tree-optimization/83963
3161 * graphite-scop-detection.c (scop_detection::get_sese): Delay
3162 including the loop exit block.
3163 (scop_detection::merge_sese): Likewise.
3164 (scop_detection::add_scop): Do it here instead.
3166 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3168 * doc/sourcebuild.texi (arm_softfloat): Document.
3170 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
3173 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
3174 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
3175 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
3177 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3178 David Edelsohn <dje.gcc@gmail.com>
3181 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3182 Change "crset eq" to "crset 2".
3183 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3184 (*call_indirect_aix<mode>_nospec): Likewise.
3185 (*call_value_indirect_aix<mode>_nospec): Likewise.
3186 (*call_indirect_elfv2<mode>_nospec): Likewise.
3187 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
3188 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
3189 change assembly output from . to $.
3190 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3191 (indirect_jump<mode>_nospec): Change assembly output from . to $.
3192 (*tablejump<mode>_internal1_nospec): Likewise.
3194 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
3197 * config/sh/sh_optimize_sett_clrt.cc:
3198 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
3200 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
3202 PR tree-optimization/83940
3203 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
3204 offset_dt to vect_constant_def rather than vect_unknown_def_type.
3205 (vect_check_load_store_mask): Add a mask_dt_out parameter and
3206 use it to pass back the definition type.
3207 (vect_check_store_rhs): Likewise rhs_dt_out.
3208 (vect_build_gather_load_calls): Add a mask_dt argument and use
3209 it instead of a call to vect_is_simple_use.
3210 (vectorizable_store): Update calls to vect_check_load_store_mask
3211 and vect_check_store_rhs. Use the dt returned by the latter instead
3212 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
3213 instead of calls to vect_is_simple_use. Pass the scalar rather
3214 than the vector operand to vect_is_simple_use when handling
3215 second and subsequent copies of an rhs value.
3216 (vectorizable_load): Update calls to vect_check_load_store_mask
3217 and vect_build_gather_load_calls. Use the cached mask_dt and
3218 gs_info.offset_dt instead of calls to vect_is_simple_use.
3220 2018-01-20 Jakub Jelinek <jakub@redhat.com>
3223 * tree-emutls.c: Include gimplify.h.
3224 (lower_emutls_2): New function.
3225 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
3226 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
3227 it before further processing.
3230 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
3231 UINTVAL (trueop1) instead of INTVAL (op1).
3233 2018-01-19 Jakub Jelinek <jakub@redhat.com>
3237 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
3238 INCOMING_FRAME_SP_OFFSET if not defined.
3239 (scan_trace): Add ENTRY argument. If true and
3240 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
3241 emit a note to adjust the CFA offset.
3242 (create_cfi_notes): Adjust scan_trace callers.
3243 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
3244 INCOMING_FRAME_SP_OFFSET in the CIE.
3245 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
3246 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
3248 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
3249 * doc/tm.texi: Regenerated.
3251 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3253 PR rtl-optimization/83147
3254 * lra-constraints.c (remove_inheritance_pseudos): Use
3255 lra_substitute_pseudo_within_insn.
3257 2018-01-19 Tom de Vries <tom@codesourcery.com>
3258 Cesar Philippidis <cesar@codesourcery.com>
3261 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
3263 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
3266 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
3267 spaces for function labels.
3269 2018-01-19 Martin Liska <mliska@suse.cz>
3271 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
3272 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
3273 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
3274 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
3275 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
3276 (PRED_CONST_RETURN): Change from 69 to 65.
3277 (PRED_NULL_RETURN): Change from 91 to 71.
3278 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
3279 (PRED_LOOP_GUARD): Change from 66 to 73.
3281 2018-01-19 Martin Liska <mliska@suse.cz>
3283 * predict.c (predict_insn_def): Add new assert.
3284 (struct branch_predictor): Change type to signed integer.
3285 (test_prediction_value_range): Amend test to cover
3287 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
3288 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
3289 (PRED_LOOP_ITERATIONS_MAX): Likewise.
3290 (PRED_LOOP_IV_COMPARE): Likewise.
3291 * predict.h (PROB_UNINITIALIZED): Define new constant.
3293 2018-01-19 Martin Liska <mliska@suse.cz>
3295 * predict.c (dump_prediction): Add new format for
3296 analyze_brprob.py script which is enabled with -details
3298 * profile-count.h (precise_p): New function.
3300 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3302 PR tree-optimization/83922
3303 * tree-vect-loop.c (vect_verify_full_masking): Return false if
3304 there are no statements that need masking.
3305 (vect_active_double_reduction_p): New function.
3306 (vect_analyze_loop_operations): Use it when handling phis that
3307 are not in the loop header.
3309 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3311 PR tree-optimization/83914
3312 * tree-vect-loop.c (vectorizable_induction): Don't convert
3313 init_expr or apply the peeling adjustment for inductions
3314 that are nested within the vectorized loop.
3316 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3318 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
3321 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3325 * function.h (gimplify_parameters): Add gimple_seq * argument.
3326 * function.c: Include gimple.h and options.h.
3327 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
3328 for the added local temporaries if needed.
3329 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
3330 if there are any parameter cleanups, wrap whole body into a
3331 try/finally with the cleanups.
3333 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
3336 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
3337 Use GET_MODE_CLASS for scalar floating point.
3339 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3343 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
3344 Fix call of call_cgraph_insertion_hooks.
3346 2018-01-18 Martin Sebor <msebor@redhat.com>
3348 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
3350 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3353 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
3356 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
3359 * common.opt: (-ffile-prefix-map): New option.
3360 * opts.c (common_handle_option): Defer it.
3361 * opts-global.c (handle_common_deferred_options): Handle it.
3362 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
3363 * file-prefix-map.h: New file.
3364 (remap_debug_filename, add_debug_prefix_map): ...here.
3365 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
3366 * final.c (debug_prefix_map, add_debug_prefix_map
3367 remap_debug_filename): Move to...
3368 * file-prefix-map.c: New file.
3369 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
3370 generalize, get rid of alloca(), use strrchr() instead of strchr().
3371 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
3372 Implement in terms of add_prefix_map().
3373 (remap_macro_filename, remap_debug_filename): Implement in term of
3375 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
3376 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
3377 * dbxout.c: Include file-prefix-map.h.
3378 * varasm.c: Likewise.
3379 * vmsdbgout.c: Likewise.
3380 * xcoffout.c: Likewise.
3381 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
3382 * doc/cppopts.texi (-fmacro-prefix-map): Document.
3383 * doc/invoke.texi (-ffile-prefix-map): Document.
3384 (-fdebug-prefix-map): Update description.
3386 2018-01-18 Martin Liska <mliska@suse.cz>
3388 * config/i386/i386.c (indirect_thunk_name): Document that also
3390 (output_indirect_thunk): Document why both instructions
3391 (pause and lfence) are generated.
3393 2018-01-18 Richard Biener <rguenther@suse.de>
3395 PR tree-optimization/83887
3396 * graphite-scop-detection.c
3397 (scop_detection::get_nearest_dom_with_single_entry): Remove.
3398 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
3399 (scop_detection::merge_sese): Re-implement with a flood-fill
3400 algorithm that properly finds a SESE region if it exists.
3402 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3405 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
3406 pointer_diff optimizations use view_convert instead of convert.
3408 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3410 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3411 Generate different code for -mno-speculate-indirect-jumps.
3412 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3413 (*call_indirect_aix<mode>): Disable for
3414 -mno-speculate-indirect-jumps.
3415 (*call_indirect_aix<mode>_nospec): New define_insn.
3416 (*call_value_indirect_aix<mode>): Disable for
3417 -mno-speculate-indirect-jumps.
3418 (*call_value_indirect_aix<mode>_nospec): New define_insn.
3419 (*sibcall_nonlocal_sysv<mode>): Generate different code for
3420 -mno-speculate-indirect-jumps.
3421 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3423 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3425 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
3426 long double type, set the flags for noting the default long double
3427 type, even if we don't pass or return a long double type.
3429 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
3432 * ipa-inline.c (flatten_function): Do not overwrite final inlining
3435 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
3437 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
3438 support for merge[hl].
3439 (fold_mergehl_helper): New helper function.
3440 (tree-vector-builder.h): New #include for tree_vector_builder usage.
3441 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
3442 (altivec_vmrglw_direct): Add xxmrglw insn.
3444 2018-01-17 Andrew Waterman <andrew@sifive.com>
3446 * config/riscv/riscv.c (riscv_conditional_register_usage): If
3447 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
3449 2018-01-17 David Malcolm <dmalcolm@redhat.com>
3452 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
3453 call the lto_location_cache before reading the
3454 DECL_SOURCE_LOCATION of the types.
3456 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3457 Richard Sandiford <richard.sandiford@linaro.org>
3459 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
3460 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
3461 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
3462 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
3463 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
3465 * config/aarch64/constraints.md (aarch64_movti_operand):
3467 * config/aarch64/predicates.md (Uti): Add new constraint.
3469 2018-01-17 Carl Love <cel@us.ibm.com>
3471 * config/rs6000/vsx.md (define_expand xl_len_r,
3472 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
3473 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
3475 (define_expand, define_insn): Move the shift left from the
3476 define_insn to the define_expand for lxvl and stxvl instructions.
3477 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
3478 and XL_LEN_R definitions to PURE.
3480 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
3482 * config/i386/i386.c (indirect_thunk_name): Declare regno
3483 as unsigned int. Compare regno with INVALID_REGNUM.
3484 (output_indirect_thunk): Ditto.
3485 (output_indirect_thunk_function): Ditto.
3486 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
3487 in the call to output_indirect_thunk_function.
3489 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
3492 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
3493 rather than the size of inner_type to determine the stack slot size
3494 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
3496 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
3499 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
3502 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
3504 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
3505 endian Linux systems to optionally enable multilibs for selecting
3506 the long double type if the user configured an explicit type.
3507 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
3508 have no long double multilibs if not defined.
3509 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
3510 warn if the user used -mabi={ieee,ibm}longdouble and we built
3511 multilibs for long double.
3512 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
3513 appropriate multilib option.
3514 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
3516 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
3517 for building long double multilibs.
3518 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
3520 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
3522 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
3525 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
3527 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
3530 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
3533 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
3536 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3538 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
3539 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
3541 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
3543 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
3544 different rtl trees depending on TARGET_64BIT.
3545 (rs6000_gen_lvx): Likewise.
3547 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3549 * config/visium/visium.md (nop): Tweak comment.
3550 (hazard_nop): Likewise.
3552 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3554 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
3555 -mspeculate-indirect-jumps.
3556 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
3557 for -mno-speculate-indirect-jumps.
3558 (*call_indirect_elfv2<mode>_nospec): New define_insn.
3559 (*call_value_indirect_elfv2<mode>): Disable for
3560 -mno-speculate-indirect-jumps.
3561 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
3562 (indirect_jump): Emit different RTL for
3563 -mno-speculate-indirect-jumps.
3564 (*indirect_jump<mode>): Disable for
3565 -mno-speculate-indirect-jumps.
3566 (*indirect_jump<mode>_nospec): New define_insn.
3567 (tablejump): Emit different RTL for
3568 -mno-speculate-indirect-jumps.
3569 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
3570 (tablejumpsi_nospec): New define_expand.
3571 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
3572 (tablejumpdi_nospec): New define_expand.
3573 (*tablejump<mode>_internal1): Disable for
3574 -mno-speculate-indirect-jumps.
3575 (*tablejump<mode>_internal1_nospec): New define_insn.
3576 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
3579 2018-01-16 Artyom Skrobov tyomitch@gmail.com
3581 * caller-save.c (insert_save): Drop unnecessary parameter. All
3584 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3585 Richard Biener <rguenth@suse.de>
3588 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
3589 return early, inline manually is_gimple_sizepos. Make sure if we
3590 call gimplify_expr we don't end up with a gimple constant.
3591 * tree.c (variably_modified_type_p): Don't return true for
3592 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
3593 * gimplify.h (is_gimple_sizepos): Remove.
3595 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3597 PR tree-optimization/83857
3598 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
3599 vectorizable_live_operation for pure SLP statements.
3600 (vectorizable_live_operation): Handle PHIs.
3602 2018-01-16 Richard Biener <rguenther@suse.de>
3604 PR tree-optimization/83867
3605 * tree-vect-stmts.c (vect_transform_stmt): Precompute
3606 nested_in_vect_loop_p since the scalar stmt may get invalidated.
3608 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3611 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
3612 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
3613 If off is not INTEGER_CST, issue a may not be aligned warning
3614 rather than isn't aligned. Use isn%'t rather than isn't.
3615 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
3617 <case MULT_EXPR>: Improve the case when bottom and one of the
3618 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
3619 operand, in that case check if the other operand is multiple of
3620 bottom divided by the INTEGER_CST operand.
3622 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3625 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
3626 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
3627 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
3628 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
3629 * config/pa/pa.c (pa_function_arg_advance): Likewise.
3630 (pa_function_arg, pa_arg_partial_bytes): Likewise.
3631 (pa_function_arg_size): New function.
3633 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3635 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
3636 in a separate statement.
3638 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3640 PR tree-optimization/83847
3641 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
3642 group gathers and scatters.
3644 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3646 PR rtl-optimization/86620
3647 * params.def (max-sched-ready-insns): Bump minimum value to 1.
3649 PR rtl-optimization/83213
3650 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
3651 to last if both are JUMP_INSNs.
3653 PR tree-optimization/83843
3654 * gimple-ssa-store-merging.c
3655 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
3656 store_immediate_info for bswap/nop orig_stores.
3658 2018-01-15 Andrew Waterman <andrew@sifive.com>
3660 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
3662 <UDIV>: Increase cost if !TARGET_DIV.
3664 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
3666 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
3667 (define_attr "cr_logical_3op"): New.
3668 (cceq_ior_compare): Adjust.
3669 (cceq_ior_compare_complement): Adjust.
3670 (*cceq_rev_compare): Adjust.
3671 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
3672 (is_cracked_insn): Adjust.
3673 (insn_must_be_first_in_group): Adjust.
3674 * config/rs6000/40x.md: Adjust.
3675 * config/rs6000/440.md: Adjust.
3676 * config/rs6000/476.md: Adjust.
3677 * config/rs6000/601.md: Adjust.
3678 * config/rs6000/603.md: Adjust.
3679 * config/rs6000/6xx.md: Adjust.
3680 * config/rs6000/7450.md: Adjust.
3681 * config/rs6000/7xx.md: Adjust.
3682 * config/rs6000/8540.md: Adjust.
3683 * config/rs6000/cell.md: Adjust.
3684 * config/rs6000/e300c2c3.md: Adjust.
3685 * config/rs6000/e500mc.md: Adjust.
3686 * config/rs6000/e500mc64.md: Adjust.
3687 * config/rs6000/e5500.md: Adjust.
3688 * config/rs6000/e6500.md: Adjust.
3689 * config/rs6000/mpc.md: Adjust.
3690 * config/rs6000/power4.md: Adjust.
3691 * config/rs6000/power5.md: Adjust.
3692 * config/rs6000/power6.md: Adjust.
3693 * config/rs6000/power7.md: Adjust.
3694 * config/rs6000/power8.md: Adjust.
3695 * config/rs6000/power9.md: Adjust.
3696 * config/rs6000/rs64.md: Adjust.
3697 * config/rs6000/titan.md: Adjust.
3699 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3701 * config/i386/predicates.md (indirect_branch_operand): Rewrite
3702 ix86_indirect_branch_register logic.
3704 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3706 * config/i386/constraints.md (Bs): Update
3707 ix86_indirect_branch_register check. Don't check
3708 ix86_indirect_branch_register with GOT_memory_operand.
3710 * config/i386/predicates.md (GOT_memory_operand): Don't check
3711 ix86_indirect_branch_register here.
3712 (GOT32_symbol_operand): Likewise.
3714 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3716 * config/i386/predicates.md (constant_call_address_operand):
3717 Rewrite ix86_indirect_branch_register logic.
3718 (sibcall_insn_operand): Likewise.
3720 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3722 * config/i386/constraints.md (Bs): Replace
3723 ix86_indirect_branch_thunk_register with
3724 ix86_indirect_branch_register.
3726 * config/i386/i386.md (indirect_jump): Likewise.
3727 (tablejump): Likewise.
3728 (*sibcall_memory): Likewise.
3729 (*sibcall_value_memory): Likewise.
3730 Peepholes of indirect call and jump via memory: Likewise.
3731 * config/i386/i386.opt: Likewise.
3732 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3733 (GOT_memory_operand): Likewise.
3734 (call_insn_operand): Likewise.
3735 (sibcall_insn_operand): Likewise.
3736 (GOT32_symbol_operand): Likewise.
3738 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3741 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
3742 type rather than type addr's type points to.
3743 (expand_omp_atomic_mutex): Likewise.
3744 (expand_omp_atomic): Likewise.
3746 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3749 * config/i386/i386.c (output_indirect_thunk_function): Use
3750 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
3751 for __x86_return_thunk.
3753 2018-01-15 Richard Biener <rguenther@suse.de>
3756 * expmed.c (extract_bit_field_1): Fix typo.
3758 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3761 * config/arm/iterators.md (VF): New mode iterator.
3762 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
3763 Remove integer-related logic from pattern.
3764 (neon_vabd<mode>_3): Likewise.
3766 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3769 * common.opt (fstrict-overflow): No longer an alias.
3770 (fwrapv-pointer): New option.
3771 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
3772 also for pointer types based on flag_wrapv_pointer.
3773 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
3774 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
3775 opts->x_flag_wrapv got set.
3776 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
3777 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
3778 POINTER_TYPE_OVERFLOW_UNDEFINED.
3779 * match.pd: Likewise in address comparison pattern.
3780 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
3782 2018-01-15 Richard Biener <rguenther@suse.de>
3785 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
3786 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
3787 Reset type names to their identifier if their TYPE_DECL doesn't
3788 have linkage (and thus is used for ODR and devirt).
3789 (save_debug_info_for_decl): Remove.
3790 (save_debug_info_for_type): Likewise.
3791 (add_tree_to_fld_list): Adjust.
3792 * tree-pretty-print.c (dump_generic_node): Make dumping of
3793 type names more robust.
3795 2018-01-15 Richard Biener <rguenther@suse.de>
3797 * BASE-VER: Bump to 8.0.1.
3799 2018-01-14 Martin Sebor <msebor@redhat.com>
3802 * builtins.c (check_access): Avoid warning when the no-warning bit
3805 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
3807 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
3808 * ira-color (allocno_hard_regs_compare): Likewise.
3810 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
3813 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
3814 Use .pushsection/.popsection.
3816 2018-01-14 Martin Sebor <msebor@redhat.com>
3819 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
3821 2018-01-14 Jakub Jelinek <jakub@redhat.com>
3823 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
3824 entry from extra_headers.
3825 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
3826 extra_headers, make the list bitwise identical to the i?86-*-* one.
3828 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3830 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
3831 -mcmodel=large with -mindirect-branch=thunk,
3832 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
3833 -mfunction-return=thunk-extern.
3834 * doc/invoke.texi: Document -mcmodel=large is incompatible with
3835 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
3836 -mfunction-return=thunk and -mfunction-return=thunk-extern.
3838 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3840 * config/i386/i386.c (print_reg): Print the name of the full
3841 integer register without '%'.
3842 (ix86_print_operand): Handle 'V'.
3843 * doc/extend.texi: Document 'V' modifier.
3845 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3847 * config/i386/constraints.md (Bs): Disallow memory operand for
3848 -mindirect-branch-register.
3850 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3851 (GOT_memory_operand): Likewise.
3852 (call_insn_operand): Likewise.
3853 (sibcall_insn_operand): Likewise.
3854 (GOT32_symbol_operand): Likewise.
3855 * config/i386/i386.md (indirect_jump): Call convert_memory_address
3856 for -mindirect-branch-register.
3857 (tablejump): Likewise.
3858 (*sibcall_memory): Likewise.
3859 (*sibcall_value_memory): Likewise.
3860 Disallow peepholes of indirect call and jump via memory for
3861 -mindirect-branch-register.
3862 (*call_pop): Replace m with Bw.
3863 (*call_value_pop): Likewise.
3864 (*sibcall_pop_memory): Replace m with Bs.
3865 * config/i386/i386.opt (mindirect-branch-register): New option.
3866 * doc/invoke.texi: Document -mindirect-branch-register option.
3868 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3870 * config/i386/i386-protos.h (ix86_output_function_return): New.
3871 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
3872 set function_return_type.
3873 (indirect_thunk_name): Add ret_p to indicate thunk for function
3875 (output_indirect_thunk_function): Pass false to
3876 indirect_thunk_name.
3877 (ix86_output_indirect_branch_via_reg): Likewise.
3878 (ix86_output_indirect_branch_via_push): Likewise.
3879 (output_indirect_thunk_function): Create alias for function
3880 return thunk if regno < 0.
3881 (ix86_output_function_return): New function.
3882 (ix86_handle_fndecl_attribute): Handle function_return.
3883 (ix86_attribute_table): Add function_return.
3884 * config/i386/i386.h (machine_function): Add
3885 function_return_type.
3886 * config/i386/i386.md (simple_return_internal): Use
3887 ix86_output_function_return.
3888 (simple_return_internal_long): Likewise.
3889 * config/i386/i386.opt (mfunction-return=): New option.
3890 (indirect_branch): Mention -mfunction-return=.
3891 * doc/extend.texi: Document function_return function attribute.
3892 * doc/invoke.texi: Document -mfunction-return= option.
3894 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3896 * config/i386/i386-opts.h (indirect_branch): New.
3897 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
3898 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
3899 with local indirect jump when converting indirect call and jump.
3900 (ix86_set_indirect_branch_type): New.
3901 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
3902 (indirectlabelno): New.
3903 (indirect_thunk_needed): Likewise.
3904 (indirect_thunk_bnd_needed): Likewise.
3905 (indirect_thunks_used): Likewise.
3906 (indirect_thunks_bnd_used): Likewise.
3907 (INDIRECT_LABEL): Likewise.
3908 (indirect_thunk_name): Likewise.
3909 (output_indirect_thunk): Likewise.
3910 (output_indirect_thunk_function): Likewise.
3911 (ix86_output_indirect_branch_via_reg): Likewise.
3912 (ix86_output_indirect_branch_via_push): Likewise.
3913 (ix86_output_indirect_branch): Likewise.
3914 (ix86_output_indirect_jmp): Likewise.
3915 (ix86_code_end): Call output_indirect_thunk_function if needed.
3916 (ix86_output_call_insn): Call ix86_output_indirect_branch if
3918 (ix86_handle_fndecl_attribute): Handle indirect_branch.
3919 (ix86_attribute_table): Add indirect_branch.
3920 * config/i386/i386.h (machine_function): Add indirect_branch_type
3921 and has_local_indirect_jump.
3922 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
3924 (tablejump): Likewise.
3925 (*indirect_jump): Use ix86_output_indirect_jmp.
3926 (*tablejump_1): Likewise.
3927 (simple_return_indirect_internal): Likewise.
3928 * config/i386/i386.opt (mindirect-branch=): New option.
3929 (indirect_branch): New.
3932 (thunk-inline): Likewise.
3933 (thunk-extern): Likewise.
3934 * doc/extend.texi: Document indirect_branch function attribute.
3935 * doc/invoke.texi: Document -mindirect-branch= option.
3937 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
3940 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
3942 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
3944 * ipa-inline.c (want_inline_small_function_p): Return false if
3945 inlining has already failed with CIF_FINAL_ERROR.
3946 (update_caller_keys): Call want_inline_small_function_p before
3948 (update_callee_keys): Likewise.
3950 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3952 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
3954 (rs6000_quadword_masked_address_p): Likewise.
3955 (quad_aligned_load_p): Likewise.
3956 (quad_aligned_store_p): Likewise.
3957 (const_load_sequence_p): Add comment to describe the outer-most loop.
3958 (mimic_memory_attributes_and_flags): New function.
3959 (rs6000_gen_stvx): Likewise.
3960 (replace_swapped_aligned_store): Likewise.
3961 (rs6000_gen_lvx): Likewise.
3962 (replace_swapped_aligned_load): Likewise.
3963 (replace_swapped_load_constant): Capitalize argument name in
3964 comment describing this function.
3965 (rs6000_analyze_swaps): Add a third pass to search for vector loads
3966 and stores that access quad-word aligned addresses and replace
3967 with stvx or lvx instructions when appropriate.
3968 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
3969 New function prototype.
3970 (rs6000_quadword_masked_address_p): Likewise.
3971 (rs6000_gen_lvx): Likewise.
3972 (rs6000_gen_stvx): Likewise.
3973 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
3974 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
3975 when memory address is aligned.
3976 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
3977 this split to select lvx instruction when memory address is aligned.
3978 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
3979 instruction when memory address is aligned.
3980 (*vsx_le_perm_load_v16qi): Likewise.
3981 (four unnamed splitters): Modify to select the stvx instruction
3982 when memory is aligned.
3984 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
3986 * predict.c (determine_unlikely_bbs): Handle correctly BBs
3987 which appears in the queue multiple times.
3989 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3990 Alan Hayward <alan.hayward@arm.com>
3991 David Sherwood <david.sherwood@arm.com>
3993 * tree-vectorizer.h (vec_lower_bound): New structure.
3994 (_loop_vec_info): Add check_nonzero and lower_bounds.
3995 (LOOP_VINFO_CHECK_NONZERO): New macro.
3996 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
3997 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
3998 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
3999 fields. Make seg_len the distance travelled, not including the
4001 (dr_direction_indicator): Declare.
4002 (dr_zero_step_indicator): Likewise.
4003 (dr_known_forward_stride_p): Likewise.
4004 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
4006 (runtime_alias_check_p): Allow runtime alias checks with
4008 (operator ==): Compare access_size and align.
4009 (prune_runtime_alias_test_list): Rework for new distinction between
4010 the access_size and seg_len.
4011 (create_intersect_range_checks_index): Likewise. Cope with polynomial
4013 (get_segment_min_max): New function.
4014 (create_intersect_range_checks): Use it.
4015 (dr_step_indicator): New function.
4016 (dr_direction_indicator): Likewise.
4017 (dr_zero_step_indicator): Likewise.
4018 (dr_known_forward_stride_p): Likewise.
4019 * tree-loop-distribution.c (data_ref_segment_size): Return
4020 DR_STEP * (niters - 1).
4021 (compute_alias_check_pairs): Update call to the dr_with_seg_len
4023 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
4024 (vect_preserves_scalar_order_p): New function, split out from...
4025 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
4026 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
4027 (vect_vfa_access_size): New function.
4028 (vect_vfa_align): Likewise.
4029 (vect_compile_time_alias): Take access_size_a and access_b arguments.
4030 (dump_lower_bound): New function.
4031 (vect_check_lower_bound): Likewise.
4032 (vect_small_gap_p): Likewise.
4033 (vectorizable_with_step_bound_p): Likewise.
4034 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
4035 depencies if the vectorization factor is 1. Convert the checks
4036 for nonzero steps into checks on the bounds of DR_STEP. Try using
4037 a bunds check for variable steps if the minimum required step is
4038 relatively small. Update calls to the dr_with_seg_len
4039 constructor and to vect_compile_time_alias.
4040 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
4042 (vect_loop_versioning): Call it.
4043 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
4045 (vect_estimate_min_profitable_iters): Account for any bounds checks.
4047 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4048 Alan Hayward <alan.hayward@arm.com>
4049 David Sherwood <david.sherwood@arm.com>
4051 * doc/sourcebuild.texi (vect_scatter_store): Document.
4052 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
4054 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
4056 * genopinit.c (main): Add supports_vec_scatter_store and
4057 supports_vec_scatter_store_cached to target_optabs.
4058 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
4059 IFN_MASK_SCATTER_STORE.
4060 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
4062 * internal-fn.h (internal_store_fn_p): Declare.
4063 (internal_fn_stored_value_index): Likewise.
4064 * internal-fn.c (scatter_store_direct): New macro.
4065 (expand_scatter_store_optab_fn): New function.
4066 (direct_scatter_store_optab_supported_p): New macro.
4067 (internal_store_fn_p): New function.
4068 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
4069 IFN_MASK_SCATTER_STORE.
4070 (internal_fn_mask_index): Likewise.
4071 (internal_fn_stored_value_index): New function.
4072 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
4074 * optabs-query.h (supports_vec_scatter_store_p): Declare.
4075 * optabs-query.c (supports_vec_scatter_store_p): New function.
4076 * tree-vectorizer.h (vect_get_store_rhs): Declare.
4077 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
4078 true for scatter stores.
4079 (vect_gather_scatter_fn_p): Handle scatter stores too.
4080 (vect_check_gather_scatter): Consider using scatter stores if
4081 supports_vec_scatter_store_p.
4082 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
4084 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4085 internal_fn_stored_value_index.
4086 (check_load_store_masking): Handle scatter stores too.
4087 (vect_get_store_rhs): Make public.
4088 (vectorizable_call): Use internal_store_fn_p.
4089 (vectorizable_store): Handle scatter store internal functions.
4090 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
4091 when deciding whether the end of the group has been reached.
4092 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
4093 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
4094 (mask_scatter_store<mode>): New insns.
4096 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4097 Alan Hayward <alan.hayward@arm.com>
4098 David Sherwood <david.sherwood@arm.com>
4100 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
4101 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
4102 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
4104 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
4105 Use vect_truncate_gather_scatter_offset if we can't treat the
4106 operation as a normal gather load or scatter store.
4107 (get_group_load_store_type): Take the gather_scatter_info
4108 as argument. Try using a gather load or scatter store for
4109 single-element groups.
4110 (get_load_store_type): Update calls to get_group_load_store_type
4111 and vect_use_strided_gather_scatters_p.
4113 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4114 Alan Hayward <alan.hayward@arm.com>
4115 David Sherwood <david.sherwood@arm.com>
4117 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
4118 optional tree argument.
4119 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
4121 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
4122 but continue to use the current value as a fallback.
4123 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
4124 to compare the updates.
4125 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
4126 (get_load_store_type): Use it when handling a strided access.
4127 (vect_get_strided_load_store_ops): New function.
4128 (vect_get_data_ptr_increment): Likewise.
4129 (vectorizable_load): Handle strided gather loads. Always pass
4130 a step to vect_create_data_ref_ptr and bump_vector_ptr.
4132 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4133 Alan Hayward <alan.hayward@arm.com>
4134 David Sherwood <david.sherwood@arm.com>
4136 * doc/md.texi (gather_load@var{m}): Document.
4137 (mask_gather_load@var{m}): Likewise.
4138 * genopinit.c (main): Add supports_vec_gather_load and
4139 supports_vec_gather_load_cached to target_optabs.
4140 * optabs-tree.c (init_tree_optimization_optabs): Use
4141 ggc_cleared_alloc to allocate target_optabs.
4142 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
4143 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
4145 * internal-fn.h (internal_load_fn_p): Declare.
4146 (internal_gather_scatter_fn_p): Likewise.
4147 (internal_fn_mask_index): Likewise.
4148 (internal_gather_scatter_fn_supported_p): Likewise.
4149 * internal-fn.c (gather_load_direct): New macro.
4150 (expand_gather_load_optab_fn): New function.
4151 (direct_gather_load_optab_supported_p): New macro.
4152 (direct_internal_fn_optab): New function.
4153 (internal_load_fn_p): Likewise.
4154 (internal_gather_scatter_fn_p): Likewise.
4155 (internal_fn_mask_index): Likewise.
4156 (internal_gather_scatter_fn_supported_p): Likewise.
4157 * optabs-query.c (supports_at_least_one_mode_p): New function.
4158 (supports_vec_gather_load_p): Likewise.
4159 * optabs-query.h (supports_vec_gather_load_p): Declare.
4160 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
4161 and memory_type field.
4162 (NUM_PATTERNS): Bump to 15.
4163 * tree-vect-data-refs.c: Include internal-fn.h.
4164 (vect_gather_scatter_fn_p): New function.
4165 (vect_describe_gather_scatter_call): Likewise.
4166 (vect_check_gather_scatter): Try using internal functions for
4167 gather loads. Recognize existing calls to a gather load function.
4168 (vect_analyze_data_refs): Consider using gather loads if
4169 supports_vec_gather_load_p.
4170 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
4171 (vect_get_gather_scatter_offset_type): Likewise.
4172 (vect_convert_mask_for_vectype): Likewise.
4173 (vect_add_conversion_to_patterm): Likewise.
4174 (vect_try_gather_scatter_pattern): Likewise.
4175 (vect_recog_gather_scatter_pattern): New pattern recognizer.
4176 (vect_vect_recog_func_ptrs): Add it.
4177 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4178 internal_fn_mask_index and internal_gather_scatter_fn_p.
4179 (check_load_store_masking): Take the gather_scatter_info as an
4180 argument and handle gather loads.
4181 (vect_get_gather_scatter_ops): New function.
4182 (vectorizable_call): Check internal_load_fn_p.
4183 (vectorizable_load): Likewise. Handle gather load internal
4185 (vectorizable_store): Update call to check_load_store_masking.
4186 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
4187 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
4188 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
4189 (aarch64_gather_scale_operand_d): New predicates.
4190 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
4191 (mask_gather_load<mode>): New insns.
4193 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4194 Alan Hayward <alan.hayward@arm.com>
4195 David Sherwood <david.sherwood@arm.com>
4197 * optabs.def (fold_left_plus_optab): New optab.
4198 * doc/md.texi (fold_left_plus_@var{m}): Document.
4199 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
4200 * internal-fn.c (fold_left_direct): Define.
4201 (expand_fold_left_optab_fn): Likewise.
4202 (direct_fold_left_optab_supported_p): Likewise.
4203 * fold-const-call.c (fold_const_fold_left): New function.
4204 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
4205 * tree-parloops.c (valid_reduction_p): New function.
4206 (gather_scalar_reductions): Use it.
4207 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
4208 (vect_finish_replace_stmt): Declare.
4209 * tree-vect-loop.c (fold_left_reduction_fn): New function.
4210 (needs_fold_left_reduction_p): New function, split out from...
4211 (vect_is_simple_reduction): ...here. Accept reductions that
4212 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
4213 (vect_force_simple_reduction): Also store the reduction type in
4214 the assignment's STMT_VINFO_REDUC_TYPE.
4215 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
4216 (merge_with_identity): New function.
4217 (vect_expand_fold_left): Likewise.
4218 (vectorize_fold_left_reduction): Likewise.
4219 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
4220 scalar phi in place for it. Check for target support and reject
4221 cases that would reassociate the operation. Defer the transform
4222 phase to vectorize_fold_left_reduction.
4223 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
4224 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
4225 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
4227 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4229 * tree-if-conv.c (predicate_mem_writes): Remove redundant
4230 call to ifc_temp_var.
4232 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4233 Alan Hayward <alan.hayward@arm.com>
4234 David Sherwood <david.sherwood@arm.com>
4236 * target.def (legitimize_address_displacement): Take the original
4237 offset as a poly_int.
4238 * targhooks.h (default_legitimize_address_displacement): Update
4240 * targhooks.c (default_legitimize_address_displacement): Likewise.
4241 * doc/tm.texi: Regenerate.
4242 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
4243 as an argument, moving assert of ad->disp == ad->disp_term to...
4244 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
4245 Try calling targetm.legitimize_address_displacement before expanding
4246 the address rather than afterwards, and adjust for the new interface.
4247 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
4248 Match the new hook interface. Handle SVE addresses.
4249 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
4252 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4254 * Makefile.in (OBJS): Add early-remat.o.
4255 * target.def (select_early_remat_modes): New hook.
4256 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
4257 * doc/tm.texi: Regenerate.
4258 * targhooks.h (default_select_early_remat_modes): Declare.
4259 * targhooks.c (default_select_early_remat_modes): New function.
4260 * timevar.def (TV_EARLY_REMAT): New timevar.
4261 * passes.def (pass_early_remat): New pass.
4262 * tree-pass.h (make_pass_early_remat): Declare.
4263 * early-remat.c: New file.
4264 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
4266 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
4268 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4269 Alan Hayward <alan.hayward@arm.com>
4270 David Sherwood <david.sherwood@arm.com>
4272 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
4273 vfm1 with a bound_epilog parameter.
4274 (vect_do_peeling): Update calls accordingly, and move the prologue
4275 call earlier in the function. Treat the base bound_epilog as 0 for
4276 fully-masked loops and retain vf - 1 for other loops. Add 1 to
4277 this base when peeling for gaps.
4278 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
4279 with fully-masked loops.
4280 (vect_estimate_min_profitable_iters): Handle the single peeled
4281 iteration in that case.
4283 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4284 Alan Hayward <alan.hayward@arm.com>
4285 David Sherwood <david.sherwood@arm.com>
4287 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
4288 single-element interleaving even if the size is not a power of 2.
4289 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
4290 accesses for single-element interleaving if the group size is
4293 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4294 Alan Hayward <alan.hayward@arm.com>
4295 David Sherwood <david.sherwood@arm.com>
4297 * doc/md.texi (fold_extract_last_@var{m}): Document.
4298 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
4299 * optabs.def (fold_extract_last_optab): New optab.
4300 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
4301 * internal-fn.c (fold_extract_direct): New macro.
4302 (expand_fold_extract_optab_fn): Likewise.
4303 (direct_fold_extract_optab_supported_p): Likewise.
4304 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
4305 * tree-vect-loop.c (vect_model_reduction_cost): Handle
4306 EXTRACT_LAST_REDUCTION.
4307 (get_initial_def_for_reduction): Do not create an initial vector
4308 for EXTRACT_LAST_REDUCTION reductions.
4309 (vectorizable_reduction): Leave the scalar phi in place for
4310 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
4311 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
4312 epilogue code for EXTRACT_LAST_REDUCTION and defer the
4313 transform phase to vectorizable_condition.
4314 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
4316 (vect_finish_stmt_generation): ...here.
4317 (vect_finish_replace_stmt): New function.
4318 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
4319 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
4321 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
4323 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4324 Alan Hayward <alan.hayward@arm.com>
4325 David Sherwood <david.sherwood@arm.com>
4327 * doc/md.texi (extract_last_@var{m}): Document.
4328 * optabs.def (extract_last_optab): New optab.
4329 * internal-fn.def (EXTRACT_LAST): New internal function.
4330 * internal-fn.c (cond_unary_direct): New macro.
4331 (expand_cond_unary_optab_fn): Likewise.
4332 (direct_cond_unary_optab_supported_p): Likewise.
4333 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
4334 loops using EXTRACT_LAST.
4335 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
4336 (extract_last_<mode>): ...this optab.
4337 (vec_extract<mode><Vel>): Update accordingly.
4339 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4340 Alan Hayward <alan.hayward@arm.com>
4341 David Sherwood <david.sherwood@arm.com>
4343 * target.def (empty_mask_is_expensive): New hook.
4344 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
4345 * doc/tm.texi: Regenerate.
4346 * targhooks.h (default_empty_mask_is_expensive): Declare.
4347 * targhooks.c (default_empty_mask_is_expensive): New function.
4348 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
4349 if the target says that empty masks are expensive.
4350 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
4352 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
4354 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4355 Alan Hayward <alan.hayward@arm.com>
4356 David Sherwood <david.sherwood@arm.com>
4358 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
4359 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
4360 (vect_use_loop_mask_for_alignment_p): New function.
4361 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
4362 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
4363 niters_skip argument. Make sure that the first niters_skip elements
4364 of the first iteration are inactive.
4365 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
4366 Update call to vect_set_loop_masks_directly.
4367 (get_misalign_in_elems): New function, split out from...
4368 (vect_gen_prolog_loop_niters): ...here.
4369 (vect_update_init_of_dr): Take a code argument that specifies whether
4370 the adjustment should be added or subtracted.
4371 (vect_update_init_of_drs): Likewise.
4372 (vect_prepare_for_masked_peels): New function.
4373 (vect_do_peeling): Skip prologue peeling if we're using a mask
4374 instead. Update call to vect_update_inits_of_drs.
4375 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4377 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
4378 alignment. Do not include the number of peeled iterations in
4379 the minimum threshold in that case.
4380 (vectorizable_induction): Adjust the start value down by
4381 LOOP_VINFO_MASK_SKIP_NITERS iterations.
4382 (vect_transform_loop): Call vect_prepare_for_masked_peels.
4383 Take the number of skipped iterations into account when calculating
4385 * tree-vect-stmts.c (vect_gen_while_not): New function.
4387 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4388 Alan Hayward <alan.hayward@arm.com>
4389 David Sherwood <david.sherwood@arm.com>
4391 * doc/sourcebuild.texi (vect_fully_masked): Document.
4392 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
4394 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
4396 (vect_analyze_loop_2): ...here. Don't check the vectorization
4397 factor against the number of loop iterations if the loop is
4400 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4401 Alan Hayward <alan.hayward@arm.com>
4402 David Sherwood <david.sherwood@arm.com>
4404 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
4405 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
4406 (dump_groups): Update accordingly.
4407 (iv_use::mem_type): New member variable.
4408 (address_p): New function.
4409 (record_use): Add a mem_type argument and initialize the new
4411 (record_group_use): Add a mem_type argument. Use address_p.
4412 Remove obsolete null checks of base_object. Update call to record_use.
4413 (find_interesting_uses_op): Update call to record_group_use.
4414 (find_interesting_uses_cond): Likewise.
4415 (find_interesting_uses_address): Likewise.
4416 (get_mem_type_for_internal_fn): New function.
4417 (find_address_like_use): Likewise.
4418 (find_interesting_uses_stmt): Try find_address_like_use before
4419 calling find_interesting_uses_op.
4420 (addr_offset_valid_p): Use the iv mem_type field as the type
4421 of the addressed memory.
4422 (add_autoinc_candidates): Likewise.
4423 (get_address_cost): Likewise.
4424 (split_small_address_groups_p): Use address_p.
4425 (split_address_groups): Likewise.
4426 (add_iv_candidate_for_use): Likewise.
4427 (autoinc_possible_for_pair): Likewise.
4428 (rewrite_groups): Likewise.
4429 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
4430 (determine_group_iv_cost): Update after split of USE_ADDRESS.
4431 (get_alias_ptr_type_for_ptr_address): New function.
4432 (rewrite_use_address): Rewrite address uses in calls that were
4433 identified by find_address_like_use.
4435 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4436 Alan Hayward <alan.hayward@arm.com>
4437 David Sherwood <david.sherwood@arm.com>
4439 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
4441 * gimple-expr.h (is_gimple_addressable: Likewise.
4442 * gimple-expr.c (is_gimple_address): Likewise.
4443 * internal-fn.c (expand_call_mem_ref): New function.
4444 (expand_mask_load_optab_fn): Use it.
4445 (expand_mask_store_optab_fn): Likewise.
4447 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4448 Alan Hayward <alan.hayward@arm.com>
4449 David Sherwood <david.sherwood@arm.com>
4451 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
4452 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
4453 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
4454 (cond_umax@var{mode}): Document.
4455 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
4456 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
4457 (cond_umin_optab, cond_umax_optab): New optabs.
4458 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
4459 (COND_IOR, COND_XOR): New internal functions.
4460 * internal-fn.h (get_conditional_internal_fn): Declare.
4461 * internal-fn.c (cond_binary_direct): New macro.
4462 (expand_cond_binary_optab_fn): Likewise.
4463 (direct_cond_binary_optab_supported_p): Likewise.
4464 (get_conditional_internal_fn): New function.
4465 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
4466 Cope with reduction statements that are vectorized as calls rather
4468 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
4469 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
4470 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
4471 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4472 (UNSPEC_COND_EOR): New unspecs.
4473 (optab): Add mappings for them.
4474 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
4475 (sve_int_op, sve_fp_op): New int attributes.
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 * optabs.def (while_ult_optab): New optab.
4482 * doc/md.texi (while_ult@var{m}@var{n}): Document.
4483 * internal-fn.def (WHILE_ULT): New internal function.
4484 * internal-fn.h (direct_internal_fn_supported_p): New override
4485 that takes two types as argument.
4486 * internal-fn.c (while_direct): New macro.
4487 (expand_while_optab_fn): New function.
4488 (convert_optab_supported_p): Likewise.
4489 (direct_while_optab_supported_p): New macro.
4490 * wide-int.h (wi::udiv_ceil): New function.
4491 * tree-vectorizer.h (rgroup_masks): New structure.
4492 (vec_loop_masks): New typedef.
4493 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
4495 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
4496 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
4497 (vect_max_vf): New function.
4498 (slpeel_make_loop_iterate_ntimes): Delete.
4499 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
4500 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
4501 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
4502 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
4503 internal-fn.h, stor-layout.h and optabs-query.h.
4504 (vect_set_loop_mask): New function.
4505 (add_preheader_seq): Likewise.
4506 (add_header_seq): Likewise.
4507 (interleave_supported_p): Likewise.
4508 (vect_maybe_permute_loop_masks): Likewise.
4509 (vect_set_loop_masks_directly): Likewise.
4510 (vect_set_loop_condition_masked): Likewise.
4511 (vect_set_loop_condition_unmasked): New function, split out from
4512 slpeel_make_loop_iterate_ntimes.
4513 (slpeel_make_loop_iterate_ntimes): Rename to..
4514 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
4515 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
4516 (vect_do_peeling): Update call accordingly.
4517 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
4519 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4520 mask_compare_type, can_fully_mask_p and fully_masked_p.
4521 (release_vec_loop_masks): New function.
4522 (_loop_vec_info): Use it to free the loop masks.
4523 (can_produce_all_loop_masks_p): New function.
4524 (vect_get_max_nscalars_per_iter): Likewise.
4525 (vect_verify_full_masking): Likewise.
4526 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
4527 retries, and free the mask rgroups before retrying. Check loop-wide
4528 reasons for disallowing fully-masked loops. Make the final decision
4529 about whether use a fully-masked loop or not.
4530 (vect_estimate_min_profitable_iters): Do not assume that peeling
4531 for the number of iterations will be needed for fully-masked loops.
4532 (vectorizable_reduction): Disable fully-masked loops.
4533 (vectorizable_live_operation): Likewise.
4534 (vect_halve_mask_nunits): New function.
4535 (vect_double_mask_nunits): Likewise.
4536 (vect_record_loop_mask): Likewise.
4537 (vect_get_loop_mask): Likewise.
4538 (vect_transform_loop): Handle the case in which the final loop
4539 iteration might handle a partial vector. Call vect_set_loop_condition
4540 instead of slpeel_make_loop_iterate_ntimes.
4541 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
4542 (check_load_store_masking): New function.
4543 (prepare_load_store_mask): Likewise.
4544 (vectorizable_store): Handle fully-masked loops.
4545 (vectorizable_load): Likewise.
4546 (supportable_widening_operation): Use vect_halve_mask_nunits for
4548 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
4549 (vect_gen_while): New function.
4550 * config/aarch64/aarch64.md (umax<mode>3): New expander.
4551 (aarch64_uqdec<mode>): New insn.
4553 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4554 Alan Hayward <alan.hayward@arm.com>
4555 David Sherwood <david.sherwood@arm.com>
4557 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
4558 (reduc_xor_scal_optab): New optabs.
4559 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
4560 (reduc_xor_scal_@var{m}): Document.
4561 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
4562 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
4564 * fold-const-call.c (fold_const_call): Handle them.
4565 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
4566 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
4567 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
4568 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
4569 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
4570 (UNSPEC_XORV): New unspecs.
4571 (optab): Add entries for them.
4572 (BITWISEV): New int iterator.
4573 (bit_reduc_op): New int attributes.
4575 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4576 Alan Hayward <alan.hayward@arm.com>
4577 David Sherwood <david.sherwood@arm.com>
4579 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
4580 * internal-fn.def (VEC_SHL_INSERT): New internal function.
4581 * optabs.def (vec_shl_insert_optab): New optab.
4582 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
4583 (duplicate_and_interleave): Likewise.
4584 * tree-vect-loop.c: Include internal-fn.h.
4585 (neutral_op_for_slp_reduction): New function, split out from
4586 get_initial_defs_for_reduction.
4587 (get_initial_def_for_reduction): Handle option 2 for variable-length
4588 vectors by loading the neutral value into a vector and then shifting
4589 the initial value into element 0.
4590 (get_initial_defs_for_reduction): Replace the code argument with
4591 the neutral value calculated by neutral_op_for_slp_reduction.
4592 Use gimple_build_vector for constant-length vectors.
4593 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
4594 but the first group_size elements have a neutral value.
4595 Use duplicate_and_interleave otherwise.
4596 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
4597 Update call to get_initial_defs_for_reduction. Handle SLP
4598 reductions for variable-length vectors by creating one vector
4599 result for each scalar result, with the elements associated
4600 with other scalar results stubbed out with the neutral value.
4601 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
4602 Require IFN_VEC_SHL_INSERT for double reductions on
4603 variable-length vectors, or SLP reductions that have
4604 a neutral value. Require can_duplicate_and_interleave_p
4605 support for variable-length unchained SLP reductions if there
4606 is no neutral value, such as for MIN/MAX reductions. Also require
4607 the number of vector elements to be a multiple of the number of
4608 SLP statements when doing variable-length unchained SLP reductions.
4609 Update call to vect_create_epilog_for_reduction.
4610 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
4611 and remove initial values.
4612 (duplicate_and_interleave): Make public.
4613 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
4614 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
4616 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4617 Alan Hayward <alan.hayward@arm.com>
4618 David Sherwood <david.sherwood@arm.com>
4620 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
4621 (can_duplicate_and_interleave_p): New function.
4622 (vect_get_and_check_slp_defs): Take the vector of statements
4623 rather than just the current one. Remove excess parentheses.
4624 Restriction rejectinon of vect_constant_def and vect_external_def
4625 for variable-length vectors to boolean types, or types for which
4626 can_duplicate_and_interleave_p is false.
4627 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
4628 (duplicate_and_interleave): New function.
4629 (vect_get_constant_vectors): Use gimple_build_vector for
4630 constant-length vectors and suitable variable-length constant
4631 vectors. Use duplicate_and_interleave for other variable-length
4632 vectors. Don't defer the update when inserting new statements.
4634 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4635 Alan Hayward <alan.hayward@arm.com>
4636 David Sherwood <david.sherwood@arm.com>
4638 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
4639 min_profitable_iters doesn't go negative.
4641 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4642 Alan Hayward <alan.hayward@arm.com>
4643 David Sherwood <david.sherwood@arm.com>
4645 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
4646 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
4647 * optabs.def (vec_mask_load_lanes_optab): New optab.
4648 (vec_mask_store_lanes_optab): Likewise.
4649 * internal-fn.def (MASK_LOAD_LANES): New internal function.
4650 (MASK_STORE_LANES): Likewise.
4651 * internal-fn.c (mask_load_lanes_direct): New macro.
4652 (mask_store_lanes_direct): Likewise.
4653 (expand_mask_load_optab_fn): Handle masked operations.
4654 (expand_mask_load_lanes_optab_fn): New macro.
4655 (expand_mask_store_optab_fn): Handle masked operations.
4656 (expand_mask_store_lanes_optab_fn): New macro.
4657 (direct_mask_load_lanes_optab_supported_p): Likewise.
4658 (direct_mask_store_lanes_optab_supported_p): Likewise.
4659 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
4661 (vect_load_lanes_supported): Likewise.
4662 * tree-vect-data-refs.c (strip_conversion): New function.
4663 (can_group_stmts_p): Likewise.
4664 (vect_analyze_data_ref_accesses): Use it instead of checking
4665 for a pair of assignments.
4666 (vect_store_lanes_supported): Take a masked_p parameter.
4667 (vect_load_lanes_supported): Likewise.
4668 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
4669 vect_store_lanes_supported and vect_load_lanes_supported.
4670 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4671 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
4672 parameter. Don't allow gaps for masked accesses.
4673 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
4674 and vect_load_lanes_supported.
4675 (get_load_store_type): Take a masked_p parameter and update
4676 call to get_group_load_store_type.
4677 (vectorizable_store): Update call to get_load_store_type.
4678 Handle IFN_MASK_STORE_LANES.
4679 (vectorizable_load): Update call to get_load_store_type.
4680 Handle IFN_MASK_LOAD_LANES.
4682 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4683 Alan Hayward <alan.hayward@arm.com>
4684 David Sherwood <david.sherwood@arm.com>
4686 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
4688 * config/aarch64/aarch64-protos.h
4689 (aarch64_sve_struct_memory_operand_p): Declare.
4690 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
4691 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
4692 (VPRED, vpred): Handle SVE structure modes.
4693 * config/aarch64/constraints.md (Utx): New constraint.
4694 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
4695 (aarch64_sve_struct_nonimmediate_operand): New predicates.
4696 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
4697 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
4698 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
4699 structure modes. Split into pieces after RA.
4700 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
4701 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
4703 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
4704 SVE structure modes.
4705 (aarch64_classify_address): Likewise.
4706 (sizetochar): Move earlier in file.
4707 (aarch64_print_operand): Handle SVE register lists.
4708 (aarch64_array_mode): New function.
4709 (aarch64_sve_struct_memory_operand_p): Likewise.
4710 (TARGET_ARRAY_MODE): Redefine.
4712 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4713 Alan Hayward <alan.hayward@arm.com>
4714 David Sherwood <david.sherwood@arm.com>
4716 * target.def (array_mode): New target hook.
4717 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
4718 * doc/tm.texi: Regenerate.
4719 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
4720 * hooks.c (hook_optmode_mode_uhwi_none): New function.
4721 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
4723 * stor-layout.c (mode_for_array): Likewise. Support polynomial
4726 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4727 Alan Hayward <alan.hayward@arm.com>
4728 David Sherwood <david.sherwood@arm.com>
4730 * fold-const.c (fold_binary_loc): Check the argument types
4731 rather than the result type when testing for a vector operation.
4733 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4735 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
4736 * doc/tm.texi: Regenerate.
4738 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4739 Alan Hayward <alan.hayward@arm.com>
4740 David Sherwood <david.sherwood@arm.com>
4742 * doc/invoke.texi (-msve-vector-bits=): Document new option.
4743 (sve): Document new AArch64 extension.
4744 * doc/md.texi (w): Extend the description of the AArch64
4745 constraint to include SVE vectors.
4746 (Upl, Upa): Document new AArch64 predicate constraints.
4747 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
4749 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
4750 (msve-vector-bits=): New option.
4751 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
4752 SVE when these are disabled.
4753 (sve): New extension.
4754 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
4755 modes. Adjust their number of units based on aarch64_sve_vg.
4756 (MAX_BITSIZE_MODE_ANY_MODE): Define.
4757 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
4758 aarch64_addr_query_type.
4759 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
4760 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
4761 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
4762 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
4763 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
4764 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
4765 (aarch64_simd_imm_zero_p): Delete.
4766 (aarch64_check_zero_based_sve_index_immediate): Declare.
4767 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4768 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4769 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4770 (aarch64_sve_float_mul_immediate_p): Likewise.
4771 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4773 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
4774 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
4775 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
4776 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
4777 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
4778 (aarch64_regmode_natural_size): Likewise.
4779 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
4780 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
4782 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
4783 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
4784 for VG and the SVE predicate registers.
4785 (V_ALIASES): Add a "z"-prefixed alias.
4786 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
4787 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
4788 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
4789 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
4790 (REG_CLASS_NAMES): Add entries for them.
4791 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
4792 and the predicate registers.
4793 (aarch64_sve_vg): Declare.
4794 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
4795 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
4796 (REGMODE_NATURAL_SIZE): Define.
4797 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
4799 * config/aarch64/aarch64.c: Include cfgrtl.h.
4800 (simd_immediate_info): Add a constructor for series vectors,
4801 and an associated step field.
4802 (aarch64_sve_vg): New variable.
4803 (aarch64_dbx_register_number): Handle VG and the predicate registers.
4804 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
4805 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
4806 (VEC_ANY_DATA, VEC_STRUCT): New constants.
4807 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
4808 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
4809 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
4810 (aarch64_get_mask_mode): New functions.
4811 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
4812 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
4813 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
4814 predicate modes and predicate registers. Explicitly restrict
4815 GPRs to modes of 16 bytes or smaller. Only allow FP registers
4816 to store a vector mode if it is recognized by
4817 aarch64_classify_vector_mode.
4818 (aarch64_regmode_natural_size): New function.
4819 (aarch64_hard_regno_caller_save_mode): Return the original mode
4821 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
4822 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
4823 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
4824 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
4826 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
4827 does not overlap dest if the function is frame-related. Handle
4829 (aarch64_split_add_offset): New function.
4830 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
4831 them aarch64_add_offset.
4832 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
4833 and update call to aarch64_sub_sp.
4834 (aarch64_add_cfa_expression): New function.
4835 (aarch64_expand_prologue): Pass extra temporary registers to the
4836 functions above. Handle the case in which we need to emit new
4837 DW_CFA_expressions for registers that were originally saved
4838 relative to the stack pointer, but now have to be expressed
4839 relative to the frame pointer.
4840 (aarch64_output_mi_thunk): Pass extra temporary registers to the
4842 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
4843 IP0 and IP1 values for SVE frames.
4844 (aarch64_expand_vec_series): New function.
4845 (aarch64_expand_sve_widened_duplicate): Likewise.
4846 (aarch64_expand_sve_const_vector): Likewise.
4847 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
4848 Handle SVE constants. Use emit_move_insn to move a force_const_mem
4849 into the register, rather than emitting a SET directly.
4850 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
4851 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
4852 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
4853 (offset_9bit_signed_scaled_p): New functions.
4854 (aarch64_replicate_bitmask_imm): New function.
4855 (aarch64_bitmask_imm): Use it.
4856 (aarch64_cannot_force_const_mem): Reject expressions involving
4857 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
4858 (aarch64_classify_index): Handle SVE indices, by requiring
4859 a plain register index with a scale that matches the element size.
4860 (aarch64_classify_address): Handle SVE addresses. Assert that
4861 the mode of the address is VOIDmode or an integer mode.
4862 Update call to aarch64_classify_symbol.
4863 (aarch64_classify_symbolic_expression): Update call to
4864 aarch64_classify_symbol.
4865 (aarch64_const_vec_all_in_range_p): New function.
4866 (aarch64_print_vector_float_operand): Likewise.
4867 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
4868 "vN" for FP registers with SVE modes. Handle (const ...) vectors
4869 and the FP immediates 1.0 and 0.5.
4870 (aarch64_print_address_internal): Handle SVE addresses.
4871 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
4872 (aarch64_regno_regclass): Handle predicate registers.
4873 (aarch64_secondary_reload): Handle big-endian reloads of SVE
4875 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
4876 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
4877 (aarch64_convert_sve_vector_bits): New function.
4878 (aarch64_override_options): Use it to handle -msve-vector-bits=.
4879 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4881 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
4882 Handle SVE vector and predicate modes. Accept VL-based constants
4883 that need only one temporary register, and VL offsets that require
4884 no temporary registers.
4885 (aarch64_conditional_register_usage): Mark the predicate registers
4886 as fixed if SVE isn't available.
4887 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
4888 Return true for SVE vector and predicate modes.
4889 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
4890 rather than an unsigned int. Handle SVE modes.
4891 (aarch64_preferred_simd_mode): Update call accordingly. Handle
4893 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
4895 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4896 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4897 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4898 (aarch64_sve_float_mul_immediate_p): New functions.
4899 (aarch64_sve_valid_immediate): New function.
4900 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
4901 Explicitly reject structure modes. Check for INDEX constants.
4902 Handle PTRUE and PFALSE constants.
4903 (aarch64_check_zero_based_sve_index_immediate): New function.
4904 (aarch64_simd_imm_zero_p): Delete.
4905 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
4906 vector modes. Accept constants in the range of CNT[BHWD].
4907 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
4908 ask for an Advanced SIMD mode.
4909 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
4910 (aarch64_simd_vector_alignment): Handle SVE predicates.
4911 (aarch64_vectorize_preferred_vector_alignment): New function.
4912 (aarch64_simd_vector_alignment_reachable): Use it instead of
4914 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
4915 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
4917 (MAX_VECT_LEN): Delete.
4918 (expand_vec_perm_d): Add a vec_flags field.
4919 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
4920 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
4921 (aarch64_evpc_ext): Don't apply a big-endian lane correction
4923 (aarch64_evpc_rev): Rename to...
4924 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
4925 (aarch64_evpc_rev_global): New function.
4926 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
4927 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
4929 (aarch64_evpc_sve_tbl): New function.
4930 (aarch64_expand_vec_perm_const_1): Update after rename of
4931 aarch64_evpc_rev. Handle SVE permutes too, trying
4932 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
4933 than aarch64_evpc_tbl.
4934 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
4935 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
4936 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
4937 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
4938 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
4939 (aarch64_expand_sve_vcond): New functions.
4940 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
4941 of aarch64_vector_mode_p.
4942 (aarch64_dwarf_poly_indeterminate_value): New function.
4943 (aarch64_compute_pressure_classes): Likewise.
4944 (aarch64_can_change_mode_class): Likewise.
4945 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
4946 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
4947 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
4948 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
4949 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
4950 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
4951 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
4952 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
4954 (Dn, Dl, Dr): Accept const as well as const_vector.
4955 (Dz): Likewise. Compare against CONST0_RTX.
4956 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
4957 of "vector" where appropriate.
4958 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
4959 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
4960 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
4961 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
4962 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
4963 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
4964 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
4965 (v_int_equiv): Extend to SVE modes.
4966 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
4968 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
4969 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
4970 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
4971 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
4972 (SVE_COND_FP_CMP): New int iterators.
4973 (perm_hilo): Handle the new unpack unspecs.
4974 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
4976 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
4977 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
4978 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
4979 (aarch64_equality_operator, aarch64_constant_vector_operand)
4980 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
4981 (aarch64_sve_nonimmediate_operand): Likewise.
4982 (aarch64_sve_general_operand): Likewise.
4983 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
4984 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
4985 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
4986 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
4987 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
4988 (aarch64_sve_float_arith_immediate): Likewise.
4989 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
4990 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
4991 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
4992 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
4993 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
4994 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
4995 (aarch64_sve_float_arith_operand): Likewise.
4996 (aarch64_sve_float_arith_with_sub_operand): Likewise.
4997 (aarch64_sve_float_mul_operand): Likewise.
4998 (aarch64_sve_vec_perm_operand): Likewise.
4999 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
5000 (aarch64_mov_operand): Accept const_poly_int and const_vector.
5001 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
5002 as well as const_vector.
5003 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
5004 in file. Use CONST0_RTX and CONSTM1_RTX.
5005 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
5006 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
5007 Use aarch64_simd_imm_zero.
5008 * config/aarch64/aarch64-sve.md: New file.
5009 * config/aarch64/aarch64.md: Include it.
5010 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
5011 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
5012 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
5013 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
5014 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
5015 (sve): New attribute.
5016 (enabled): Disable instructions with the sve attribute unless
5018 (movqi, movhi): Pass CONST_POLY_INT operaneds through
5019 aarch64_expand_mov_immediate.
5020 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
5021 CNT[BHSD] immediates.
5022 (movti): Split CONST_POLY_INT moves into two halves.
5023 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
5024 Split additions that need a temporary here if the destination
5025 is the stack pointer.
5026 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
5027 (*add<mode>3_poly_1): New instruction.
5028 (set_clobber_cc): New expander.
5030 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5032 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
5033 parameter and use it instead of GET_MODE_SIZE (innermode). Use
5034 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
5035 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
5036 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
5037 Change innermode from fixed_mode_size to machine_mode.
5038 (simplify_subreg): Update call accordingly. Handle a constant-sized
5039 subreg of a variable-length CONST_VECTOR.
5041 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5042 Alan Hayward <alan.hayward@arm.com>
5043 David Sherwood <david.sherwood@arm.com>
5045 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
5046 (add_offset_to_base): New function, split out from...
5047 (create_mem_ref): ...here. When handling a scale other than 1,
5048 check first whether the address is valid without the offset.
5049 Add it into the base if so, leaving the index and scale as-is.
5051 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5054 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
5055 fold_for_warn before checking if arg2 is INTEGER_CST.
5057 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
5059 * config/rs6000/predicates.md (load_multiple_operation): Delete.
5060 (store_multiple_operation): Delete.
5061 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
5062 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
5063 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
5064 guarded by TARGET_STRING.
5065 (rs6000_output_load_multiple): Delete.
5066 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
5067 OPTION_MASK_STRING / TARGET_STRING handling.
5068 (print_operand) <'N', 'O'>: Add comment that these are unused now.
5069 (const rs6000_opt_masks) <"string">: Change mask to 0.
5070 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
5071 (MASK_STRING): Delete.
5072 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
5074 (load_multiple): Delete.
5081 (store_multiple): Delete.
5088 (movmemsi_8reg): Delete.
5089 (corresponding unnamed define_insn): Delete.
5090 (movmemsi_6reg): Delete.
5091 (corresponding unnamed define_insn): Delete.
5092 (movmemsi_4reg): Delete.
5093 (corresponding unnamed define_insn): Delete.
5094 (movmemsi_2reg): Delete.
5095 (corresponding unnamed define_insn): Delete.
5096 (movmemsi_1reg): Delete.
5097 (corresponding unnamed define_insn): Delete.
5098 * config/rs6000/rs6000.opt (mno-string): New.
5099 (mstring): Replace by deprecation warning stub.
5100 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
5102 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5104 * regrename.c (regrename_do_replace): If replacing the same
5105 reg multiple times, try to reuse last created gen_raw_REG.
5108 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
5109 main to workaround a bug in GDB.
5111 2018-01-12 Tom de Vries <tom@codesourcery.com>
5114 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
5116 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
5118 PR rtl-optimization/80481
5119 * ira-color.c (get_cap_member): New function.
5120 (allocnos_conflict_by_live_ranges_p): Use it.
5121 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
5122 (setup_slot_coalesced_allocno_live_ranges): Ditto.
5124 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
5127 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
5128 (*saddl_se_1): Ditto.
5130 (*ssubl_se_1): Ditto.
5132 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5134 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
5135 rather than wi::to_widest for DR_INITs.
5136 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
5137 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
5138 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
5140 (vect_analyze_group_access_1): Note that here.
5142 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5144 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
5145 polynomial type sizes.
5147 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5149 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
5150 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
5151 (gimple_add_tmp_var): Likewise.
5153 2018-01-12 Martin Liska <mliska@suse.cz>
5155 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
5156 (gimple_alloc_sizes): Likewise.
5157 (dump_gimple_statistics): Use PRIu64 in printf format.
5158 * gimple.h: Change uint64_t to int.
5160 2018-01-12 Martin Liska <mliska@suse.cz>
5162 * tree-core.h: Use uint64_t instead of int.
5163 * tree.c (tree_node_counts): Likewise.
5164 (tree_node_sizes): Likewise.
5165 (dump_tree_statistics): Use PRIu64 in printf format.
5167 2018-01-12 Martin Liska <mliska@suse.cz>
5169 * Makefile.in: As qsort_chk is implemented in vec.c, add
5170 vec.o to linkage of gencfn-macros.
5171 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
5172 passing the info to record_node_allocation_statistics.
5173 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
5175 * ggc-common.c (struct ggc_usage): Add operator== and use
5176 it in operator< and compare function.
5177 * mem-stats.h (struct mem_usage): Likewise.
5178 * vec.c (struct vec_usage): Remove operator< and compare
5179 function. Can be simply inherited.
5181 2018-01-12 Martin Jambor <mjambor@suse.cz>
5184 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
5185 * tree-ssa-math-opts.c: Include domwalk.h.
5186 (convert_mult_to_fma_1): New function.
5187 (fma_transformation_info): New type.
5188 (fma_deferring_state): Likewise.
5189 (cancel_fma_deferring): New function.
5190 (result_of_phi): Likewise.
5191 (last_fma_candidate_feeds_initial_phi): Likewise.
5192 (convert_mult_to_fma): Added deferring logic, split actual
5193 transformation to convert_mult_to_fma_1.
5194 (math_opts_dom_walker): New type.
5195 (math_opts_dom_walker::after_dom_children): New method, body moved
5196 here from pass_optimize_widening_mul::execute, added deferring logic
5198 (pass_optimize_widening_mul::execute): Moved most of code to
5199 math_opts_dom_walker::after_dom_children.
5200 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
5201 * config/i386/i386.c (ix86_option_override_internal): Added
5202 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
5204 2018-01-12 Richard Biener <rguenther@suse.de>
5207 * dwarf2out.c (gen_variable_die): Do not reset old_die for
5208 inline instance vars.
5210 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
5213 * config/rx/rx.c (rx_is_restricted_memory_address):
5216 2018-01-12 Richard Biener <rguenther@suse.de>
5218 PR tree-optimization/80846
5219 * target.def (split_reduction): New target hook.
5220 * targhooks.c (default_split_reduction): New function.
5221 * targhooks.h (default_split_reduction): Declare.
5222 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
5223 target requests first reduce vectors by combining low and high
5225 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
5226 (get_vectype_for_scalar_type_and_size): Export.
5227 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
5228 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
5229 * doc/tm.texi: Regenerate.
5230 * config/i386/i386.c (ix86_split_reduction): Implement
5231 TARGET_VECTORIZE_SPLIT_REDUCTION.
5233 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5236 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
5237 in PIC mode except for TARGET_VXWORKS_RTP.
5238 * config/sparc/sparc.c: Include cfgrtl.h.
5239 (TARGET_INIT_PIC_REG): Define.
5240 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
5241 (sparc_pic_register_p): New predicate.
5242 (sparc_legitimate_address_p): Use it.
5243 (sparc_legitimize_pic_address): Likewise.
5244 (sparc_delegitimize_address): Likewise.
5245 (sparc_mode_dependent_address_p): Likewise.
5246 (gen_load_pcrel_sym): Remove 4th parameter.
5247 (load_got_register): Adjust call to above. Remove obsolete stuff.
5248 (sparc_expand_prologue): Do not call load_got_register here.
5249 (sparc_flat_expand_prologue): Likewise.
5250 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
5251 (sparc_use_pseudo_pic_reg): New function.
5252 (sparc_init_pic_reg): Likewise.
5253 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5254 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
5256 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
5258 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
5259 Add item for branch_cost.
5261 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5263 PR rtl-optimization/83565
5264 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
5265 not extend the result to a larger mode for rotate operations.
5266 (num_sign_bit_copies1): Likewise.
5268 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5271 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
5273 Use values-Xc.o for -pedantic.
5274 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
5276 2018-01-12 Martin Liska <mliska@suse.cz>
5279 * ipa-devirt.c (final_warning_record::grow_type_warnings):
5281 (possible_polymorphic_call_targets): Use it.
5282 (ipa_devirt): Likewise.
5284 2018-01-12 Martin Liska <mliska@suse.cz>
5286 * profile-count.h (enum profile_quality): Use 0 as invalid
5287 enum value of profile_quality.
5289 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
5291 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
5292 -mext-string options.
5294 2018-01-12 Richard Biener <rguenther@suse.de>
5296 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
5297 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
5298 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
5300 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
5302 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5304 * configure.ac (--with-long-double-format): Add support for the
5305 configuration option to change the default long double format on
5307 * config.gcc (powerpc*-linux*-*): Likewise.
5308 * configure: Regenerate.
5309 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
5310 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
5311 used without modification.
5313 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5315 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
5316 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
5317 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
5318 MISC_BUILTIN_SPEC_BARRIER.
5319 (rs6000_init_builtins): Likewise.
5320 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
5322 (speculation_barrier): New define_insn.
5323 * doc/extend.texi: Document __builtin_speculation_barrier.
5325 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5328 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
5329 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
5330 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
5332 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
5333 integral modes instead of "ss" and "sd".
5334 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
5335 vectors with 32-bit and 64-bit elements.
5336 (vecdupssescalarmodesuffix): New mode attribute.
5337 (vec_dup<mode>): Use it.
5339 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
5342 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
5343 frame if argument is passed on stack.
5345 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5348 * ree.c (combine_reaching_defs): Optimize also
5349 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
5350 reg2=any_extend(exp); reg1=reg2;, formatting fix.
5352 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5355 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
5357 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5360 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
5361 after they are computed.
5363 2018-01-11 Bin Cheng <bin.cheng@arm.com>
5365 PR tree-optimization/83695
5366 * gimple-loop-linterchange.cc
5367 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
5368 reset cached scev information after interchange.
5369 (pass_linterchange::execute): Remove call to scev_reset_htab.
5371 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5373 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
5374 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
5375 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
5376 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
5377 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
5378 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
5379 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
5380 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
5381 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
5382 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
5383 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
5384 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
5385 (V_lane_reg): Likewise.
5386 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
5388 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
5389 (vfmal_lane_low<mode>_intrinsic,
5390 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
5391 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
5392 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
5393 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
5394 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
5395 vfmsl_lane_high<mode>_intrinsic): New define_insns.
5397 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5399 * config/arm/arm-cpus.in (fp16fml): New feature.
5400 (ALL_SIMD): Add fp16fml.
5401 (armv8.2-a): Add fp16fml as an option.
5402 (armv8.3-a): Likewise.
5403 (armv8.4-a): Add fp16fml as part of fp16.
5404 * config/arm/arm.h (TARGET_FP16FML): Define.
5405 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
5407 * config/arm/arm-modes.def (V2HF): Define.
5408 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
5409 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
5410 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
5411 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
5412 vfmsl_low, vfmsl_high): New set of builtins.
5413 * config/arm/iterators.md (PLUSMINUS): New code iterator.
5414 (vfml_op): New code attribute.
5415 (VFMLHALVES): New int iterator.
5416 (VFML, VFMLSEL): New mode attributes.
5417 (V_reg): Define mapping for V2HF.
5418 (V_hi, V_lo): New mode attributes.
5419 (VF_constraint): Likewise.
5420 (vfml_half, vfml_half_selector): New int attributes.
5421 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
5423 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
5424 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
5426 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
5427 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
5428 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
5429 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
5431 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
5432 Document new effective target and option set.
5434 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5436 * config/arm/arm-cpus.in (armv8_4): New feature.
5437 (ARMv8_4a): New fgroup.
5438 (armv8.4-a): New arch.
5439 * config/arm/arm-tables.opt: Regenerate.
5440 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
5441 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
5442 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
5443 Add matching rules for -march=armv8.4-a and extensions.
5444 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
5446 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
5449 * config/rx/rx.md (BW): New mode attribute.
5450 (sync_lock_test_and_setsi): Add mode suffix to insn output.
5452 2018-01-11 Richard Biener <rguenther@suse.de>
5454 PR tree-optimization/83435
5455 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
5456 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
5457 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
5459 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5460 Alan Hayward <alan.hayward@arm.com>
5461 David Sherwood <david.sherwood@arm.com>
5463 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
5465 (aarch64_classify_address): Initialize it. Track polynomial offsets.
5466 (aarch64_print_address_internal): Use it to check for a zero offset.
5468 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5469 Alan Hayward <alan.hayward@arm.com>
5470 David Sherwood <david.sherwood@arm.com>
5472 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
5473 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
5474 Return a poly_int64 rather than a HOST_WIDE_INT.
5475 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
5476 rather than a HOST_WIDE_INT.
5477 * config/aarch64/aarch64.h (aarch64_frame): Protect with
5478 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
5479 hard_fp_offset, frame_size, initial_adjust, callee_offset and
5480 final_offset from HOST_WIDE_INT to poly_int64.
5481 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
5482 to_constant when getting the number of units in an Advanced SIMD
5484 (aarch64_builtin_vectorized_function): Check for a constant number
5486 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
5488 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
5489 attribute instead of GET_MODE_NUNITS.
5490 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
5491 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
5492 GET_MODE_SIZE for fixed-size registers.
5493 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
5494 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
5495 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
5496 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
5497 (aarch64_print_operand, aarch64_print_address_internal)
5498 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
5499 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
5500 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
5501 Handle polynomial GET_MODE_SIZE.
5502 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
5503 wider than SImode without modification.
5504 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
5505 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
5506 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
5507 passing and returning SVE modes.
5508 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
5509 rather than GEN_INT.
5510 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
5511 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
5512 (aarch64_allocate_and_probe_stack_space): Likewise.
5513 (aarch64_layout_frame): Cope with polynomial offsets.
5514 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
5515 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
5517 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
5518 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
5519 poly_int64 rather than a HOST_WIDE_INT.
5520 (aarch64_get_separate_components, aarch64_process_components)
5521 (aarch64_expand_prologue, aarch64_expand_epilogue)
5522 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
5523 (aarch64_anchor_offset): New function, split out from...
5524 (aarch64_legitimize_address): ...here.
5525 (aarch64_builtin_vectorization_cost): Handle polynomial
5526 TYPE_VECTOR_SUBPARTS.
5527 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
5529 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
5530 number of elements from the PARALLEL rather than the mode.
5531 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
5532 rather than GET_MODE_BITSIZE.
5533 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
5534 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
5535 (aarch64_expand_vec_perm_const_1): Handle polynomial
5536 d->perm.length () and d->perm elements.
5537 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
5538 Apply to_constant to d->perm elements.
5539 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
5540 polynomial CONST_VECTOR_NUNITS.
5541 (aarch64_move_pointer): Take amount as a poly_int64 rather
5543 (aarch64_progress_pointer): Avoid temporary variable.
5544 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
5545 the mode attribute instead of GET_MODE.
5547 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5548 Alan Hayward <alan.hayward@arm.com>
5549 David Sherwood <david.sherwood@arm.com>
5551 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
5552 x exists before using it.
5553 (aarch64_add_constant_internal): Rename to...
5554 (aarch64_add_offset_1): ...this. Replace regnum with separate
5555 src and dest rtxes. Handle the case in which they're different,
5556 including when the offset is zero. Replace scratchreg with an rtx.
5557 Use 2 additions if there is no spare register into which we can
5558 move a 16-bit constant.
5559 (aarch64_add_constant): Delete.
5560 (aarch64_add_offset): Replace reg with separate src and dest
5561 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
5562 Use aarch64_add_offset_1.
5563 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
5564 an rtx rather than an int. Take the delta as a poly_int64
5565 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
5566 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
5567 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
5568 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
5569 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
5571 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
5572 aarch64_add_constant.
5574 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5576 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
5577 Use scalar_float_mode.
5579 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5581 * config/aarch64/aarch64-simd.md
5582 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
5583 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
5584 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
5585 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
5586 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
5587 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
5588 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
5589 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
5590 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
5591 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
5593 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5596 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
5597 targ_options->x_arm_arch_string is non NULL.
5599 2018-01-11 Tamar Christina <tamar.christina@arm.com>
5601 * config/aarch64/aarch64.h
5602 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
5604 2018-01-11 Sudakshina Das <sudi.das@arm.com>
5607 * expmed.c (emit_store_flag_force): Swap if const op0
5608 and change VOIDmode to mode of op0.
5610 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5612 PR rtl-optimization/83761
5613 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
5614 than bytes to mode_for_size.
5616 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5619 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
5620 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
5623 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5626 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
5627 when in layout mode.
5628 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
5629 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
5632 2018-01-10 Michael Collison <michael.collison@arm.com>
5634 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
5635 * config/aarch64/aarch64-option-extension.def: Add
5636 AARCH64_OPT_EXTENSION of 'fp16fml'.
5637 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5638 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
5639 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
5640 * config/aarch64/constraints.md (Ui7): New constraint.
5641 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
5642 (VFMLA_SEL_W): Ditto.
5645 (VFMLA16_LOW): New int iterator.
5646 (VFMLA16_HIGH): Ditto.
5647 (UNSPEC_FMLAL): New unspec.
5648 (UNSPEC_FMLSL): Ditto.
5649 (UNSPEC_FMLAL2): Ditto.
5650 (UNSPEC_FMLSL2): Ditto.
5651 (f16mac): New code attribute.
5652 * config/aarch64/aarch64-simd-builtins.def
5653 (aarch64_fmlal_lowv2sf): Ditto.
5654 (aarch64_fmlsl_lowv2sf): Ditto.
5655 (aarch64_fmlalq_lowv4sf): Ditto.
5656 (aarch64_fmlslq_lowv4sf): Ditto.
5657 (aarch64_fmlal_highv2sf): Ditto.
5658 (aarch64_fmlsl_highv2sf): Ditto.
5659 (aarch64_fmlalq_highv4sf): Ditto.
5660 (aarch64_fmlslq_highv4sf): Ditto.
5661 (aarch64_fmlal_lane_lowv2sf): Ditto.
5662 (aarch64_fmlsl_lane_lowv2sf): Ditto.
5663 (aarch64_fmlal_laneq_lowv2sf): Ditto.
5664 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
5665 (aarch64_fmlalq_lane_lowv4sf): Ditto.
5666 (aarch64_fmlsl_lane_lowv4sf): Ditto.
5667 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
5668 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
5669 (aarch64_fmlal_lane_highv2sf): Ditto.
5670 (aarch64_fmlsl_lane_highv2sf): Ditto.
5671 (aarch64_fmlal_laneq_highv2sf): Ditto.
5672 (aarch64_fmlsl_laneq_highv2sf): Ditto.
5673 (aarch64_fmlalq_lane_highv4sf): Ditto.
5674 (aarch64_fmlsl_lane_highv4sf): Ditto.
5675 (aarch64_fmlalq_laneq_highv4sf): Ditto.
5676 (aarch64_fmlsl_laneq_highv4sf): Ditto.
5677 * config/aarch64/aarch64-simd.md:
5678 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
5679 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5680 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
5681 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5682 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
5683 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
5684 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
5685 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
5686 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
5687 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
5688 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
5689 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
5690 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
5691 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
5692 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
5693 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
5694 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
5695 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
5696 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
5697 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
5698 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
5699 (vfmlsl_low_u32): Ditto.
5700 (vfmlalq_low_u32): Ditto.
5701 (vfmlslq_low_u32): Ditto.
5702 (vfmlal_high_u32): Ditto.
5703 (vfmlsl_high_u32): Ditto.
5704 (vfmlalq_high_u32): Ditto.
5705 (vfmlslq_high_u32): Ditto.
5706 (vfmlal_lane_low_u32): Ditto.
5707 (vfmlsl_lane_low_u32): Ditto.
5708 (vfmlal_laneq_low_u32): Ditto.
5709 (vfmlsl_laneq_low_u32): Ditto.
5710 (vfmlalq_lane_low_u32): Ditto.
5711 (vfmlslq_lane_low_u32): Ditto.
5712 (vfmlalq_laneq_low_u32): Ditto.
5713 (vfmlslq_laneq_low_u32): Ditto.
5714 (vfmlal_lane_high_u32): Ditto.
5715 (vfmlsl_lane_high_u32): Ditto.
5716 (vfmlal_laneq_high_u32): Ditto.
5717 (vfmlsl_laneq_high_u32): Ditto.
5718 (vfmlalq_lane_high_u32): Ditto.
5719 (vfmlslq_lane_high_u32): Ditto.
5720 (vfmlalq_laneq_high_u32): Ditto.
5721 (vfmlslq_laneq_high_u32): Ditto.
5722 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
5723 (AARCH64_FL_FOR_ARCH8_4): New.
5724 (AARCH64_ISA_F16FML): New ISA flag.
5725 (TARGET_F16FML): New feature flag for fp16fml.
5726 (doc/invoke.texi): Document new fp16fml option.
5728 2018-01-10 Michael Collison <michael.collison@arm.com>
5730 * config/aarch64/aarch64-builtins.c:
5731 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
5732 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5733 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
5734 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
5735 (AARCH64_ISA_SHA3): New ISA flag.
5736 (TARGET_SHA3): New feature flag for sha3.
5737 * config/aarch64/iterators.md (sha512_op): New int attribute.
5738 (CRYPTO_SHA512): New int iterator.
5739 (UNSPEC_SHA512H): New unspec.
5740 (UNSPEC_SHA512H2): Ditto.
5741 (UNSPEC_SHA512SU0): Ditto.
5742 (UNSPEC_SHA512SU1): Ditto.
5743 * config/aarch64/aarch64-simd-builtins.def
5744 (aarch64_crypto_sha512hqv2di): New builtin.
5745 (aarch64_crypto_sha512h2qv2di): Ditto.
5746 (aarch64_crypto_sha512su0qv2di): Ditto.
5747 (aarch64_crypto_sha512su1qv2di): Ditto.
5748 (aarch64_eor3qv8hi): Ditto.
5749 (aarch64_rax1qv2di): Ditto.
5750 (aarch64_xarqv2di): Ditto.
5751 (aarch64_bcaxqv8hi): Ditto.
5752 * config/aarch64/aarch64-simd.md:
5753 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
5754 (aarch64_crypto_sha512su0qv2di): Ditto.
5755 (aarch64_crypto_sha512su1qv2di): Ditto.
5756 (aarch64_eor3qv8hi): Ditto.
5757 (aarch64_rax1qv2di): Ditto.
5758 (aarch64_xarqv2di): Ditto.
5759 (aarch64_bcaxqv8hi): Ditto.
5760 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
5761 (vsha512h2q_u64): Ditto.
5762 (vsha512su0q_u64): Ditto.
5763 (vsha512su1q_u64): Ditto.
5764 (veor3q_u16): Ditto.
5765 (vrax1q_u64): Ditto.
5767 (vbcaxq_u16): Ditto.
5768 * config/arm/types.md (crypto_sha512): New type attribute.
5769 (crypto_sha3): Ditto.
5770 (doc/invoke.texi): Document new sha3 option.
5772 2018-01-10 Michael Collison <michael.collison@arm.com>
5774 * config/aarch64/aarch64-builtins.c:
5775 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
5776 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5777 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
5778 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
5779 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
5780 (AARCH64_ISA_SM4): New ISA flag.
5781 (TARGET_SM4): New feature flag for sm4.
5782 * config/aarch64/aarch64-simd-builtins.def
5783 (aarch64_sm3ss1qv4si): Ditto.
5784 (aarch64_sm3tt1aq4si): Ditto.
5785 (aarch64_sm3tt1bq4si): Ditto.
5786 (aarch64_sm3tt2aq4si): Ditto.
5787 (aarch64_sm3tt2bq4si): Ditto.
5788 (aarch64_sm3partw1qv4si): Ditto.
5789 (aarch64_sm3partw2qv4si): Ditto.
5790 (aarch64_sm4eqv4si): Ditto.
5791 (aarch64_sm4ekeyqv4si): Ditto.
5792 * config/aarch64/aarch64-simd.md:
5793 (aarch64_sm3ss1qv4si): Ditto.
5794 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
5795 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
5796 (aarch64_sm4eqv4si): Ditto.
5797 (aarch64_sm4ekeyqv4si): Ditto.
5798 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
5799 (sm3part_op): Ditto.
5800 (CRYPTO_SM3TT): Ditto.
5801 (CRYPTO_SM3PART): Ditto.
5802 (UNSPEC_SM3SS1): New unspec.
5803 (UNSPEC_SM3TT1A): Ditto.
5804 (UNSPEC_SM3TT1B): Ditto.
5805 (UNSPEC_SM3TT2A): Ditto.
5806 (UNSPEC_SM3TT2B): Ditto.
5807 (UNSPEC_SM3PARTW1): Ditto.
5808 (UNSPEC_SM3PARTW2): Ditto.
5809 (UNSPEC_SM4E): Ditto.
5810 (UNSPEC_SM4EKEY): Ditto.
5811 * config/aarch64/constraints.md (Ui2): New constraint.
5812 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
5813 * config/arm/types.md (crypto_sm3): New type attribute.
5814 (crypto_sm4): Ditto.
5815 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
5816 (vsm3tt1aq_u32): Ditto.
5817 (vsm3tt1bq_u32): Ditto.
5818 (vsm3tt2aq_u32): Ditto.
5819 (vsm3tt2bq_u32): Ditto.
5820 (vsm3partw1q_u32): Ditto.
5821 (vsm3partw2q_u32): Ditto.
5822 (vsm4eq_u32): Ditto.
5823 (vsm4ekeyq_u32): Ditto.
5824 (doc/invoke.texi): Document new sm4 option.
5826 2018-01-10 Michael Collison <michael.collison@arm.com>
5828 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
5829 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
5830 (AARCH64_FL_FOR_ARCH8_4): New.
5831 (AARCH64_FL_V8_4): New flag.
5832 (doc/invoke.texi): Document new armv8.4-a option.
5834 2018-01-10 Michael Collison <michael.collison@arm.com>
5836 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5837 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
5838 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
5839 * config/aarch64/aarch64-option-extension.def: Add
5840 AARCH64_OPT_EXTENSION of 'sha2'.
5841 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
5842 (crypto): Disable sha2 and aes if crypto disabled.
5843 (crypto): Enable aes and sha2 if enabled.
5844 (simd): Disable sha2 and aes if simd disabled.
5845 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
5847 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
5848 (TARGET_SHA2): New feature flag for sha2.
5849 (TARGET_AES): New feature flag for aes.
5850 * config/aarch64/aarch64-simd.md:
5851 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
5852 conditional on TARGET_AES.
5853 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
5854 (aarch64_crypto_sha1hsi): Make pattern conditional
5856 (aarch64_crypto_sha1hv4si): Ditto.
5857 (aarch64_be_crypto_sha1hv4si): Ditto.
5858 (aarch64_crypto_sha1su1v4si): Ditto.
5859 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
5860 (aarch64_crypto_sha1su0v4si): Ditto.
5861 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
5862 (aarch64_crypto_sha256su0v4si): Ditto.
5863 (aarch64_crypto_sha256su1v4si): Ditto.
5864 (doc/invoke.texi): Document new aes and sha2 options.
5866 2018-01-10 Martin Sebor <msebor@redhat.com>
5868 PR tree-optimization/83781
5869 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
5872 2018-01-11 Martin Sebor <msebor@gmail.com>
5873 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5875 PR tree-optimization/83501
5876 PR tree-optimization/81703
5878 * tree-ssa-strlen.c (get_string_cst): Rename...
5879 (get_string_len): ...to this. Handle global constants.
5880 (handle_char_store): Adjust.
5882 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
5883 Jim Wilson <jimw@sifive.com>
5885 * config/riscv/riscv-protos.h (riscv_output_return): New.
5886 * config/riscv/riscv.c (struct machine_function): New naked_p field.
5887 (riscv_attribute_table, riscv_output_return),
5888 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
5889 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
5890 (riscv_compute_frame_info): Only compute frame->mask if not a naked
5892 (riscv_expand_prologue): Add early return for naked function.
5893 (riscv_expand_epilogue): Likewise.
5894 (riscv_function_ok_for_sibcall): Return false for naked function.
5895 (riscv_set_current_function): New.
5896 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
5897 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
5898 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
5899 * doc/extend.texi (RISC-V Function Attributes): New.
5901 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
5903 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
5904 check for 128-bit long double before checking TCmode.
5905 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
5906 128-bit long doubles before checking TFmode or TCmode.
5907 (FLOAT128_IBM_P): Likewise.
5909 2018-01-10 Martin Sebor <msebor@redhat.com>
5911 PR tree-optimization/83671
5912 * builtins.c (c_strlen): Unconditionally return zero for the empty
5914 Use -Warray-bounds for warnings.
5915 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
5916 for non-constant array indices with COMPONENT_REF, arrays of
5917 arrays, and pointers to arrays.
5918 (gimple_fold_builtin_strlen): Determine and set length range for
5919 non-constant character arrays.
5921 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
5924 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
5927 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
5929 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
5931 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
5934 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
5935 VECTOR_MEM_ALTIVEC_OR_VSX_P.
5936 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
5937 indexed_or_indirect_operand predicate.
5938 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
5939 (*vsx_le_perm_load_v8hi): Likewise.
5940 (*vsx_le_perm_load_v16qi): Likewise.
5941 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
5942 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
5943 (*vsx_le_perm_store_v8hi): Likewise.
5944 (*vsx_le_perm_store_v16qi): Likewise.
5945 (eight unnamed splitters): Likewise.
5947 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
5949 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
5950 * config/rs6000/emmintrin.h: Likewise.
5951 * config/rs6000/mmintrin.h: Likewise.
5952 * config/rs6000/xmmintrin.h: Likewise.
5954 2018-01-10 David Malcolm <dmalcolm@redhat.com>
5957 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
5959 * tree.c (tree_nop_conversion): Return true for location wrapper
5961 (maybe_wrap_with_location): New function.
5962 (selftest::check_strip_nops): New function.
5963 (selftest::test_location_wrappers): New function.
5964 (selftest::tree_c_tests): Call it.
5965 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
5966 (maybe_wrap_with_location): New decl.
5967 (EXPR_LOCATION_WRAPPER_P): New macro.
5968 (location_wrapper_p): New inline function.
5969 (tree_strip_any_location_wrapper): New inline function.
5971 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
5974 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
5975 stack_realign_offset for the largest alignment of stack slot
5977 (ix86_find_max_used_stack_alignment): New function.
5978 (ix86_finalize_stack_frame_flags): Use it. Set
5979 max_used_stack_alignment if we don't realign stack.
5980 * config/i386/i386.h (machine_function): Add
5981 max_used_stack_alignment.
5983 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
5985 * config/arm/arm.opt (-mbranch-cost): New option.
5986 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
5989 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
5992 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
5993 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
5995 2018-01-10 Richard Biener <rguenther@suse.de>
5998 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
5999 early out so it also covers the case where we have a non-NULL
6002 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6004 PR tree-optimization/83753
6005 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
6006 for non-strided grouped accesses if the number of elements is 1.
6008 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6011 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
6012 * i386.h (TARGET_USE_GATHER): Define.
6013 * x86-tune.def (X86_TUNE_USE_GATHER): New.
6015 2018-01-10 Martin Liska <mliska@suse.cz>
6018 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
6019 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
6021 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
6022 CLEANUP_NO_PARTITIONING is not set.
6024 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6026 * doc/rtl.texi: Remove documentation of (const ...) wrappers
6027 for vectors, as a partial revert of r254296.
6028 * rtl.h (const_vec_p): Delete.
6029 (const_vec_duplicate_p): Don't test for vector CONSTs.
6030 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
6031 * expmed.c (make_tree): Likewise.
6034 * common.md (E, F): Use CONSTANT_P instead of checking for
6036 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
6037 checking for CONST_VECTOR.
6039 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6042 * predict.c (force_edge_cold): Handle in more sane way edges
6045 2018-01-09 Carl Love <cel@us.ibm.com>
6047 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
6049 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
6050 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
6051 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
6052 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
6053 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
6054 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
6055 * config/rs6000/rs6000-protos.h: Add extern defition for
6056 rs6000_generate_float2_double_code.
6057 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
6059 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
6060 (float2_v2df): Add define_expand.
6062 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
6065 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
6066 op_mode in the force_to_mode call.
6068 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6070 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
6071 instead of checking each element individually.
6072 (aarch64_evpc_uzp): Likewise.
6073 (aarch64_evpc_zip): Likewise.
6074 (aarch64_evpc_ext): Likewise.
6075 (aarch64_evpc_rev): Likewise.
6076 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
6077 instead of checking each element individually. Return true without
6079 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
6080 whether all selected elements come from the same input, instead of
6081 checking each element individually. Remove calls to gen_rtx_REG,
6082 start_sequence and end_sequence and instead assert that no rtl is
6085 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6087 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
6088 order of HIGH and CONST checks.
6090 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6092 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
6093 if the destination isn't an SSA_NAME.
6095 2018-01-09 Richard Biener <rguenther@suse.de>
6097 PR tree-optimization/83668
6098 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
6100 (canonicalize_loop_form): ... here, renamed from ...
6101 (canonicalize_loop_closed_ssa_form): ... this and amended to
6102 swap successor edges for loop exit blocks to make us use
6103 the RPO order we need for initial schedule generation.
6105 2018-01-09 Joseph Myers <joseph@codesourcery.com>
6107 PR tree-optimization/64811
6108 * match.pd: When optimizing comparisons with Inf, avoid
6109 introducing or losing exceptions from comparisons with NaN.
6111 2018-01-09 Martin Liska <mliska@suse.cz>
6114 * asan.c (shadow_mem_size): Add gcc_assert.
6116 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
6118 Don't save registers in main().
6121 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
6122 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
6123 * config/avr/avr.c (avr_set_current_function): Don't error if
6124 naked, OS_task or OS_main are specified at the same time.
6125 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
6127 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
6129 * common/config/avr/avr-common.c (avr_option_optimization_table):
6130 Switch on -mmain-is-OS_task for optimizing compilations.
6132 2018-01-09 Richard Biener <rguenther@suse.de>
6134 PR tree-optimization/83572
6135 * graphite.c: Include cfganal.h.
6136 (graphite_transform_loops): Connect infinite loops to exit
6137 and remove fake edges at the end.
6139 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6141 * ipa-inline.c (edge_badness): Revert accidental checkin.
6143 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6146 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
6147 symbols; not inline clones.
6149 2018-01-09 Jakub Jelinek <jakub@redhat.com>
6152 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
6153 hard registers. Formatting fixes.
6155 PR preprocessor/83722
6156 * gcc.c (try_generate_repro): Pass
6157 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
6158 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
6161 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
6162 Kito Cheng <kito.cheng@gmail.com>
6164 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
6165 (riscv_leaf_function_p): Delete.
6166 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
6168 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6170 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
6172 (do_ifelse): New function.
6173 (do_isel): New function.
6174 (do_sub3): New function.
6175 (do_add3): New function.
6176 (do_load_mask_compare): New function.
6177 (do_overlap_load_compare): New function.
6178 (expand_compare_loop): New function.
6179 (expand_block_compare): Call expand_compare_loop() when appropriate.
6180 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
6182 (-mblock-compare-inline-loop-limit): New option.
6184 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6187 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
6188 Reverse order of second and third operands in first alternative.
6189 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
6190 of first and second elements in UNSPEC_VPERMR vector.
6191 (altivec_expand_vec_perm_le): Likewise.
6193 2018-01-08 Jeff Law <law@redhat.com>
6195 PR rtl-optimizatin/81308
6196 * tree-switch-conversion.c (cfg_altered): New file scoped static.
6197 (process_switch): If group_case_labels makes a change, then set
6199 (pass_convert_switch::execute): If a switch is converted, then
6200 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
6202 PR rtl-optimization/81308
6203 * recog.c (split_all_insns): Conditionally cleanup the CFG after
6206 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
6208 PR target/83663 - Revert r255946
6209 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
6210 generation for cases where splatting a value is not useful.
6211 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
6212 across a vec_duplicate and a paradoxical subreg forming a vector
6213 mode to a vec_concat.
6215 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6217 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
6218 -march=armv8.3-a variants.
6219 * config/arm/t-multilib: Likewise.
6220 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
6222 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6224 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
6226 (cceq_ior_compare_complement): Give it a name so I can use it, and
6227 change boolean_or_operator predicate to boolean_operator so it can
6228 be used to generate a crand.
6229 (eqne): New code iterator.
6230 (bd/bd_neg): New code_attrs.
6231 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
6232 a single define_insn.
6233 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
6234 decrement (bdnzt/bdnzf/bdzt/bdzf).
6235 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
6236 with the new names of the branch decrement patterns, and added the
6237 names of the branch decrement conditional patterns.
6239 2018-01-08 Richard Biener <rguenther@suse.de>
6241 PR tree-optimization/83563
6242 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
6245 2018-01-08 Richard Biener <rguenther@suse.de>
6248 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
6250 2018-01-08 Richard Biener <rguenther@suse.de>
6252 PR tree-optimization/83685
6253 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
6254 references to abnormals.
6256 2018-01-08 Richard Biener <rguenther@suse.de>
6259 * dwarf2out.c (output_indirect_strings): Handle empty
6260 skeleton_debug_str_hash.
6261 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
6263 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6265 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
6266 (emit_store_direct): Likewise.
6267 (arc_trampoline_adjust_address): Likewise.
6268 (arc_asm_trampoline_template): New function.
6269 (arc_initialize_trampoline): Use asm_trampoline_template.
6270 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
6271 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
6272 * config/arc/arc.md (flush_icache): Delete pattern.
6274 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6276 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
6277 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
6280 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6283 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
6284 by not USED_FOR_TARGET.
6285 (make_pass_resolve_sw_modes): Likewise.
6287 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6289 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
6292 2018-01-08 Richard Biener <rguenther@suse.de>
6295 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
6297 2018-01-08 Richard Biener <rguenther@suse.de>
6300 * match.pd ((t * 2) / 2) -> t): Add missing :c.
6302 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
6305 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
6306 basic blocks with a small number of successors.
6307 (convert_control_dep_chain_into_preds): Improve handling of
6309 (dump_predicates): Split apart into...
6310 (dump_pred_chain): ...here...
6311 (dump_pred_info): ...and here.
6312 (can_one_predicate_be_invalidated_p): Add debugging printfs.
6313 (can_chain_union_be_invalidated_p): Improve check for invalidation
6315 (uninit_uses_cannot_happen): Avoid unnecessary if
6316 convert_control_dep_chain_into_preds yielded nothing.
6318 2018-01-06 Martin Sebor <msebor@redhat.com>
6320 PR tree-optimization/83640
6321 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
6322 subtracting negative offset from size.
6323 (builtin_access::overlap): Adjust offset bounds of the access to fall
6324 within the size of the object if possible.
6326 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
6328 PR rtl-optimization/83699
6329 * expmed.c (extract_bit_field_1): Restrict the vector usage of
6330 extract_bit_field_as_subreg to cases in which the extracted
6331 value is also a vector.
6333 * lra-constraints.c (process_alt_operands): Test for the equivalence
6334 substitutions when detecting a possible reload cycle.
6336 2018-01-06 Jakub Jelinek <jakub@redhat.com>
6339 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
6340 by default if flag_selective_schedling{,2}. Formatting fixes.
6342 PR rtl-optimization/83682
6343 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
6344 if it has non-VECTOR_MODE element mode.
6345 (vec_duplicate_p): Likewise.
6348 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
6349 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
6351 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6354 * config/i386/i386-builtin.def
6355 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
6356 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
6357 Require also OPTION_MASK_ISA_AVX512F in addition to
6358 OPTION_MASK_ISA_GFNI.
6359 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
6360 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
6361 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
6362 to OPTION_MASK_ISA_GFNI.
6363 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
6364 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
6365 OPTION_MASK_ISA_AVX512BW.
6366 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
6367 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
6368 addition to OPTION_MASK_ISA_GFNI.
6369 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
6370 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
6371 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
6372 to OPTION_MASK_ISA_GFNI.
6373 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
6374 a requirement for all ISAs rather than any of them with a few
6376 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
6378 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
6379 bitmasks to be enabled with 3 exceptions, instead of requiring any
6380 enabled ISA with lots of exceptions.
6381 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
6382 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
6383 Change avx512bw in isa attribute to avx512f.
6384 * config/i386/sgxintrin.h: Add license boilerplate.
6385 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
6386 to __AVX512F__ and __AVX512VL to __AVX512VL__.
6387 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
6388 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
6390 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
6391 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
6392 temporarily sse2 rather than sse if not enabled already.
6395 * config/i386/sse.md (VI248_VLBW): Rename to ...
6396 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
6397 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
6398 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
6399 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
6400 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
6401 mode iterator instead of VI248_VLBW.
6403 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
6405 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
6406 (record_modified): Skip clobbers; add debug output.
6407 (param_change_prob): Use sreal frequencies.
6409 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6411 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
6412 punt for user-aligned variables.
6414 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6416 * tree-chrec.c (chrec_contains_symbols): Return true for
6419 2018-01-05 Sudakshina Das <sudi.das@arm.com>
6422 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
6423 of (x|y) == x for BICS pattern.
6425 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6427 PR tree-optimization/83605
6428 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
6429 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
6432 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
6434 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
6435 * config/epiphany/rtems.h: New file.
6437 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6438 Uros Bizjak <ubizjak@gmail.com>
6441 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
6442 QIreg_operand instead of register_operand predicate.
6443 * config/i386/i386.c (ix86_rop_should_change_byte_p,
6444 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
6445 comments instead of -fmitigate[-_]rop.
6447 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6450 * cgraphunit.c (symbol_table::compile): Switch to text_section
6451 before calling assembly_start debug hook.
6452 * run-rtl-passes.c (run_rtl_passes): Likewise.
6455 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6457 * tree-vrp.c (extract_range_from_binary_expr_1): Check
6458 range_int_cst_p rather than !symbolic_range_p before calling
6459 extract_range_from_multiplicative_op_1.
6461 2018-01-04 Jeff Law <law@redhat.com>
6463 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
6464 redundant test in assertion.
6466 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6468 * doc/rtl.texi: Document machine_mode wrapper classes.
6470 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6472 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
6475 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6477 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
6478 the VEC_PERM_EXPR fold to fail.
6480 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6483 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
6484 to switched_sections.
6486 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6489 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
6492 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
6495 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
6496 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
6498 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6501 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
6502 is BLKmode and bitpos not zero or mode change is needed.
6504 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6507 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
6510 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
6513 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
6514 instead of MULT rtx. Update all corresponding splitters.
6516 (*ssub<modesuffix>): Ditto.
6518 (*cmp_sadd_di): Update split patterns.
6519 (*cmp_sadd_si): Ditto.
6520 (*cmp_sadd_sidi): Ditto.
6521 (*cmp_ssub_di): Ditto.
6522 (*cmp_ssub_si): Ditto.
6523 (*cmp_ssub_sidi): Ditto.
6524 * config/alpha/predicates.md (const23_operand): New predicate.
6525 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
6526 Look for ASHIFT, not MULT inner operand.
6527 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
6529 2018-01-04 Martin Liska <mliska@suse.cz>
6531 PR gcov-profile/83669
6532 * gcov.c (output_intermediate_file): Add version to intermediate
6534 * doc/gcov.texi: Document new field 'version' in intermediate
6535 file format. Fix location of '-k' option of gcov command.
6537 2018-01-04 Martin Liska <mliska@suse.cz>
6540 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
6542 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6544 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
6546 2018-01-03 Martin Sebor <msebor@redhat.com>
6548 PR tree-optimization/83655
6549 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
6550 checking calls with invalid arguments.
6552 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6554 * tree-vect-stmts.c (vect_get_store_rhs): New function.
6555 (vectorizable_mask_load_store): Delete.
6556 (vectorizable_call): Return false for masked loads and stores.
6557 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
6558 instead of gimple_assign_rhs1.
6559 (vectorizable_load): Handle IFN_MASK_LOAD.
6560 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
6562 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6564 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
6566 (vectorizable_mask_load_store): ...here.
6567 (vectorizable_load): ...and here.
6569 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6571 * tree-vect-stmts.c (vect_build_all_ones_mask)
6572 (vect_build_zero_merge_argument): New functions, split out from...
6573 (vectorizable_load): ...here.
6575 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6577 * tree-vect-stmts.c (vect_check_store_rhs): New function,
6579 (vectorizable_mask_load_store): ...here.
6580 (vectorizable_store): ...and here.
6582 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6584 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
6586 (vectorizable_mask_load_store): ...here.
6588 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6590 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
6591 (vect_model_store_cost): Take a vec_load_store_type instead of a
6593 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
6594 (vect_model_store_cost): Take a vec_load_store_type instead of a
6596 (vectorizable_mask_load_store): Update accordingly.
6597 (vectorizable_store): Likewise.
6598 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
6600 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6602 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
6603 IFN_MASK_LOAD calls here rather than...
6604 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
6606 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6607 Alan Hayward <alan.hayward@arm.com>
6608 David Sherwood <david.sherwood@arm.com>
6610 * expmed.c (extract_bit_field_1): For vector extracts,
6611 fall back to extract_bit_field_as_subreg if vec_extract
6614 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6615 Alan Hayward <alan.hayward@arm.com>
6616 David Sherwood <david.sherwood@arm.com>
6618 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
6619 they are variable or constant sized.
6620 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
6621 slots for constant-sized data.
6623 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6624 Alan Hayward <alan.hayward@arm.com>
6625 David Sherwood <david.sherwood@arm.com>
6627 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
6628 handling COND_EXPRs with boolean comparisons, try to find a better
6629 basis for the mask type than the boolean itself.
6631 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6633 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
6634 is calculated and how it can be overridden.
6635 * genmodes.c (max_bitsize_mode_any_mode): New variable.
6636 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
6638 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
6641 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6642 Alan Hayward <alan.hayward@arm.com>
6643 David Sherwood <david.sherwood@arm.com>
6645 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
6646 Remove the mode argument.
6647 (aarch64_simd_valid_immediate): Remove the mode and inverse
6649 * config/aarch64/iterators.md (bitsize): New iterator.
6650 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
6651 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
6652 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
6653 aarch64_simd_valid_immediate.
6654 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
6655 (aarch64_reg_or_bic_imm): Likewise.
6656 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
6657 with an insn_type enum and msl with a modifier_type enum.
6658 Replace element_width with a scalar_mode. Change the shift
6659 to unsigned int. Add constructors for scalar_float_mode and
6660 scalar_int_mode elements.
6661 (aarch64_vect_float_const_representable_p): Delete.
6662 (aarch64_can_const_movi_rtx_p)
6663 (aarch64_simd_scalar_immediate_valid_for_move)
6664 (aarch64_simd_make_constant): Update call to
6665 aarch64_simd_valid_immediate.
6666 (aarch64_advsimd_valid_immediate_hs): New function.
6667 (aarch64_advsimd_valid_immediate): Likewise.
6668 (aarch64_simd_valid_immediate): Remove mode and inverse
6669 arguments. Rewrite to use the above. Use const_vec_duplicate_p
6670 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
6671 and aarch64_float_const_representable_p on the result.
6672 (aarch64_output_simd_mov_immediate): Remove mode argument.
6673 Update call to aarch64_simd_valid_immediate and use of
6674 simd_immediate_info.
6675 (aarch64_output_scalar_simd_mov_immediate): Update call
6678 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6679 Alan Hayward <alan.hayward@arm.com>
6680 David Sherwood <david.sherwood@arm.com>
6682 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
6683 (mode_nunits): Likewise CONST_MODE_NUNITS.
6684 * machmode.def (ADJUST_NUNITS): Document.
6685 * genmodes.c (mode_data::need_nunits_adj): New field.
6686 (blank_mode): Update accordingly.
6687 (adj_nunits): New variable.
6688 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
6690 (emit_mode_size_inline): Set need_bytesize_adj for all modes
6691 listed in adj_nunits.
6692 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
6693 listed in adj_nunits. Don't emit case statements for such modes.
6694 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
6695 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
6696 nothing if adj_nunits is nonnull.
6697 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
6698 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
6699 (emit_mode_fbit): Update use of print_maybe_const_decl.
6700 (emit_move_size): Likewise. Treat the array as non-const
6702 (emit_mode_adjustments): Handle adj_nunits.
6704 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6706 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
6707 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
6708 (VECTOR_MODES): Use it.
6709 (make_vector_modes): Take the prefix as an argument.
6711 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6712 Alan Hayward <alan.hayward@arm.com>
6713 David Sherwood <david.sherwood@arm.com>
6715 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
6716 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
6717 for MODE_VECTOR_BOOL.
6718 * machmode.def (VECTOR_BOOL_MODE): Document.
6719 * genmodes.c (VECTOR_BOOL_MODE): New macro.
6720 (make_vector_bool_mode): New function.
6721 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
6723 * lto-streamer-in.c (lto_input_mode_table): Likewise.
6724 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
6726 * stor-layout.c (int_mode_for_mode): Likewise.
6727 * tree.c (build_vector_type_for_mode): Likewise.
6728 * varasm.c (output_constant_pool_2): Likewise.
6729 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
6730 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
6731 for MODE_VECTOR_BOOL.
6732 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
6733 of mode class checks.
6734 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
6735 instead of a list of mode class checks.
6736 (expand_vector_scalar_condition): Likewise.
6737 (type_for_widest_vector_mode): Handle BImode as an inner mode.
6739 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6740 Alan Hayward <alan.hayward@arm.com>
6741 David Sherwood <david.sherwood@arm.com>
6743 * machmode.h (mode_size): Change from unsigned short to
6745 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
6746 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6747 or if measurement_type is not polynomial.
6748 (fixed_size_mode::includes_p): Check for constant-sized modes.
6749 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
6750 return a poly_uint16 rather than an unsigned short.
6751 (emit_mode_size): Change the type of mode_size from unsigned short
6752 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
6753 (emit_mode_adjustments): Cope with polynomial vector sizes.
6754 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6756 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6758 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
6759 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
6760 * caller-save.c (setup_save_areas): Likewise.
6761 (replace_reg_with_saved_mem): Likewise.
6762 * calls.c (emit_library_call_value_1): Likewise.
6763 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
6764 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
6765 (gen_lowpart_for_combine): Likewise.
6766 * convert.c (convert_to_integer_1): Likewise.
6767 * cse.c (equiv_constant, cse_insn): Likewise.
6768 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
6769 (cselib_subst_to_values): Likewise.
6770 * dce.c (word_dce_process_block): Likewise.
6771 * df-problems.c (df_word_lr_mark_ref): Likewise.
6772 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
6773 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
6774 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
6775 (rtl_for_decl_location): Likewise.
6776 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
6777 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
6778 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
6779 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
6780 (expand_expr_real_1): Likewise.
6781 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
6782 (pad_below): Likewise.
6783 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6784 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
6785 * ira.c (get_subreg_tracking_sizes): Likewise.
6786 * ira-build.c (ira_create_allocno_objects): Likewise.
6787 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
6788 (ira_sort_regnos_for_alter_reg): Likewise.
6789 * ira-costs.c (record_operand_costs): Likewise.
6790 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
6791 (resolve_simple_move): Likewise.
6792 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
6793 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
6794 (lra_constraints): Likewise.
6795 (CONST_POOL_OK_P): Reject variable-sized modes.
6796 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
6797 (add_pseudo_to_slot, lra_spill): Likewise.
6798 * omp-low.c (omp_clause_aligned_alignment): Likewise.
6799 * optabs-query.c (get_best_extraction_insn): Likewise.
6800 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
6801 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
6802 (expand_mult_highpart, valid_multiword_target_p): Likewise.
6803 * recog.c (offsettable_address_addr_space_p): Likewise.
6804 * regcprop.c (maybe_mode_change): Likewise.
6805 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
6806 * regrename.c (build_def_use): Likewise.
6807 * regstat.c (dump_reg_info): Likewise.
6808 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
6809 (find_reloads, find_reloads_subreg_address): Likewise.
6810 * reload1.c (eliminate_regs_1): Likewise.
6811 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
6812 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
6813 (simplify_binary_operation_1, simplify_subreg): Likewise.
6814 * targhooks.c (default_function_arg_padding): Likewise.
6815 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
6816 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
6817 (verify_gimple_assign_ternary): Likewise.
6818 * tree-inline.c (estimate_move_cost): Likewise.
6819 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
6820 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
6821 (get_address_cost_ainc): Likewise.
6822 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
6823 (vect_supportable_dr_alignment): Likewise.
6824 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6825 (vectorizable_reduction): Likewise.
6826 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
6827 (vectorizable_operation, vectorizable_load): Likewise.
6828 * tree.c (build_same_sized_truth_vector_type): Likewise.
6829 * valtrack.c (cleanup_auto_inc_dec): Likewise.
6830 * var-tracking.c (emit_note_insn_var_location): Likewise.
6831 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
6832 (ADDR_VEC_ALIGN): Likewise.
6834 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6835 Alan Hayward <alan.hayward@arm.com>
6836 David Sherwood <david.sherwood@arm.com>
6838 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
6840 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6841 or if measurement_type is polynomial.
6842 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
6843 * combine.c (make_extraction): Likewise.
6844 * dse.c (find_shift_sequence): Likewise.
6845 * dwarf2out.c (mem_loc_descriptor): Likewise.
6846 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
6847 (extract_bit_field, extract_low_bits): Likewise.
6848 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
6849 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
6850 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
6851 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
6852 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6853 * reload.c (find_reloads): Likewise.
6854 * reload1.c (alter_reg): Likewise.
6855 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
6856 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
6857 * tree-if-conv.c (predicate_mem_writes): Likewise.
6858 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
6859 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
6860 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
6861 * valtrack.c (dead_debug_insert_temp): Likewise.
6862 * varasm.c (mergeable_constant_section): Likewise.
6863 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
6865 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6866 Alan Hayward <alan.hayward@arm.com>
6867 David Sherwood <david.sherwood@arm.com>
6869 * expr.c (expand_assignment): Cope with polynomial mode sizes
6870 when assigning to a CONCAT.
6872 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6873 Alan Hayward <alan.hayward@arm.com>
6874 David Sherwood <david.sherwood@arm.com>
6876 * machmode.h (mode_precision): Change from unsigned short to
6878 (mode_to_precision): Return a poly_uint16 rather than an unsigned
6880 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
6881 or if measurement_type is not polynomial.
6882 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
6883 in which the mode is already known to be a scalar_int_mode.
6884 * genmodes.c (emit_mode_precision): Change the type of mode_precision
6885 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
6887 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6888 for GET_MODE_PRECISION.
6889 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6890 for GET_MODE_PRECISION.
6891 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
6893 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
6894 (expand_field_assignment, make_extraction): Likewise.
6895 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
6896 (get_last_value): Likewise.
6897 * convert.c (convert_to_integer_1): Likewise.
6898 * cse.c (cse_insn): Likewise.
6899 * expr.c (expand_expr_real_1): Likewise.
6900 * lra-constraints.c (simplify_operand_subreg): Likewise.
6901 * optabs-query.c (can_atomic_load_p): Likewise.
6902 * optabs.c (expand_atomic_load): Likewise.
6903 (expand_atomic_store): Likewise.
6904 * ree.c (combine_reaching_defs): Likewise.
6905 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
6906 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
6907 * tree.h (type_has_mode_precision_p): Likewise.
6908 * ubsan.c (instrument_si_overflow): Likewise.
6910 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6911 Alan Hayward <alan.hayward@arm.com>
6912 David Sherwood <david.sherwood@arm.com>
6914 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
6915 polynomial numbers of units.
6916 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
6917 (valid_vector_subparts_p): New function.
6918 (build_vector_type): Remove temporary shim and take the number
6919 of units as a poly_uint64 rather than an int.
6920 (build_opaque_vector_type): Take the number of units as a
6921 poly_uint64 rather than an int.
6922 * tree.c (build_vector_from_ctor): Handle polynomial
6923 TYPE_VECTOR_SUBPARTS.
6924 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
6925 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
6926 (build_vector_from_val): If the number of units is variable,
6927 use build_vec_duplicate_cst for constant operands and
6928 VEC_DUPLICATE_EXPR otherwise.
6929 (make_vector_type): Remove temporary is_constant ().
6930 (build_vector_type, build_opaque_vector_type): Take the number of
6931 units as a poly_uint64 rather than an int.
6932 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
6934 * cfgexpand.c (expand_debug_expr): Likewise.
6935 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
6936 (store_constructor, expand_expr_real_1): Likewise.
6937 (const_scalar_mask_from_tree): Likewise.
6938 * fold-const-call.c (fold_const_reduction): Likewise.
6939 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
6940 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
6941 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
6942 (fold_relational_const): Likewise.
6943 (native_interpret_vector): Likewise. Change the size from an
6944 int to an unsigned int.
6945 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
6946 TYPE_VECTOR_SUBPARTS.
6947 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
6948 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
6949 duplicating a non-constant operand into a variable-length vector.
6950 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
6951 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
6952 * ipa-icf.c (sem_variable::equals): Likewise.
6953 * match.pd: Likewise.
6954 * omp-simd-clone.c (simd_clone_subparts): Likewise.
6955 * print-tree.c (print_node): Likewise.
6956 * stor-layout.c (layout_type): Likewise.
6957 * targhooks.c (default_builtin_vectorization_cost): Likewise.
6958 * tree-cfg.c (verify_gimple_comparison): Likewise.
6959 (verify_gimple_assign_binary): Likewise.
6960 (verify_gimple_assign_ternary): Likewise.
6961 (verify_gimple_assign_single): Likewise.
6962 * tree-pretty-print.c (dump_generic_node): Likewise.
6963 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
6964 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
6965 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
6966 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
6967 (vect_shift_permute_load_chain): Likewise.
6968 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
6969 (expand_vector_condition, optimize_vector_constructor): Likewise.
6970 (lower_vec_perm, get_compute_type): Likewise.
6971 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6972 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
6973 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
6974 (vect_recog_mask_conversion_pattern): Likewise.
6975 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
6976 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
6977 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
6978 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
6979 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
6980 (vectorizable_shift, vectorizable_operation, vectorizable_store)
6981 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
6982 (supportable_widening_operation): Likewise.
6983 (supportable_narrowing_operation): Likewise.
6984 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
6986 * varasm.c (output_constant): Likewise.
6988 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6989 Alan Hayward <alan.hayward@arm.com>
6990 David Sherwood <david.sherwood@arm.com>
6992 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
6993 so that both the length == 3 and length != 3 cases set up their
6994 own permute vectors. Add comments explaining why we know the
6995 number of elements is constant.
6996 (vect_permute_load_chain): Likewise.
6998 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6999 Alan Hayward <alan.hayward@arm.com>
7000 David Sherwood <david.sherwood@arm.com>
7002 * machmode.h (mode_nunits): Change from unsigned char to
7004 (ONLY_FIXED_SIZE_MODES): New macro.
7005 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
7006 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
7007 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
7009 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
7010 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
7011 or if measurement_type is not polynomial.
7012 * genmodes.c (ZERO_COEFFS): New macro.
7013 (emit_mode_nunits_inline): Make mode_nunits_inline return a
7015 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
7016 Use ZERO_COEFFS when emitting initializers.
7017 * data-streamer.h (bp_pack_poly_value): New function.
7018 (bp_unpack_poly_value): Likewise.
7019 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7020 for GET_MODE_NUNITS.
7021 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7022 for GET_MODE_NUNITS.
7023 * tree.c (make_vector_type): Remove temporary shim and make
7024 the real function take the number of units as a poly_uint64
7026 (build_vector_type_for_mode): Handle polynomial nunits.
7027 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
7028 * emit-rtl.c (const_vec_series_p_1): Likewise.
7029 (gen_rtx_CONST_VECTOR): Likewise.
7030 * fold-const.c (test_vec_duplicate_folding): Likewise.
7031 * genrecog.c (validate_pattern): Likewise.
7032 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
7033 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7034 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
7035 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
7036 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
7037 * rtlanal.c (subreg_get_info): Likewise.
7038 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7039 (vect_grouped_load_supported): Likewise.
7040 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
7041 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
7042 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
7043 (simplify_const_unary_operation, simplify_binary_operation_1)
7044 (simplify_const_binary_operation, simplify_ternary_operation)
7045 (test_vector_ops_duplicate, test_vector_ops): Likewise.
7046 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
7047 instead of CONST_VECTOR_NUNITS.
7048 * varasm.c (output_constant_pool_2): Likewise.
7049 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
7050 explicit-encoded elements in the XVEC for variable-length vectors.
7052 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7054 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
7056 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7057 Alan Hayward <alan.hayward@arm.com>
7058 David Sherwood <david.sherwood@arm.com>
7060 * coretypes.h (fixed_size_mode): Declare.
7061 (fixed_size_mode_pod): New typedef.
7062 * builtins.h (target_builtins::x_apply_args_mode)
7063 (target_builtins::x_apply_result_mode): Change type to
7064 fixed_size_mode_pod.
7065 * builtins.c (apply_args_size, apply_result_size, result_vector)
7066 (expand_builtin_apply_args_1, expand_builtin_apply)
7067 (expand_builtin_return): Update accordingly.
7069 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7071 * cse.c (hash_rtx_cb): Hash only the encoded elements.
7072 * cselib.c (cselib_hash_rtx): Likewise.
7073 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
7074 CONST_VECTOR encoding.
7076 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7077 Jeff Law <law@redhat.com>
7080 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
7081 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
7082 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
7083 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
7086 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
7087 explicitly probe *sp in a noreturn function if there were any callee
7088 register saves or frame pointer is needed.
7090 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7093 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
7094 BLKmode for ternary, binary or unary expressions.
7097 * var-tracking.c (delete_vta_debug_insn): New inline function.
7098 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
7099 insns from get_insns () to NULL instead of each bb separately.
7100 Use delete_vta_debug_insn. No longer static.
7101 (vt_debug_insns_local, variable_tracking_main_1): Adjust
7102 delete_vta_debug_insns callers.
7103 * rtl.h (delete_vta_debug_insns): Declare.
7104 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
7105 instead of variable_tracking_main.
7107 2018-01-03 Martin Sebor <msebor@redhat.com>
7109 PR tree-optimization/83603
7110 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
7111 arguments past the endof the argument list in functions declared
7112 without a prototype.
7113 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
7114 Avoid checking when arguments are null.
7116 2018-01-03 Martin Sebor <msebor@redhat.com>
7119 * doc/extend.texi (attribute const): Fix a typo.
7120 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
7121 issuing -Wsuggest-attribute for void functions.
7123 2018-01-03 Martin Sebor <msebor@redhat.com>
7125 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
7126 offset_int::from instead of wide_int::to_shwi.
7127 (maybe_diag_overlap): Remove assertion.
7128 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
7129 * gimple-ssa-sprintf.c (format_directive): Same.
7130 (parse_directive): Same.
7131 (sprintf_dom_walker::compute_format_length): Same.
7132 (try_substitute_return_value): Same.
7134 2018-01-03 Jeff Law <law@redhat.com>
7137 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
7138 non-constant residual for zero at runtime and avoid probing in
7139 that case. Reorganize code for trailing problem to mirror handling
7142 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7144 PR tree-optimization/83501
7145 * tree-ssa-strlen.c (get_string_cst): New.
7146 (handle_char_store): Call get_string_cst.
7148 2018-01-03 Martin Liska <mliska@suse.cz>
7150 PR tree-optimization/83593
7151 * tree-ssa-strlen.c: Include tree-cfg.h.
7152 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
7153 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
7154 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
7156 (strlen_dom_walker::before_dom_children): Call
7157 gimple_purge_dead_eh_edges. Dump tranformation with details
7159 (strlen_dom_walker::before_dom_children): Update call by adding
7160 new argument cleanup_eh.
7161 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
7163 2018-01-03 Martin Liska <mliska@suse.cz>
7166 * cif-code.def (VARIADIC_THUNK): New enum value.
7167 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
7170 2018-01-03 Jan Beulich <jbeulich@suse.com>
7172 * sse.md (mov<mode>_internal): Tighten condition for when to use
7173 vmovdqu<ssescalarsize> for TI and OI modes.
7175 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7177 Update copyright years.
7179 2018-01-03 Martin Liska <mliska@suse.cz>
7182 * ipa-visibility.c (function_and_variable_visibility): Skip
7183 functions with noipa attribure.
7185 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7187 * gcc.c (process_command): Update copyright notice dates.
7188 * gcov-dump.c (print_version): Ditto.
7189 * gcov.c (print_version): Ditto.
7190 * gcov-tool.c (print_version): Ditto.
7191 * gengtype.c (create_file): Ditto.
7192 * doc/cpp.texi: Bump @copying's copyright year.
7193 * doc/cppinternals.texi: Ditto.
7194 * doc/gcc.texi: Ditto.
7195 * doc/gccint.texi: Ditto.
7196 * doc/gcov.texi: Ditto.
7197 * doc/install.texi: Ditto.
7198 * doc/invoke.texi: Ditto.
7200 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7202 * vector-builder.h (vector_builder::m_full_nelts): Change from
7203 unsigned int to poly_uint64.
7204 (vector_builder::full_nelts): Update prototype accordingly.
7205 (vector_builder::new_vector): Likewise.
7206 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
7207 (vector_builder::operator ==): Likewise.
7208 (vector_builder::finalize): Likewise.
7209 * int-vector-builder.h (int_vector_builder::int_vector_builder):
7210 Take the number of elements as a poly_uint64 rather than an
7212 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
7213 from unsigned int to poly_uint64.
7214 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
7215 (vec_perm_indices::new_vector): Likewise.
7216 (vec_perm_indices::length): Likewise.
7217 (vec_perm_indices::nelts_per_input): Likewise.
7218 (vec_perm_indices::input_nelts): Likewise.
7219 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
7220 number of elements per input as a poly_uint64 rather than an
7221 unsigned int. Use the original encoding for variable-length
7222 vectors, rather than clamping each individual element.
7223 For the second and subsequent elements in each pattern,
7224 clamp the step and base before clamping their sum.
7225 (vec_perm_indices::series_p): Handle polynomial element counts.
7226 (vec_perm_indices::all_in_range_p): Likewise.
7227 (vec_perm_indices_to_tree): Likewise.
7228 (vec_perm_indices_to_rtx): Likewise.
7229 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
7230 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
7231 (tree_vector_builder::new_binary_operation): Handle polynomial
7232 element counts. Return false if we need to know the number
7233 of elements at compile time.
7234 * fold-const.c (fold_vec_perm): Punt if the number of elements
7235 isn't known at compile time.
7237 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7239 * vec-perm-indices.h (vec_perm_builder): Change element type
7240 from HOST_WIDE_INT to poly_int64.
7241 (vec_perm_indices::element_type): Update accordingly.
7242 (vec_perm_indices::clamp): Handle polynomial element_types.
7243 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7244 (vec_perm_indices::all_in_range_p): Likewise.
7245 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
7247 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
7248 polynomial vec_perm_indices element types.
7249 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
7250 * fold-const.c (fold_vec_perm): Likewise.
7251 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
7252 * tree-vect-generic.c (lower_vec_perm): Likewise.
7253 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7254 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
7255 element type to HOST_WIDE_INT.
7257 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7258 Alan Hayward <alan.hayward@arm.com>
7259 David Sherwood <david.sherwood@arm.com>
7261 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
7262 rather than an int. Use plus_constant.
7263 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
7264 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
7266 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7267 Alan Hayward <alan.hayward@arm.com>
7268 David Sherwood <david.sherwood@arm.com>
7270 * calls.c (emit_call_1, expand_call): Change struct_value_size from
7271 a HOST_WIDE_INT to a poly_int64.
7273 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7274 Alan Hayward <alan.hayward@arm.com>
7275 David Sherwood <david.sherwood@arm.com>
7277 * calls.c (load_register_parameters): Cope with polynomial
7278 mode sizes. Require a constant size for BLKmode parameters
7279 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
7280 forces a parameter to be padded at the lsb end in order to
7281 fill a complete number of words, require the parameter size
7282 to be ordered wrt UNITS_PER_WORD.
7284 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7285 Alan Hayward <alan.hayward@arm.com>
7286 David Sherwood <david.sherwood@arm.com>
7288 * reload1.c (spill_stack_slot_width): Change element type
7289 from unsigned int to poly_uint64_pod.
7290 (alter_reg): Treat mode sizes as polynomial.
7292 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7293 Alan Hayward <alan.hayward@arm.com>
7294 David Sherwood <david.sherwood@arm.com>
7296 * reload.c (complex_word_subreg_p): New function.
7297 (reload_inner_reg_of_subreg, push_reload): Use it.
7299 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7300 Alan Hayward <alan.hayward@arm.com>
7301 David Sherwood <david.sherwood@arm.com>
7303 * lra-constraints.c (process_alt_operands): Reject matched
7304 operands whose sizes aren't ordered.
7305 (match_reload): Refer to this check here.
7307 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7308 Alan Hayward <alan.hayward@arm.com>
7309 David Sherwood <david.sherwood@arm.com>
7311 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
7312 that the mode size is in the set {1, 2, 4, 8, 16}.
7314 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7315 Alan Hayward <alan.hayward@arm.com>
7316 David Sherwood <david.sherwood@arm.com>
7318 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
7319 Use plus_constant instead of gen_rtx_PLUS.
7321 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7322 Alan Hayward <alan.hayward@arm.com>
7323 David Sherwood <david.sherwood@arm.com>
7325 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
7326 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
7327 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
7328 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
7329 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
7330 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
7331 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
7332 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
7333 * config/i386/i386.c (ix86_push_rounding): ...this new function.
7334 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
7336 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
7337 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
7338 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
7339 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
7340 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
7341 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
7342 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
7343 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
7344 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
7345 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
7347 * expr.c (emit_move_resolve_push): Treat the input and result
7348 of PUSH_ROUNDING as a poly_int64.
7349 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
7350 (emit_push_insn): Likewise.
7351 * lra-eliminations.c (mark_not_eliminable): Likewise.
7352 * recog.c (push_operand): Likewise.
7353 * reload1.c (elimination_effects): Likewise.
7354 * rtlanal.c (nonzero_bits1): Likewise.
7355 * calls.c (store_one_arg): Likewise. Require the padding to be
7356 known at compile time.
7358 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7359 Alan Hayward <alan.hayward@arm.com>
7360 David Sherwood <david.sherwood@arm.com>
7362 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
7363 Use plus_constant instead of gen_rtx_PLUS.
7365 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7366 Alan Hayward <alan.hayward@arm.com>
7367 David Sherwood <david.sherwood@arm.com>
7369 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
7372 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7373 Alan Hayward <alan.hayward@arm.com>
7374 David Sherwood <david.sherwood@arm.com>
7376 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
7377 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
7378 via stack temporaries. Treat the mode size as polynomial too.
7380 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7381 Alan Hayward <alan.hayward@arm.com>
7382 David Sherwood <david.sherwood@arm.com>
7384 * expr.c (expand_expr_real_2): When handling conversions involving
7385 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
7386 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
7387 as a poly_uint64 too.
7389 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7390 Alan Hayward <alan.hayward@arm.com>
7391 David Sherwood <david.sherwood@arm.com>
7393 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
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 * combine.c (can_change_dest_mode): Handle polynomial
7400 REGMODE_NATURAL_SIZE.
7401 * expmed.c (store_bit_field_1): Likewise.
7402 * expr.c (store_constructor): Likewise.
7403 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
7404 and polynomial REGMODE_NATURAL_SIZE.
7405 (gen_lowpart_common): Likewise.
7406 * reginfo.c (record_subregs_of_mode): Likewise.
7407 * rtlanal.c (read_modify_subreg_p): Likewise.
7409 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7410 Alan Hayward <alan.hayward@arm.com>
7411 David Sherwood <david.sherwood@arm.com>
7413 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
7414 numbers of elements.
7416 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7417 Alan Hayward <alan.hayward@arm.com>
7418 David Sherwood <david.sherwood@arm.com>
7420 * match.pd: Cope with polynomial numbers of vector elements.
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 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
7427 in a POINTER_PLUS_EXPR.
7429 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7430 Alan Hayward <alan.hayward@arm.com>
7431 David Sherwood <david.sherwood@arm.com>
7433 * omp-simd-clone.c (simd_clone_subparts): New function.
7434 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
7435 (ipa_simd_modify_function_body): Likewise.
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 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
7442 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
7443 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
7444 (expand_vector_condition, vector_element): Likewise.
7445 (subparts_gt): New function.
7446 (get_compute_type): Use subparts_gt.
7447 (count_type_subparts): Delete.
7448 (expand_vector_operations_1): Use subparts_gt instead of
7449 count_type_subparts.
7451 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7452 Alan Hayward <alan.hayward@arm.com>
7453 David Sherwood <david.sherwood@arm.com>
7455 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
7456 (vect_compile_time_alias): ...this new function. Do the calculation
7457 on poly_ints rather than trees.
7458 (vect_prune_runtime_alias_test_list): Update call accordingly.
7460 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7461 Alan Hayward <alan.hayward@arm.com>
7462 David Sherwood <david.sherwood@arm.com>
7464 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
7466 (vect_schedule_slp_instance): Likewise.
7468 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7469 Alan Hayward <alan.hayward@arm.com>
7470 David Sherwood <david.sherwood@arm.com>
7472 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
7473 constant and extern definitions for variable-length vectors.
7474 (vect_get_constant_vectors): Note that the number of units
7475 is known to be constant.
7477 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7478 Alan Hayward <alan.hayward@arm.com>
7479 David Sherwood <david.sherwood@arm.com>
7481 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
7482 of units as polynomial. Choose between WIDE and NARROW based
7485 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7486 Alan Hayward <alan.hayward@arm.com>
7487 David Sherwood <david.sherwood@arm.com>
7489 * tree-vect-stmts.c (simd_clone_subparts): New function.
7490 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
7492 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7493 Alan Hayward <alan.hayward@arm.com>
7494 David Sherwood <david.sherwood@arm.com>
7496 * tree-vect-stmts.c (vectorizable_call): Treat the number of
7497 vectors as polynomial. Use build_index_vector for
7500 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7501 Alan Hayward <alan.hayward@arm.com>
7502 David Sherwood <david.sherwood@arm.com>
7504 * tree-vect-stmts.c (get_load_store_type): Treat the number of
7505 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
7506 for variable-length vectors.
7507 (vectorizable_mask_load_store): Treat the number of units as
7508 polynomial, asserting that it is constant if the condition has
7509 already been enforced.
7510 (vectorizable_store, vectorizable_load): Likewise.
7512 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7513 Alan Hayward <alan.hayward@arm.com>
7514 David Sherwood <david.sherwood@arm.com>
7516 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
7517 of units as polynomial. Punt if we can't tell at compile time
7518 which vector contains the final result.
7520 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7521 Alan Hayward <alan.hayward@arm.com>
7522 David Sherwood <david.sherwood@arm.com>
7524 * tree-vect-loop.c (vectorizable_induction): Treat the number
7525 of units as polynomial. Punt on SLP inductions. Use an integer
7526 VEC_SERIES_EXPR for variable-length integer reductions. Use a
7527 cast of such a series for variable-length floating-point
7530 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7531 Alan Hayward <alan.hayward@arm.com>
7532 David Sherwood <david.sherwood@arm.com>
7534 * tree.h (build_index_vector): Declare.
7535 * tree.c (build_index_vector): New function.
7536 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
7537 of units as polynomial, forcibly converting it to a constant if
7538 vectorizable_reduction has already enforced the condition.
7539 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
7540 to create a {1,2,3,...} vector.
7541 (vectorizable_reduction): Treat the number of units as polynomial.
7542 Choose vectype_in based on the largest scalar element size rather
7543 than the smallest number of units. Enforce the restrictions
7546 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7547 Alan Hayward <alan.hayward@arm.com>
7548 David Sherwood <david.sherwood@arm.com>
7550 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
7551 number of units as polynomial.
7553 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7554 Alan Hayward <alan.hayward@arm.com>
7555 David Sherwood <david.sherwood@arm.com>
7557 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
7558 * target.def (autovectorize_vector_sizes): Return the vector sizes
7559 by pointer, using vector_sizes rather than a bitmask.
7560 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
7561 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
7562 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
7564 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
7565 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
7566 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
7567 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
7568 * omp-general.c (omp_max_vf): Likewise.
7569 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7570 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
7571 * tree-vect-loop.c (vect_analyze_loop): Likewise.
7572 * tree-vect-slp.c (vect_slp_bb): Likewise.
7573 * doc/tm.texi: Regenerate.
7574 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
7576 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
7577 the vector size as a poly_uint64 rather than an unsigned int.
7578 (current_vector_size): Change from an unsigned int to a poly_uint64.
7579 (get_vectype_for_scalar_type): Update accordingly.
7580 * tree.h (build_truth_vector_type): Take the size and number of
7581 units as a poly_uint64 rather than an unsigned int.
7582 (build_vector_type): Add a temporary overload that takes
7583 the number of units as a poly_uint64 rather than an unsigned int.
7584 * tree.c (make_vector_type): Likewise.
7585 (build_truth_vector_type): Take the number of units as a poly_uint64
7586 rather than an unsigned int.
7588 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7589 Alan Hayward <alan.hayward@arm.com>
7590 David Sherwood <david.sherwood@arm.com>
7592 * target.def (get_mask_mode): Take the number of units and length
7593 as poly_uint64s rather than unsigned ints.
7594 * targhooks.h (default_get_mask_mode): Update accordingly.
7595 * targhooks.c (default_get_mask_mode): Likewise.
7596 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
7597 * doc/tm.texi: Regenerate.
7599 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7600 Alan Hayward <alan.hayward@arm.com>
7601 David Sherwood <david.sherwood@arm.com>
7603 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
7604 * omp-general.c (omp_max_vf): Likewise.
7605 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
7606 (expand_omp_simd): Handle polynomial safelen.
7607 * omp-low.c (omplow_simd_context): Add a default constructor.
7608 (omplow_simd_context::max_vf): Change from int to poly_uint64.
7609 (lower_rec_simd_input_clauses): Update accordingly.
7610 (lower_rec_input_clauses): Likewise.
7612 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7613 Alan Hayward <alan.hayward@arm.com>
7614 David Sherwood <david.sherwood@arm.com>
7616 * tree-vectorizer.h (vect_nunits_for_cost): New function.
7617 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
7618 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
7619 (vect_analyze_slp_cost): Likewise.
7620 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
7621 (vect_model_load_cost): Likewise.
7623 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7624 Alan Hayward <alan.hayward@arm.com>
7625 David Sherwood <david.sherwood@arm.com>
7627 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
7628 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
7629 from an unsigned int * to a poly_uint64_pod *.
7630 (calculate_unrolling_factor): New function.
7631 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
7633 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7634 Alan Hayward <alan.hayward@arm.com>
7635 David Sherwood <david.sherwood@arm.com>
7637 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
7638 from an unsigned int to a poly_uint64.
7639 (_loop_vec_info::slp_unrolling_factor): Likewise.
7640 (_loop_vec_info::vectorization_factor): Change from an int
7642 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
7643 (vect_get_num_vectors): New function.
7644 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
7645 (vect_get_num_copies): Use vect_get_num_vectors.
7646 (vect_analyze_data_ref_dependences): Change max_vf from an int *
7647 to an unsigned int *.
7648 (vect_analyze_data_refs): Change min_vf from an int * to a
7650 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7651 than an unsigned HOST_WIDE_INT.
7652 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
7653 (vect_analyze_data_ref_dependence): Change max_vf from an int *
7654 to an unsigned int *.
7655 (vect_analyze_data_ref_dependences): Likewise.
7656 (vect_compute_data_ref_alignment): Handle polynomial vf.
7657 (vect_enhance_data_refs_alignment): Likewise.
7658 (vect_prune_runtime_alias_test_list): Likewise.
7659 (vect_shift_permute_load_chain): Likewise.
7660 (vect_supportable_dr_alignment): Likewise.
7661 (dependence_distance_ge_vf): Take the vectorization factor as a
7662 poly_uint64 rather than an unsigned HOST_WIDE_INT.
7663 (vect_analyze_data_refs): Change min_vf from an int * to a
7665 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
7666 vfm1 as a poly_uint64 rather than an int. Make the same change
7667 for the returned bound_scalar.
7668 (vect_gen_vector_loop_niters): Handle polynomial vf.
7669 (vect_do_peeling): Likewise. Update call to
7670 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
7671 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
7673 * tree-vect-loop.c (vect_determine_vectorization_factor)
7674 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
7675 (vect_get_known_peeling_cost): Likewise.
7676 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
7677 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
7678 (vect_transform_loop): Likewise. Use the lowest possible VF when
7679 updating the upper bounds of the loop.
7680 (vect_min_worthwhile_factor): Make static. Return an unsigned int
7682 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
7683 polynomial unroll factors.
7684 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
7685 (vect_make_slp_decision): Likewise.
7686 (vect_supported_load_permutation_p): Likewise, and polynomial
7688 (vect_analyze_slp_cost): Handle polynomial vf.
7689 (vect_slp_analyze_node_operations): Likewise.
7690 (vect_slp_analyze_bb_1): Likewise.
7691 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7692 than an unsigned HOST_WIDE_INT.
7693 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
7694 (vectorizable_load): Handle polynomial vf.
7695 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
7697 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
7699 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7700 Alan Hayward <alan.hayward@arm.com>
7701 David Sherwood <david.sherwood@arm.com>
7703 * match.pd: Handle bit operations involving three constants
7704 and try to fold one pair.
7706 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7708 * tree-vect-loop-manip.c: Include gimple-fold.h.
7709 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
7710 niters_maybe_zero parameters. Handle other cases besides a step of 1.
7711 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
7712 Add a path that uses a step of VF instead of 1, but disable it
7714 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
7715 and niters_no_overflow parameters. Update calls to
7716 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
7717 Create a new SSA name if the latter choses to use a ste other
7718 than zero, and return it via niters_vector_mult_vf_var.
7719 * tree-vect-loop.c (vect_transform_loop): Update calls to
7720 vect_do_peeling, vect_gen_vector_loop_niters and
7721 slpeel_make_loop_iterate_ntimes.
7722 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
7723 (vect_gen_vector_loop_niters): Update declarations after above changes.
7725 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
7727 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
7728 128-bit round to integer instructions.
7729 (ceil<mode>2): Likewise.
7730 (btrunc<mode>2): Likewise.
7731 (round<mode>2): Likewise.
7733 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7735 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
7736 unaligned VSX load/store on P8/P9.
7737 (expand_block_clear): Allow the use of unaligned VSX
7738 load/store on P8/P9.
7740 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7742 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
7744 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
7745 swap associated with both a load and a store.
7747 2018-01-02 Andrew Waterman <andrew@sifive.com>
7749 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
7750 * config/riscv/riscv.md (clear_cache): Use it.
7752 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
7754 * web.c: Remove out-of-date comment.
7756 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7758 * expr.c (fixup_args_size_notes): Check that any existing
7759 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
7760 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
7761 (emit_single_push_insn): ...here.
7763 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7765 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
7766 (const_vector_encoded_nelts): New function.
7767 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
7768 (const_vector_int_elt, const_vector_elt): Declare.
7769 * emit-rtl.c (const_vector_int_elt_1): New function.
7770 (const_vector_elt): Likewise.
7771 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
7772 of CONST_VECTOR_ELT.
7774 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7776 * expr.c: Include rtx-vector-builder.h.
7777 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
7778 directly on the tree encoding.
7779 (const_vector_from_tree): Likewise.
7780 * optabs.c: Include rtx-vector-builder.h.
7781 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
7782 sequence of "u" values.
7783 * vec-perm-indices.c: Include rtx-vector-builder.h.
7784 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
7785 directly on the vec_perm_indices encoding.
7787 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7789 * doc/rtl.texi (const_vector): Describe new encoding scheme.
7790 * Makefile.in (OBJS): Add rtx-vector-builder.o.
7791 * rtx-vector-builder.h: New file.
7792 * rtx-vector-builder.c: Likewise.
7793 * rtl.h (rtx_def::u2): Add a const_vector field.
7794 (CONST_VECTOR_NPATTERNS): New macro.
7795 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
7796 (CONST_VECTOR_DUPLICATE_P): Likewise.
7797 (CONST_VECTOR_STEPPED_P): Likewise.
7798 (CONST_VECTOR_ENCODED_ELT): Likewise.
7799 (const_vec_duplicate_p): Check for a duplicated vector encoding.
7800 (unwrap_const_vec_duplicate): Likewise.
7801 (const_vec_series_p): Check for a non-duplicated vector encoding.
7802 Say that the function only returns true for integer vectors.
7803 * emit-rtl.c: Include rtx-vector-builder.h.
7804 (gen_const_vec_duplicate_1): Delete.
7805 (gen_const_vector): Call gen_const_vec_duplicate instead of
7806 gen_const_vec_duplicate_1.
7807 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
7808 (gen_const_vec_duplicate): Use rtx_vector_builder.
7809 (gen_const_vec_series): Likewise.
7810 (gen_rtx_CONST_VECTOR): Likewise.
7811 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
7812 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7813 Build a new vector rather than modifying a CONST_VECTOR in-place.
7814 (handle_special_swappables): Update call accordingly.
7815 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
7816 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7817 Build a new vector rather than modifying a CONST_VECTOR in-place.
7818 (handle_special_swappables): Update call accordingly.
7820 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7822 * simplify-rtx.c (simplify_const_binary_operation): Use
7823 CONST_VECTOR_ELT instead of XVECEXP.
7825 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7827 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
7828 the selector elements to be different from the data elements
7829 if the selector is a VECTOR_CST.
7830 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
7831 ssizetype for the selector.
7833 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7835 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
7836 before testing each element individually.
7837 * tree-vect-generic.c (lower_vec_perm): Likewise.
7839 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7841 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
7842 * selftest-run-tests.c (selftest::run_tests): Call it.
7843 * vector-builder.h (vector_builder::operator ==): New function.
7844 (vector_builder::operator !=): Likewise.
7845 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
7846 (vec_perm_indices::all_from_input_p): New function.
7847 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7848 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
7849 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
7850 instead of reading the VECTOR_CST directly. Detect whether both
7851 vector inputs are the same before constructing the vec_perm_indices,
7852 and update the number of inputs argument accordingly. Use the
7853 utility functions added above. Only construct sel2 if we need to.
7855 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7857 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
7858 the broadcast of the low byte.
7859 (expand_mult_highpart): Use an explicit encoding for the permutes.
7860 * optabs-query.c (can_mult_highpart_p): Likewise.
7861 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
7862 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7863 (vectorizable_bswap): Likewise.
7864 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
7865 explicit encoding for the power-of-2 permutes.
7866 (vect_permute_store_chain): Likewise.
7867 (vect_grouped_load_supported): Likewise.
7868 (vect_permute_load_chain): Likewise.
7870 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7872 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
7873 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
7874 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
7875 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7876 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
7877 (vect_gen_perm_mask_any): Likewise.
7879 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7881 * int-vector-builder.h: New file.
7882 * vec-perm-indices.h: Include int-vector-builder.h.
7883 (vec_perm_indices): Redefine as an int_vector_builder.
7884 (auto_vec_perm_indices): Delete.
7885 (vec_perm_builder): Redefine as a stand-alone class.
7886 (vec_perm_indices::vec_perm_indices): New function.
7887 (vec_perm_indices::clamp): Likewise.
7888 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
7889 (vec_perm_indices::new_vector): New function.
7890 (vec_perm_indices::new_expanded_vector): Update for new
7891 vec_perm_indices class.
7892 (vec_perm_indices::rotate_inputs): New function.
7893 (vec_perm_indices::all_in_range_p): Operate directly on the
7894 encoded form, without computing elided elements.
7895 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
7896 encoding. Update for new vec_perm_indices class.
7897 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
7898 the given vec_perm_builder.
7899 (expand_vec_perm_var): Update vec_perm_builder constructor.
7900 (expand_mult_highpart): Use vec_perm_builder instead of
7901 auto_vec_perm_indices.
7902 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
7903 vec_perm_indices instead of auto_vec_perm_indices. Use a single
7904 or double series encoding as appropriate.
7905 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
7906 vec_perm_indices instead of auto_vec_perm_indices.
7907 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7908 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7909 (vect_permute_store_chain): Likewise.
7910 (vect_grouped_load_supported): Likewise.
7911 (vect_permute_load_chain): Likewise.
7912 (vect_shift_permute_load_chain): Likewise.
7913 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
7914 (vect_transform_slp_perm_load): Likewise.
7915 (vect_schedule_slp_instance): Likewise.
7916 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7917 (vectorizable_mask_load_store): Likewise.
7918 (vectorizable_bswap): Likewise.
7919 (vectorizable_store): Likewise.
7920 (vectorizable_load): Likewise.
7921 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
7922 vec_perm_indices instead of auto_vec_perm_indices. Use
7923 tree_to_vec_perm_builder to read the vector from a tree.
7924 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
7925 vec_perm_builder instead of a vec_perm_indices.
7926 (have_whole_vector_shift): Use vec_perm_builder and
7927 vec_perm_indices instead of auto_vec_perm_indices. Leave the
7928 truncation to calc_vec_perm_mask_for_shift.
7929 (vect_create_epilog_for_reduction): Likewise.
7930 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
7931 from auto_vec_perm_indices to vec_perm_indices.
7932 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
7933 instead of changing individual elements.
7934 (aarch64_vectorize_vec_perm_const): Use new_vector to install
7935 the vector in d.perm.
7936 * config/arm/arm.c (expand_vec_perm_d::perm): Change
7937 from auto_vec_perm_indices to vec_perm_indices.
7938 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
7939 instead of changing individual elements.
7940 (arm_vectorize_vec_perm_const): Use new_vector to install
7941 the vector in d.perm.
7942 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
7943 Update vec_perm_builder constructor.
7944 (rs6000_expand_interleave): Likewise.
7945 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
7946 (rs6000_expand_interleave): Likewise.
7948 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7950 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
7951 to qimode could truncate the indices.
7952 * optabs.c (expand_vec_perm_var): Likewise.
7954 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7956 * Makefile.in (OBJS): Add vec-perm-indices.o.
7957 * vec-perm-indices.h: New file.
7958 * vec-perm-indices.c: Likewise.
7959 * target.h (vec_perm_indices): Replace with a forward class
7961 (auto_vec_perm_indices): Move to vec-perm-indices.h.
7962 * optabs.h: Include vec-perm-indices.h.
7963 (expand_vec_perm): Delete.
7964 (selector_fits_mode_p, expand_vec_perm_var): Declare.
7965 (expand_vec_perm_const): Declare.
7966 * target.def (vec_perm_const_ok): Replace with...
7967 (vec_perm_const): ...this new hook.
7968 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
7969 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
7970 * doc/tm.texi: Regenerate.
7971 * optabs.def (vec_perm_const): Delete.
7972 * doc/md.texi (vec_perm_const): Likewise.
7973 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
7974 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
7975 expand_vec_perm for constant permutation vectors. Assert that
7976 the mode of variable permutation vectors is the integer equivalent
7977 of the mode that is being permuted.
7978 * optabs-query.h (selector_fits_mode_p): Declare.
7979 * optabs-query.c: Include vec-perm-indices.h.
7980 (selector_fits_mode_p): New function.
7981 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
7982 is defined, instead of checking whether the vec_perm_const_optab
7983 exists. Use targetm.vectorize.vec_perm_const instead of
7984 targetm.vectorize.vec_perm_const_ok. Check whether the indices
7985 fit in the vector mode before using a variable permute.
7986 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
7987 vec_perm_indices instead of an rtx.
7988 (expand_vec_perm): Replace with...
7989 (expand_vec_perm_const): ...this new function. Take the selector
7990 as a vec_perm_indices rather than an rtx. Also take the mode of
7991 the selector. Update call to shift_amt_for_vec_perm_mask.
7992 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
7993 Use vec_perm_indices::new_expanded_vector to expand the original
7994 selector into bytes. Check whether the indices fit in the vector
7995 mode before using a variable permute.
7996 (expand_vec_perm_var): Make global.
7997 (expand_mult_highpart): Use expand_vec_perm_const.
7998 * fold-const.c: Includes vec-perm-indices.h.
7999 * tree-ssa-forwprop.c: Likewise.
8000 * tree-vect-data-refs.c: Likewise.
8001 * tree-vect-generic.c: Likewise.
8002 * tree-vect-loop.c: Likewise.
8003 * tree-vect-slp.c: Likewise.
8004 * tree-vect-stmts.c: Likewise.
8005 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
8007 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
8008 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
8009 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
8010 (aarch64_vectorize_vec_perm_const): ...this new function.
8011 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8012 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8013 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
8014 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
8015 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8016 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8017 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
8019 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
8020 check for NEON modes.
8021 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
8022 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
8023 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
8024 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
8026 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
8027 the old VEC_PERM_CONST conditions.
8028 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
8029 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
8030 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
8031 (ia64_vectorize_vec_perm_const_ok): Merge into...
8032 (ia64_vectorize_vec_perm_const): ...this new function.
8033 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
8034 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
8035 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
8036 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
8037 * config/mips/mips.c (mips_expand_vec_perm_const)
8038 (mips_vectorize_vec_perm_const_ok): Merge into...
8039 (mips_vectorize_vec_perm_const): ...this new function.
8040 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
8041 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
8042 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
8043 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
8044 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
8045 (rs6000_expand_vec_perm_const): Delete.
8046 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
8048 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8049 (altivec_expand_vec_perm_const_le): Take each operand individually.
8050 Operate on constant selectors rather than rtxes.
8051 (altivec_expand_vec_perm_const): Likewise. Update call to
8052 altivec_expand_vec_perm_const_le.
8053 (rs6000_expand_vec_perm_const): Delete.
8054 (rs6000_vectorize_vec_perm_const_ok): Delete.
8055 (rs6000_vectorize_vec_perm_const): New function.
8056 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8057 an element count and rtx array.
8058 (rs6000_expand_extract_even): Update call accordingly.
8059 (rs6000_expand_interleave): Likewise.
8060 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
8061 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
8062 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
8063 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
8064 (rs6000_expand_vec_perm_const): Delete.
8065 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8066 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8067 (altivec_expand_vec_perm_const_le): Take each operand individually.
8068 Operate on constant selectors rather than rtxes.
8069 (altivec_expand_vec_perm_const): Likewise. Update call to
8070 altivec_expand_vec_perm_const_le.
8071 (rs6000_expand_vec_perm_const): Delete.
8072 (rs6000_vectorize_vec_perm_const_ok): Delete.
8073 (rs6000_vectorize_vec_perm_const): New function. Remove stray
8074 reference to the SPE evmerge intructions.
8075 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8076 an element count and rtx array.
8077 (rs6000_expand_extract_even): Update call accordingly.
8078 (rs6000_expand_interleave): Likewise.
8079 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
8080 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
8082 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8084 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8086 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
8087 vector mode and that that mode matches the mode of the data
8089 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
8090 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
8091 directly using expand_vec_perm_1 when forcing selectors into
8093 (expand_vec_perm_var): New function, split out from expand_vec_perm.
8095 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8097 * optabs-query.h (can_vec_perm_p): Delete.
8098 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
8099 * optabs-query.c (can_vec_perm_p): Split into...
8100 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
8101 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
8102 particular selector is valid.
8103 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8104 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8105 (vect_grouped_load_supported): Likewise.
8106 (vect_shift_permute_load_chain): Likewise.
8107 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8108 (vect_transform_slp_perm_load): Likewise.
8109 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8110 (vectorizable_bswap): Likewise.
8111 (vect_gen_perm_mask_checked): Likewise.
8112 * fold-const.c (fold_ternary_loc): Likewise. Don't take
8113 implementations of variable permutation vectors into account
8114 when deciding which selector to use.
8115 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
8116 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
8117 with a false third argument.
8118 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
8119 to test whether the constant selector is valid and can_vec_perm_var_p
8120 to test whether a variable selector is valid.
8122 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8124 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
8125 * optabs-query.c (can_vec_perm_p): Likewise.
8126 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
8127 instead of vec_perm_indices.
8128 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
8129 (vect_gen_perm_mask_checked): Likewise,
8130 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
8131 (vect_gen_perm_mask_checked): Likewise,
8133 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8135 * optabs-query.h (qimode_for_vec_perm): Declare.
8136 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
8137 (qimode_for_vec_perm): ...this new function.
8138 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
8140 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8142 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
8143 does not have a conditional at the top.
8145 2018-01-02 Richard Biener <rguenther@suse.de>
8147 * ipa-inline.c (big_speedup_p): Fix expression.
8149 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8152 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
8155 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8159 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
8160 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
8161 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
8162 cond_taken_branch_cost 3->4.
8164 2018-01-01 Jakub Jelinek <jakub@redhat.com>
8166 PR tree-optimization/83581
8167 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
8168 TODO_cleanup_cfg if any changes have been made.
8171 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
8172 convert_modes if target mode has the right side, but different mode
8176 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
8177 last argument when extracting from CONCAT. If either from_real or
8178 from_imag is NULL, use expansion through memory. If result is not
8179 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
8180 the parts directly to inner mode, if even that fails, use expansion
8184 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
8185 check for bswap in mode rather than HImode and use that in expand_unop
8188 Copyright (C) 2018 Free Software Foundation, Inc.
8190 Copying and distribution of this file, with or without modification,
8191 are permitted in any medium without royalty provided the copyright
8192 notice and this notice are preserved.