1 2018-03-06 Jakub Jelinek <jakub@redhat.com>
4 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
7 PR tree-optimization/84687
8 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
10 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
12 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
15 Rename to ppc_speculation_barrier.
16 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
17 __builtin_ppc_speculation_barrier.
19 2018-03-05 Jakub Jelinek <jakub@redhat.com>
22 * combine.c (combine_simplify_rtx): Don't try to simplify if
23 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
26 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
28 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
29 to 32 bytes when compiling for POWER9.
31 2018-03-05 Jakub Jelinek <jakub@redhat.com>
34 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
35 regparm >= 3 with no arg reg available also for calls with
36 flag_force_indirect_call. Pass decl to ix86_function_regparm.
39 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
41 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
43 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
46 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
48 2018-03-05 Richard Biener <rguenther@suse.de>
50 PR tree-optimization/84486
51 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
52 When inserting a __builtin_assume_aligned call set the LHS
53 SSA name alignment info accordingly.
55 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
57 PR tree-optimization/84114
58 * config/aarch64/aarch64.c (aarch64_reassociation_width)
59 Avoid reassociation of FLOAT_MODE addition.
61 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
63 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
64 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
65 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
66 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
67 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
68 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
69 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
71 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
72 __builtin_ia32_wbinvd): New builtins.
74 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
76 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
77 (ix86_valid_target_attribute_inner_p): Ditto.
78 (ix86_init_mmx_sse_builtins): Add special_args2.
79 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
80 TARGET_WBNOINVD_P): New.
81 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
82 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
83 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
84 * config/i386/immintrin.h (_wbinvd): New intrinsic.
85 * config/i386/pconfigintrin.h: New file.
86 * config/i386/wbnoinvdintrin.h: Ditto.
87 * config/i386/x86intrin.h: Add headers pconfigintrin.h and wbnoinvdintrin.h.
88 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
90 2018-03-05 Richard Biener <rguenther@suse.de>
92 PR tree-optimization/84670
93 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
95 (BB_VISITED_WITH_VISITED_SUCCS): New define.
96 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
97 (compute_antic_aux): Only assert the number of values in ANTIC_IN
98 doesn't grow if all successors (recursively) were visited at least
101 2018-03-05 Richard Biener <rguenther@suse.de>
103 PR tree-optimization/84650
104 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
105 if executed in the loop pipeline.
107 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
109 * doc/configfiles.texi (Configuration Files): Move info about
110 conditionalizing $target-protos.h to...
111 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
112 differs from $target-protos.h.
114 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
115 Chung-Ju Wu <jasonwucj@gmail.com>
117 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
118 * config/nds32/nds32-multiple.md (setmemsi): Define.
119 * config/nds32/nds32-memory-manipulation.c
120 (nds32_gen_dup_4_byte_to_word_value): New.
121 (emit_setmem_word_loop): New.
122 (emit_setmem_byte_loop): New.
123 (nds32_expand_setmem_loop): New.
124 (nds32_expand_setmem_loop_v3m): New.
125 (nds32_expand_setmem_unroll): New.
126 (nds32_expand_setmem): New.
128 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
129 Chung-Ju Wu <jasonwucj@gmail.com>
131 * config/nds32/nds32-memory-manipulation.c
132 (nds32_emit_load_store): New.
133 (nds32_emit_post_inc_load_store): New.
134 (nds32_emit_mem_move): New.
135 (nds32_emit_mem_move_block): New.
136 (nds32_expand_movmemsi_loop_unknown_size): New.
137 (nds32_expand_movmemsi_loop_known_size): New.
138 (nds32_expand_movmemsi_loop): New.
139 (nds32_expand_movmemsi_unroll): New.
140 (nds32_expand_movmemqi): Rename ...
141 (nds32_expand_movmemsi): ... to this.
142 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
143 (movmemsi): ... to this.
144 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
145 (nds32_expand_movmemsi): ... to this.
147 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
148 Monk Chiang <sh.chiang04@gmail.com>
149 Chung-Ju Wu <jasonwucj@gmail.com>
151 * config/nds32/nds32-protos.h
152 (nds32_expand_load_multiple): New arguments.
153 (nds32_expand_store_multiple): Ditto.
154 (nds32_valid_multiple_load_store): Rename ...
155 (nds32_valid_multiple_load_store_p): ... to this.
156 * config/nds32/nds32-memory-manipulation.c
157 (nds32_expand_load_multiple): Refine implementation.
158 (nds32_expand_store_multiple): Ditto.
159 * config/nds32/nds32-multiple.md
160 (load_multiple): Update nds32_expand_load_multiple interface.
161 (store_multiple): Update nds32_expand_store_multiple interface.
162 * config/nds32/nds32-predicates.c
163 (nds32_valid_multiple_load_store): Rename ...
164 (nds32_valid_multiple_load_store_p): ... to this and refine
166 * config/nds32/predicates.md
167 (nds32_load_multiple_and_update_address_operation): New predicate.
168 (nds32_store_multiple_and_update_address_operation): New predicate.
170 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
171 Chung-Ju Wu <jasonwucj@gmail.com>
173 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
174 (combo): New attribute.
175 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
177 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
179 * config/nds32/nds32.opt: Change -mcmodel= default value.
181 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
182 Monk Chiang <sh.chiang04@gmail.com>
183 Chung-Ju Wu <jasonwucj@gmail.com>
185 * config/nds32/constants.md (unspec_element): New enum.
186 * config/nds32/constraints.md (Umw): New constraint.
187 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
188 * config/nds32/nds32-intrinsic.md: Likewise.
189 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
190 (nds32_valid_smw_lwm_base_p): New.
191 (nds32_output_smw_single_word): New.
192 (nds32_output_lmw_single_word): New.
193 (nds32_expand_unaligned_load): New.
194 (nds32_expand_unaligned_store): New.
195 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
196 (nds32_output_smw_single_word): Declare.
197 (nds32_output_lmw_single_word): Declare.
198 (nds32_expand_unaligned_load): Declare.
199 (nds32_expand_unaligned_store): Declare.
200 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
201 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
202 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
203 NDS32_BUILTIN_UASTORE_DW.
204 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
207 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
208 Kito Cheng <kito.cheng@gmail.com>
209 Chung-Ju Wu <jasonwucj@gmail.com>
211 * config/nds32/nds32-intrinsic.c
212 (nds32_expand_builtin_null_ftype_reg): Delete.
213 (nds32_expand_builtin_reg_ftype_imm): Ditto.
214 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
215 (nds32_read_argument): New.
216 (nds32_legitimize_target): Ditto.
217 (nds32_legitimize_argument): Ditto.
218 (nds32_check_constant_argument): Ditto.
219 (nds32_expand_unop_builtin): Ditto.
220 (nds32_expand_unopimm_builtin): Ditto.
221 (nds32_expand_binop_builtin): Ditto.
222 (nds32_builtin_decl_impl): Ditto.
223 (builtin_description): Ditto.
224 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
225 (nds32_init_builtins_impl): Ditto.
226 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
227 (nds32_builtin_decl): New.
228 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
229 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
231 2018-03-02 Jeff Law <law@redhat.com>
233 * reorg.c (stop_search_p): Handle DEBUG_INSN.
234 (redundant_insn, fill_simple_delay_slots): Likewise.
235 (fill_slots_from_thread): Likewise.
236 * resource.c (mark_referenced_resources): Likewise.
237 (mark_set_resources, find_dead_or_set_registers): Likewise.
239 2018-03-02 Jakub Jelinek <jakub@redhat.com>
241 * substring-locations.h (format_warning_va): Formatting fix for
243 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
245 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
246 * substring-locations.c: Include intl.h.
247 (format_warning_va): Turned into small wrapper around
248 format_warning_n_va, renamed to ...
249 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
250 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
252 (format_warning_at_substring_n): New function.
253 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
254 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
255 format_warning_at_substring with just a shorter name instead of
256 const function pointer.
257 (fmtwarn_n): New function.
258 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
259 appropriate, get rid of all the fmtstr temporaries, move conditionals
260 with G_() wrapped string literals directly into fmtwarn arguments,
261 cast dir.len to (int), formatting fixes.
263 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
265 * doc/invoke.texi: Remove "Cilk Plus" references.
267 2018-03-02 Jakub Jelinek <jakub@redhat.com>
268 Richard Biener <rguenther@suse.de>
271 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
272 for error or warning attributes if CALL_FROM_THUNK_P is set.
275 2018-03-02 Jakub Jelinek <jakub@redhat.com>
278 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
279 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
282 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
283 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
285 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
286 instead of -1U in last predictors element's probability member.
288 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
291 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
292 arguments if they are comparable.
294 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
296 PR tree-optimization/84634
297 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
298 masks and masked_loop_p with a single loop_masks, making sure it's
299 null for bb vectorization.
301 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
303 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
304 (vect_analyze_data_ref_access): Use loop->safe_len rather than
305 loop->force_vectorize to check whether there is no alias.
307 2018-03-02 Jakub Jelinek <jakub@redhat.com>
310 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
312 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
314 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
315 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
316 instead of a loop around prev_real_insn.
317 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
321 * config/i386/i386.c (ix86_print_operand): Use conditional
322 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
325 2018-03-02 Richard Biener <rguenther@suse.de>
327 PR tree-optimization/84427
328 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
329 (bitmap_set_subtract_values): Rewrite to handle multiple
332 (prune_clobbered_mems): Likewise.
333 (phi_translate): Take edge instead of pred/phiblock.
334 (phi_translate_1): Likewise.
335 (phi_translate_set): Likewise. Insert all translated
336 exprs for a value into the set, keeping possibly multiple
337 expressions per value.
338 (compute_antic_aux): Adjust for phi_translate changes.
339 When intersecting union the expressions and prune those
340 not in the final value set, keeping possibly multiple
341 expressions per value. Do not use value-insertion
342 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
343 all expressions. Add verification that the value-sets
344 only shrink during iteration.
345 (compute_partial_antic_aux): Adjust for the phi_translate changes.
346 (do_pre_regular_insertion): Likewise.
347 (do_pre_partial_partial_insertion): Likewise.
349 2018-03-02 Richard Biener <rguenther@suse.de>
352 * config/darwin.c (saved_debug_info_level): New static global.
353 (darwin_asm_lto_start): Disable debug info generation for LTO out.
354 (darwin_asm_lto_end): Restore debug info generation settings.
356 2018-03-01 Martin Liska <mliska@suse.cz>
359 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
362 2018-03-01 Richard Biener <rguenther@suse.de>
365 * dwarf2out.c (gen_variable_die): Properly handle late VLA
366 type annotation with LTO when debug was disabled at compile-time.
368 2018-03-01 Matthew Fortune <mfortune@gmail.com>
370 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
372 (mips_final_postscan_insn): Likewise.
374 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
376 PR rtl-optimization/84528
377 * alias.c (init_alias_target): Add commentary.
378 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
379 a unique base value if the frame pointer is not eliminated
380 to the stack pointer.
382 2018-03-01 Tom de Vries <tom@codesourcery.com>
384 PR rtl-optimization/83327
385 * lra-int.h (hard_regs_spilled_into): Declare.
386 * lra.c (hard_regs_spilled_into): Define.
387 (init_reg_info): Init hard_regs_spilled_into.
388 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
389 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
390 (process_bb_lives): Handle hard_regs_spilled_into.
391 (lra_create_live_ranges_1): Before doing liveness propagation, clear
392 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
394 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
396 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
397 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
398 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
399 * config/rs6000/aix72.h: New file.
401 2018-02-28 Jakub Jelinek <jakub@redhat.com>
403 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
404 instead of warning_at with conditional singular and plural messages
408 * stor-layout.c (update_alignment_for_field): For
409 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
410 && !DECL_PACKED (field), do the alignment update, just use
411 only desired_align instead of MAX (type_align, desired_align)
413 (place_field): Don't do known_align < desired_align handling
414 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
415 is non-NULL, instead do it after rli->prev_field handling and
416 only if not within a bitfield word. For DECL_PACKED (field)
417 use type_align of BITS_PER_UNIT.
419 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
421 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
422 superfluous parentheses and trailing spaces.
424 2018-02-28 Richard Biener <rguenther@suse.de>
426 PR tree-optimization/84584
427 * graphite-scop-detection.c (scop_detection::add_scop): Discard
428 SCoPs with fake exit edge.
430 2018-02-28 Martin Liska <mliska@suse.cz>
433 * timevar.c (timer::print): Fix format to properly print 100%
436 2018-02-28 Richard Biener <rguenther@suse.de>
439 * genmatch.c (capture_info::walk_match): Do not mark
440 captured expressions without operands as expr_p given
441 they act more like predicates and should be subject to
442 "lost tail" side-effect preserving.
444 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
446 PR rtl-optimization/81611
447 * auto-inc-dec.c (attempt_change): Move dead note from
448 mem_insn if it's the next use of regno
449 (find_address): Take address use of reg holding
450 non-incremented value. Add parm to limit search to the named
452 (merge_in_block): Attempt to use a mem insn that is the next
453 use of the original regno.
455 2018-02-27 Martin Sebor <msebor@redhat.com>
458 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
459 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
461 2018-02-27 Martin Sebor <msebor@redhat.com>
464 * diagnostic-core.h (warning_n, error_n, inform_n): Change
465 n argument to unsigned HOST_WIDE_INT.
466 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
467 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
468 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
469 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
471 2018-02-27 Richard Biener <rguenther@suse.de>
473 PR tree-optimization/84512
474 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
475 Do not use the estimate returned from record_stmt_cost for
476 the scalar iteration cost but sum properly using add_stmt_cost.
478 2018-02-27 Richard Biener <rguenther@suse.de>
480 PR tree-optimization/84466
481 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
482 Adjust last change to less strictly validate use operands.
484 2018-02-27 Martin Liska <mliska@suse.cz>
486 PR gcov-profile/84548
487 * gcov.c (process_file): Allow partial overlap and consider it
488 also as group functions.
489 (output_lines): Properly calculate range of lines for a group.
491 2018-02-27 Martin Liska <mliska@suse.cz>
493 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
494 'ggc' suffixes. Change first column width.
495 (timer::print): Fix formatting of the column.
497 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
499 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
500 preserve inline entry blocks for the sake of debug inline
501 entry point markers alone.
502 (remove_unused_locals): Suggest in comments a better place to
503 force the preservation of inline entry blocks that are
504 otherwise unused, but do not preserve them.
506 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
508 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
510 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
513 * config/i386/constraints.md (Bs): Replace
514 ix86_indirect_branch_register with
515 TARGET_INDIRECT_BRANCH_REGISTER.
517 * config/i386/i386.md (indirect_jump): Likewise.
518 (tablejump): Likewise.
519 (*sibcall_memory): Likewise.
520 (*sibcall_value_memory): Likewise.
521 Peepholes of indirect call and jump via memory: Likewise.
522 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
523 (*sibcall_value_GOT_32): Likewise.
524 * config/i386/predicates.md (indirect_branch_operand): Likewise.
525 (GOT_memory_operand): Likewise.
526 (call_insn_operand): Likewise.
527 (sibcall_insn_operand): Likewise.
528 (GOT32_symbol_operand): Likewise.
529 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
531 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
533 PR rtl-optimization/83496
534 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
535 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
536 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
537 redundant insn, if any.
538 (relax_delay_slots): Likewise.
539 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
541 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
543 PR tree-optimization/83965
544 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
545 that grouped statements are part of a reduction chain. Return
546 true if the statement is not marked as a reduction itself but
548 (vect_recog_dot_prod_pattern): Don't check whether the statement
549 is part of a group here.
550 (vect_recog_sad_pattern): Likewise.
551 (vect_recog_widen_sum_pattern): Likewise.
553 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
556 * final.c (rest_of_clean_state): Also look for calls inside sequences.
558 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
561 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
563 (ix86_output_indirect_function_return): New prototype.
564 (ix86_split_simple_return_pop_internal): Likewise.
565 * config/i386/i386.c (indirect_return_via_cx): New.
566 (indirect_return_via_cx_bnd): Likewise.
567 (indirect_thunk_name): Handle return va CX_REG.
568 (output_indirect_thunk_function): Create alias for
569 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
570 (ix86_output_indirect_jmp): Remove the bool argument.
571 (ix86_output_indirect_function_return): New function.
572 (ix86_split_simple_return_pop_internal): Likewise.
573 * config/i386/i386.md (*indirect_jump): Don't pass false
574 to ix86_output_indirect_jmp.
575 (*tablejump_1): Likewise.
576 (simple_return_pop_internal): Change it to define_insn_and_split.
577 Call ix86_split_simple_return_pop_internal to split it for
579 (simple_return_indirect_internal): Call
580 ix86_output_indirect_function_return instead of
581 ix86_output_indirect_jmp.
583 2018-02-26 Jakub Jelinek <jakub@redhat.com>
586 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
587 memset and value initialization afterwards.
589 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
591 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
593 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
596 * common/config/aarch64/aarch64-common.c
597 (aarch_option_optimization_table[]): Switch
598 off fomit-frame-pointer
600 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
601 Chung-Ju Wu <jasonwucj@gmail.com>
603 * config/nds32/nds32-multiple.md (load_multiple): Disallow
605 (store_multiple): Ditto.
607 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
609 * config.gcc: Add --with-cpu support for nds32 target.
610 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
611 * config/nds32/nds32.opt: Add -mcpu= option.
613 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
615 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
616 isel=yes): Warn for these deprecated options.
618 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
620 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
621 ISA_2_5_MASKS_EMBEDDED.
623 2018-02-23 Jakub Jelinek <jakub@redhat.com>
625 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
626 p->max as pointers rather than using iterative_hash_expr.
628 2018-02-23 Carl Love <cel@us.ibm.com>
630 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
631 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
633 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
634 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
635 P8V_BUILTIN_VEC_VUNSIGNED2.
637 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
640 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
641 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
643 * lra-constraints.c (curr_insn_transform): Set up
644 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
646 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
647 * lra-eliminations.c (spill_pseudos): Ditto.
648 (process_insn_for_elimination): Ditto.
649 * lra-lives.c (reg_early_clobber_p): Use the new macros.
650 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
651 LRA_NON_CLOBBERED_ALT.
653 2018-02-22 Martin Sebor <msebor@redhat.com>
655 PR tree-optimization/84480
656 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
657 to maybe_diag_stxncpy_trunc. Call it.
658 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
659 from gimple_fold_builtin_strcpy. Print inlining stack.
660 (handle_builtin_stxncpy): Print inlining stack.
661 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
663 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
666 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
667 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
668 and -fcheck-pointer-bounds are used together.
669 (indirect_thunk_prefix): New enum.
670 (indirect_thunk_need_prefix): New function.
671 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
672 "_nt" instead of "_bnd" for NOTRACK prefix.
673 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
674 (output_indirect_thunk_function): Likewise.
676 (ix86_code_end): Update output_indirect_thunk_function calls.
677 (ix86_output_indirect_branch_via_reg): Replace
678 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
679 (ix86_output_indirect_branch_via_push): Likewise.
680 (ix86_output_function_return): Likewise.
681 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
682 incompatible with -fcf-protection=branch and
683 -fcheck-pointer-bounds.
685 2018-02-22 Steve Ellcey <sellcey@cavium.com>
688 * config/aarch64/aarch64.c (aarch64_print_address_internal):
689 Change gcc_assert call to output_operand_lossage.
691 2018-02-22 Steve Ellcey <sellcey@cavium.com>
693 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
695 2018-02-22 DJ Delorie <dj@redhat.com>
696 Sebastian Perta <sebastian.perta@renesas.com>
697 Oleg Endo <olegendo@gcc.gnu.org>
699 * config/rx/rx.c (rx_rtx_costs): New function.
700 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
702 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
704 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
706 2018-02-22 Martin Liska <mliska@suse.cz>
709 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
710 Add "native" as a possible value.
712 2018-02-22 Martin Liska <mliska@suse.cz>
715 * config/i386/i386.c (ix86_option_override_internal):
716 Add "native" as a possible value for -march and -mtune.
718 2018-02-22 Jakub Jelinek <jakub@redhat.com>
721 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
722 to all type variants.
724 PR tree-optimization/84503
725 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
726 width as info->bitpos + info->bitsize - start.
727 (merged_store_group::merge_overlapping): Simplify width computation.
728 (check_no_overlap): New function.
729 (imm_store_chain_info::try_coalesce_bswap): Compute expected
730 start + width and last_order of the group, fail if check_no_overlap
732 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
733 to group if check_no_overlap fails.
735 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
737 * config/rs6000/altivec.md: Delete contraint arguments to
738 define_expand, define_split, and define_peephole2, and in
739 define_insn_and_split if always unused.
740 * config/rs6000/darwin.md: Ditto.
741 * config/rs6000/dfp.md: Ditto.
742 * config/rs6000/rs6000.md: Ditto.
743 * config/rs6000/sync.md: Ditto.
744 * config/rs6000/vector.md: Ditto.
745 * config/rs6000/vsx.md: Ditto.
747 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
749 * config/rs6000/altivec.md: Write output control strings as braced
750 blocks instead of double-quoted strings.
751 * config/rs6000/darwin.md: Ditto.
752 * config/rs6000/rs6000.md: Ditto.
753 * config/rs6000/vector.md: Ditto.
754 * config/rs6000/vsx.md: Ditto.
756 2018-02-21 Jason Merrill <jason@redhat.com>
758 PR c++/84314 - ICE with templates and fastcall attribute.
759 * attribs.c (build_type_attribute_qual_variant): Remove assert.
761 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
763 * ipa-cp.c (determine_versionability): Fix comment typos.
765 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
768 * ipa-cp.c (determine_versionability): Do not version functions caling
771 2018-02-21 Martin Liska <mliska@suse.cz>
774 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
775 Add "native" as a possible value.
776 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
777 the macro when native cpu detection is available.
779 2018-02-21 Martin Liska <mliska@suse.cz>
782 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
783 Add "native" as a possible value.
784 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
785 when native cpu detection is available.
787 2018-02-21 Jakub Jelinek <jakub@redhat.com>
788 Martin Sebor <msebor@redhat.com>
790 PR tree-optimization/84478
791 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
793 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
794 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
795 support which is conservatively correct, for 2 only stay conservative
796 for maxlen. Formatting and comment capitalization fixes. Add STRICT
797 argument to the 2 argument get_range_strlen, adjust 6 arg
798 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
800 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
801 (gimple_fold_builtin_strlen): Pass true as last argument to
804 2018-02-20 Martin Sebor <msebor@redhat.com>
807 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
808 (builtin_memref::set_base_and_offset): Same. Handle inner references.
809 (builtin_memref::builtin_memref): Factor out parts into
810 set_base_and_offset and call it.
812 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
815 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
816 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
817 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
818 search at the associated MODE_INT.
820 2018-02-20 Jeff Law <law@redhat.com>
823 PR tree-optimization/81592
825 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
826 for range data rather than using global data.
827 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
828 range data rather than using global data.
829 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
830 pass it to children as needed.
831 (struct directive::fmtresult): Similarly.
832 (struct directive::set_width): Similarly.
833 (struct directive::set_precision): Similarly.
834 (format_integer, format_directive, parse_directive): Similarly.
835 (format_none): Accept unnamed vr_values parameter.
836 (format_percent, format_floating, format_character): Similarly.
837 (format_string, format_plain): Similarly.
838 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
839 the EVRP range analyzer for range data rather than using global data.
840 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
841 gimple-ssa-evrp-analyze.h
842 (class sprintf_dom_walker): Add after_dom_children member function.
843 Add evrp_range_analyzer member.
844 (sprintf_dom_walker::before_dom_children): Call into the EVRP
845 range analyzer as needed.
846 (sprintf_dom_walker::after_dom_children): New member function.
847 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
849 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
850 (evrp_range_analyzer::pop_to_marker): Likewise.
852 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
854 PR tree-optimization/84419
855 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
856 with the required type if its current type is compatible but
859 2018-02-20 Jakub Jelinek <jakub@redhat.com>
862 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
865 2018-02-20 Martin Liska <mliska@suse.cz>
868 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
869 possible values if we don't have a hint.
871 2018-02-20 Martin Liska <mliska@suse.cz>
875 * final.c (shorten_branches): Build align_tab array with one
877 * opts.c (finish_options): Add alignment option limit check.
878 (MAX_CODE_ALIGN): Likewise.
879 (MAX_CODE_ALIGN_VALUE): Likewise.
880 * doc/invoke.texi: Document maximum allowed option value for
881 all -falign-* options.
883 2018-02-19 Jakub Jelinek <jakub@redhat.com>
886 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
887 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
888 * var-tracking.c (emit_note_insn_var_location): Remove all references
889 to NOTE_INSN_CALL_ARG_LOCATION.
890 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
891 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
892 Use copy_rtx_if_shared.
893 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
894 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
895 (dwarf2out_var_location): Remove handling of
896 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
898 * final.c (final_scan_insn): Remove all references to
899 NOTE_INSN_CALL_ARG_LOCATION.
900 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
901 before dumping final insns.
902 * except.c (emit_note_eh_region_end): Remove all references to
903 NOTE_INSN_CALL_ARG_LOCATION.
904 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
905 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
906 * config/arc/arc.c (hwloop_optimize): Likewise.
907 * config/arm/arm.c (create_fix_barrier): Likewise.
908 * config/s390/s390.c (s390_chunkify_start): Likewise.
909 * config/sh/sh.c (find_barrier): Likewise.
910 * config/i386/i386.c (rest_of_insert_endbranch,
911 ix86_seh_fixup_eh_fallthru): Likewise.
912 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
913 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
914 * config/frv/frv.c (frv_function_prologue): Likewise.
915 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
917 (note_outside_basic_block_p): Remove all references to
918 NOTE_INSN_CALL_ARG_LOCATION.
919 * gengtype.c (adjust_field_rtx_def): Likewise.
920 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
922 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
923 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
926 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
929 PR tree-optimization/84452
930 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
931 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
934 2018-02-19 Martin Liska <mliska@suse.cz>
937 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
939 2018-02-19 Martin Liska <mliska@suse.cz>
940 Richard Sandiford <richard.sandiford@linaro.org>
942 PR tree-optimization/82491
943 * gimple-fold.c (get_base_constructor): Make earlier bail out
946 2018-02-19 Carl Love <cel@us.ibm.com>
948 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
949 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
951 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
954 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
956 * config/rl78/rl78.md (movdf): New define expand.
958 2018-02-19 Martin Liska <mliska@suse.cz>
961 * doc/invoke.texi: Fix typo.
962 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
964 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
966 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
967 handle rs6000_single_float and rs6000_double_float specially for
970 2018-02-16 Jeff Law <law@redhat.com>
972 * config/rx/rx.c (add_pop_cfi_notes): New function.;
975 2018-02-16 Jakub Jelinek <jakub@redhat.com>
978 * ipa-inline.c (inline_small_functions): Fix a typo.
980 2018-02-16 Nathan Sidwell <nathan@acm.org>
982 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
984 2018-02-16 Carl Love <cel@us.ibm.com>
986 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
987 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
988 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
989 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
990 expansion to P8V_BUILTIN_VEC_FLOAT2.
992 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
994 PR rtl-optimization/70023
995 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
996 src_regno into account.
998 2018-02-16 Carl Love <cel@us.ibm.com>
1000 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
1001 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
1002 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
1003 * config/rs6000/rs6000.c: Remove case statements for
1004 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
1005 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
1006 and P9V_BUILTIN_VEC_VINSERT4B.
1007 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
1008 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
1009 * config/rs6000/vsx.md:
1010 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
1013 2018-02-16 Carl Love <cel@us.ibm.com>
1015 * config/rs6000/altivec.h: Add builtin names vec_extract4b
1017 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
1019 * config/rs6000/rs6000-c.c: Add the definitions for
1020 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
1021 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
1022 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
1023 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
1024 definition for insert4b and define insn *insert3b_internal.
1025 * doc/extend.texi: Add documentation for vec_extract4b.
1027 2018-02-16 Nathan Sidwell <nathan@acm.org>
1029 * doc/extend.texi (Backwards Compatibility): Mention friend
1030 injection. Note for-scope is deprecated.
1031 * doc/invoke.texi (-ffriend-injection): Deprecate.
1033 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
1035 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
1036 that moved to I2, also allow destinations that are a paradoxical
1037 subreg (instead of a normal reg).
1039 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
1042 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
1045 2018-02-16 Richard Biener <rguenther@suse.de>
1047 PR tree-optimization/84037
1048 PR tree-optimization/84016
1050 * config/i386/i386.c (ix86_builtin_vectorization_cost):
1051 Adjust vec_construct for the fact we need additional higher latency
1052 128bit inserts for AVX256 and AVX512 vector builds.
1053 (ix86_add_stmt_cost): Scale vector construction cost for
1056 2018-02-16 Richard Biener <rguenther@suse.de>
1058 PR tree-optimization/84417
1059 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
1060 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
1061 (non_rewritable_lvalue_p): Likewise, use poly-ints.
1063 2018-02-16 Martin Liska <mliska@suse.cz>
1066 * internal-fn.def (ASAN_CHECK): Set proper flags.
1067 (ASAN_MARK): Likewise.
1069 2018-02-16 Julia Koval <julia.koval@intel.com>
1071 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
1072 from PTA_CANNONLAKE.
1074 2018-02-16 Jakub Jelinek <jakub@redhat.com>
1077 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
1078 Use ++iter rather than iter++ for std::list iterators.
1079 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
1080 defer deleting them until all nodes in the forest are processed. Do
1081 free even leaf nodes. Change to_process into auto_vec.
1084 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
1085 * vec.h (vec_default_construct): Use memset instead of placement new
1086 if BROKEN_VALUE_INITIALIZATION is defined.
1087 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
1088 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
1091 PR rtl-optimization/83723
1092 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
1093 * lra.c (lra_substitute_pseudo): Likewise. If true, use
1094 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
1096 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
1098 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
1100 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
1102 PR rtl-optimization/81443
1103 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
1104 from inner REGs to paradoxical SUBREGs.
1106 2018-02-16 Richard Biener <rguenther@suse.de>
1108 PR tree-optimization/84399
1109 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
1110 For operands we can analyze at their definition make sure we can
1111 analyze them at each use as well.
1113 2018-02-16 Richard Biener <rguenther@suse.de>
1115 PR tree-optimization/84190
1116 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
1117 volatile accesses if the decl isn't volatile.
1119 2018-02-15 Jason Merrill <jason@redhat.com>
1121 PR c++/84314 - ICE with templates and fastcall attribute.
1122 * attribs.c (build_type_attribute_qual_variant): Don't clobber
1123 TYPE_CANONICAL on an existing type.
1125 2018-02-15 Jakub Jelinek <jakub@redhat.com>
1127 PR tree-optimization/84383
1128 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
1129 dstoff nor call operand_equal_p if dstbase is NULL.
1131 PR tree-optimization/84334
1132 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
1133 also a CONSTANT_CLASS_P, punt.
1135 2018-02-14 Jim Wilson <jimw@sifive.com>
1137 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
1138 first SMALL_OPERAND check. New local min_second_step. Move assert
1139 to where locals are set. Add TARGET_RVC support.
1140 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
1142 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
1144 * doc/invoke.texi: Correct -Wformat-overflow code sample.
1146 2018-02-14 Martin Sebor <msebor@redhat.com>
1148 PR tree-optimization/83698
1149 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
1150 arrays constrain the offset range to their bounds.
1151 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
1152 (builtin_access::overlap): Avoid setting the size of overlap if it's
1154 (maybe_diag_overlap): Also consider arrays when deciding what values
1155 of offsets to include in diagnostics.
1157 2018-02-14 Martin Sebor <msebor@redhat.com>
1160 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
1161 that correspond to the kind of a declaration.
1163 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
1166 * config/pa/pa.md: Load address of PIC label using the linkage table
1167 if the label is nonlocal.
1169 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
1171 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
1172 warning message if user requests -maltivec=be.
1173 * doc/invoke.texi: Document deprecation of -maltivec=be.
1175 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
1178 * config/rs6000/rs6000-c.c: Update definitions for
1179 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
1180 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
1182 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
1185 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
1186 add _get_ssp intrinsics. Remove argument from
1187 __builtin_ia32_rdssp[d|q].
1188 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
1189 * config/i386/i386-builtin.def: Remove argument from
1190 __builtin_ia32_rdssp[d|q].
1191 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
1192 ix86_expand_special_args_builtin for _rdssp[d|q].
1193 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
1194 Clear register before usage.
1195 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
1196 Add documentation for new _get_ssp and _inc_ssp intrinsics.
1198 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
1200 PR tree-optimization/84357
1201 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
1202 operand 1 of an ARRAY_REF too.
1204 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
1207 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
1208 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1210 (set_of_reg): New struct.
1211 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
1212 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
1213 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
1215 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
1216 Split into bitclr, bitset, bitinvert patterns if appropriate.
1217 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
1218 use rx_fuse_in_memory_bitop.
1219 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
1220 to named insn, correct maximum insn length.
1222 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
1225 * machmode.def: Define a complex mode for PARTIAL_INT.
1226 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
1228 * doc/rtl.texi: Document CSPImode.
1229 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
1231 (msp430_hard_regno_nregs_with_padding): Likewise.
1233 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
1236 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
1238 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
1240 PR rtl-optimization/84169
1241 * combine.c (try_combine): New variable split_i2i3. Set it to true if
1242 we generated a parallel as new i3 and we split that to new i2 and i3
1243 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
1244 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
1245 those to i2, not i1. Partially rewrite this scan code.
1247 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1250 * stor-layout.c (place_field): For variable length fields, adjust
1251 offset_align afterwards not just based on the field's alignment,
1252 but also on the size.
1255 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
1256 of exps and logs in the use_exp2 case.
1258 2018-02-13 Jeff Law <law@redhat.com>
1260 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
1263 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
1266 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
1270 * common.opt (gas-loc-support, gas-locview-support): New.
1271 (ginline-points, ginternal-reset-location-views): New.
1272 * doc/invoke.texi: Document them. Use @itemx where intended.
1273 (gvariable-location-views): Adjust.
1274 * target.def (reset_location_view): New.
1275 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
1276 (TARGET_RESET_LOCATION_VIEW): New.
1277 * doc/tm.texi: Rebuilt.
1278 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
1279 (dwarf2out_default_as_locview_support): New.
1280 (output_asm_line_debug_info): Use option variables.
1281 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
1282 (output_loc_list): Likewise.
1283 (add_high_low_attributes): Check option variables.
1284 Don't output entry view attribute in strict mode.
1285 (gen_inlined_subroutine_die): Check option variables.
1286 (dwarf2out_inline_entry): Likewise.
1287 (init_sections_and_labels): Likewise.
1288 (dwarf2out_early_finish): Likewise.
1289 (maybe_reset_location_view): New, from...
1290 (dwarf2out_var_location): ... here. Call it.
1291 * debug.h (dwarf2out_default_as_loc_support): Declare.
1292 (dwarf2out_default_as_locview_support): Declare.
1293 * hooks.c (hook_int_rtx_insn_0): New.
1294 * hooks.h (hook_int_rtx_insn_0): Declare.
1295 * toplev.c (process_options): Take -gas-loc-support and
1296 -gas-locview-support from dwarf2out. Enable
1297 -gvariable-location-views by default only with locview
1298 assembler support. Enable -ginternal-reset-location-views by
1299 default only if the target defines the corresponding hook.
1300 Enable -ginline-points by default if location views are
1301 enabled; force it disabled if statement frontiers are
1303 * tree-inline.c (expand_call_inline): Check option variables.
1304 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
1306 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
1308 PR tree-optimization/84321
1309 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
1310 handling. Also check whether the anti-range contains any values
1311 that satisfy the mask; switch to a VR_RANGE if not.
1313 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
1316 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
1318 2018-02-13 Martin Jambor <mjambor@suse.cz>
1321 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
1322 of call statements, also set location of a load to a temporary.
1324 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
1326 * config/rl78/rl78.c (add_vector_labels): New function.
1327 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
1328 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
1329 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
1330 which checks that no arguments are passed.
1331 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
1332 * doc/extend.texi: Documentation for the new attribute.
1334 2018-02-13 Andreas Schwab <schwab@suse.de>
1336 * config/riscv/linux.h (CPP_SPEC): Define.
1338 2018-02-13 Jakub Jelinek <jakub@redhat.com>
1341 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
1342 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
1343 OPTION_MASK_ISA_AES as first argument to def_builtin_const
1344 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
1345 instead of OPTION_MASK_ISA_PCLMUL as first argument to
1346 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
1347 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
1348 temporarily for AES and PCLMUL builtins.
1350 PR tree-optimization/84339
1351 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
1352 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
1356 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
1357 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
1358 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
1360 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
1362 * omp-simd-clone.h: New file.
1363 * omp-simd-clone.c: Include omp-simd-clone.h.
1364 (expand_simd_clones): No longer static.
1365 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
1366 cgraph.h and omp-simd-clone.h.
1367 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
1368 (vect_recog_widen_shift_pattern): Formatting fix.
1369 (vect_pattern_recog_1): Don't check optab for calls.
1372 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
1373 operands[2] into a REG before using gen_lowpart on it.
1375 2018-02-12 Jeff Law <law@redhat.com>
1378 * config/sh/sh.c (find_barrier): Consider a sibling call
1381 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
1382 successfully back substituting a reg.
1384 2018-02-12 Richard Biener <rguenther@suse.de>
1386 PR tree-optimization/84037
1387 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
1388 parameter, move visited init to caller.
1389 (vect_slp_analyze_operations): Separate cost from validity
1390 check, initialize visited once for all instances.
1391 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
1393 * tree-vect-stmts.c (vect_model_simple_cost): Make early
1395 (vect_model_promotion_demotion_cost): Likewise.
1396 (vectorizable_bswap): Guard cost modeling with !slp_node
1397 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
1399 (vectorizable_call): Likewise.
1400 (vectorizable_conversion): Likewise.
1401 (vectorizable_assignment): Likewise.
1402 (vectorizable_shift): Likewise.
1403 (vectorizable_operation): Likewise.
1404 (vectorizable_store): Likewise.
1405 (vectorizable_load): Likewise.
1406 (vectorizable_condition): Likewise.
1407 (vectorizable_comparison): Likewise.
1409 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
1412 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
1413 (ASAN_MARK): Fix fnspec to account for return value, change pointer
1414 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
1416 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1419 * params.def (inline-min-speedup): Increase from 8 to 15.
1420 (max-inline-insns-auto): Decrease from 40 to 30.
1421 * ipa-split.c (consider_split): Add some buffer for function to
1422 be considered inlining candidate.
1423 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
1426 2018-02-12 Richard Biener <rguenther@suse.de>
1428 PR tree-optimization/84037
1429 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
1430 matched stmts if we cannot swap the non-matched ones.
1432 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1434 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
1435 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
1436 _mm_maskz_scalef_round_ss): New intrinsics.
1437 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
1438 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
1439 __builtin_ia32_scalefss_round): Remove.
1440 (__builtin_ia32_scalefsd_mask_round,
1441 __builtin_ia32_scalefss_mask_round): New intrinsics.
1442 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
1443 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
1444 ((match_operand:VF_128 2 "<round_nimm_predicate>"
1445 "<round_constraint>")): Changed to ...
1446 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
1447 "<round_scalar_constraint>")): ... this.
1448 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
1449 %0, %1, %2<round_op3>}"): Changed to ...
1450 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
1451 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
1452 %2<round_scalar_mask_op3>}"): ... this.
1453 * config/i386/subst.md (round_scalar_nimm_predicate): New.
1455 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
1457 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
1458 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
1459 (_mm_maskz_sqrt_round_ss): New intrinsics.
1460 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
1461 (__builtin_ia32_sqrtsd_mask_round)
1462 (__builtin_ia32_sqrtss_mask_round): New builtins.
1463 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
1464 (__builtin_ia32_sqrtss_round): Remove.
1465 (__builtin_ia32_sqrtsd_mask_round)
1466 (__builtin_ia32_sqrtss_mask_round): New builtins.
1467 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
1468 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
1469 ((match_operand:VF_128 1 "vector_operand"
1470 "xBm,<round_constraint>")): Changed to ...
1471 ((match_operand:VF_128 1 "vector_operand"
1472 "xBm,<round_scalar_constraint>")): ... this.
1473 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
1474 %0, %2, %<iptr>1<round_op3>}): Changed to ...
1475 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
1476 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
1477 %<iptr>1<round_scalar_mask_op3>}): ... this.
1478 ((set_attr "prefix" "<round_prefix>")): Changed to ...
1479 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
1481 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
1484 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
1485 Cast vec_cmpeq result to correct type.
1486 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
1487 Cast vec_cmpgt result to correct type.
1489 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
1491 * final.c (final_scan_insn_1): Renamed from...
1492 (final_scan_insn): ... this. New wrapper, to recover
1493 seen from the outermost call in recursive ones.
1494 * config/sparc/sparc.c (output_return): Drop seen from call.
1495 (output_sibcall): Likewise.
1496 * config/visium/visium.c (output_branch): Likewise.
1498 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
1500 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
1503 2018-02-10 Alan Modra <amodra@gmail.com>
1506 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
1507 Specify LR as an input.
1509 2018-02-10 Jakub Jelinek <jakub@redhat.com>
1512 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
1513 remove_member_access_dummy_vars): New functions.
1514 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
1515 lower_omp_1, execute_lower_omp): Use them.
1517 PR rtl-optimization/84308
1518 * shrink-wrap.c (spread_components): Release todo vector.
1520 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
1522 PR rtl-optimization/57193
1523 * ira-color.c (struct allocno_color_data): Add member
1524 conflict_allocno_hard_prefs.
1525 (update_conflict_allocno_hard_prefs): New.
1526 (bucket_allocno_compare_func): Add a preference based on
1527 conflict_allocno_hard_prefs.
1528 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
1529 (color_allocnos): Remove a dead code. Initiate
1530 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
1532 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1535 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
1536 constraint from =wa to wa. Avoid a subreg on the output operand,
1537 instead use a pseudo and subreg it in a move.
1538 (p9_xxbrd_<mode>): Changed to ...
1539 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
1540 (p9_xxbrd_v2df): New expander.
1541 (p9_xxbrw_<mode>): Changed to ...
1542 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
1543 (p9_xxbrw_v4sf): New expander.
1545 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1547 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
1549 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
1552 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
1553 multiply in 32-bit mode.
1554 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
1555 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
1558 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
1560 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
1561 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
1562 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
1563 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
1565 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
1568 * dwarf2out.c (is_trivial_indirect_ref): New function.
1569 (dwarf2out_late_global_decl): Do not generate a location
1570 attribute for variables that have a non-trivial DECL_VALUE_EXPR
1571 and that are not defined in the current unit.
1573 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1575 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
1576 instead of a libcall for UNORDERED.
1578 2018-02-09 Tamar Christina <tamar.christina@arm.com>
1581 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
1582 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
1584 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1587 * config/s390/s390.c (s390_set_current_function): Invoke
1588 s390_indirect_branch_settings also if fndecl didn't change.
1590 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1592 * config/rs6000/rs6000.md (blockage): Set length to zero.
1594 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
1596 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
1598 2018-02-09 Jakub Jelinek <jakub@redhat.com>
1601 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
1602 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
1606 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
1607 PARALLEL incoming that failed vt_get_decl_and_offset check.
1610 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
1611 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
1613 (get_variable_section): For decls in named .bss* sections pass true as
1614 second argument to bss_initializer_p.
1616 2018-02-09 Marek Polacek <polacek@redhat.com>
1617 Jakub Jelinek <jakub@redhat.com>
1620 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
1621 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
1622 Sync some changes from cxx_fold_indirect_ref.
1624 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
1626 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
1628 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
1629 (BLOCK_INLINE_ENTRY_LABEL): New.
1630 (dwarf2out_var_location): Disregard inline entry markers.
1631 (inline_entry_data): New struct.
1632 (inline_entry_data_hasher): New hashtable type.
1633 (inline_entry_data_hasher::hash): New.
1634 (inline_entry_data_hasher::equal): New.
1635 (inline_entry_data_table): New variable.
1636 (add_high_low_attributes): Add DW_AT_entry_pc and
1637 DW_AT_GNU_entry_view attributes if a pending entry is found
1638 in inline_entry_data_table. Add old entry_pc attribute only
1639 if debug nonbinding markers are disabled.
1640 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
1641 markers are enabled.
1642 (block_within_block_p, dwarf2out_inline_entry): New.
1643 (dwarf2out_finish): Check that no entries remained in
1644 inline_entry_data_table.
1645 * final.c (reemit_insn_block_notes): Handle inline entry notes.
1646 (final_scan_insn, notice_source_line): Likewise.
1647 (rest_of_clean_state): Skip inline entry markers.
1648 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
1650 * gimple.c (gimple_build_debug_inline_entry): New.
1651 * gimple.h (enum gimple_debug_subcode): Add
1652 GIMPLE_DEBUG_INLINE_ENTRY.
1653 (gimple_build_debug_inline_entry): Declare.
1654 (gimple_debug_inline_entry_p): New.
1655 (gimple_debug_nonbind_marker_p): Adjust.
1656 * insn-notes.def (INLINE_ENTRY): New.
1657 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
1658 inline entry marker notes.
1659 (print_insn): Likewise.
1660 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
1661 (INSN_DEBUG_MARKER_KIND): Likewise.
1662 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
1663 * tree-inline.c (expand_call_inline): Build and insert
1664 debug_inline_entry stmt.
1665 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
1666 inline entry blocks early, if nonbind markers are enabled.
1667 (dump_scope_block): Dump fragment info.
1668 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
1669 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
1670 (gimple_build_debug_inline_entry): New.
1671 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
1672 Enable/disable inline entry points too.
1673 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
1674 (DEBUG_INSN): Describe inline entry markers.
1676 * common.opt (gvariable-location-views): New.
1677 (gvariable-location-views=incompat5): New.
1678 * config.in: Rebuilt.
1679 * configure: Rebuilt.
1680 * configure.ac: Test assembler for view support.
1681 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
1682 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
1683 * dwarf2out.c (var_loc_view): New typedef.
1684 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
1685 (dwarf2out_locviews_in_attribute): New.
1686 (dwarf2out_locviews_in_loclist): New.
1687 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
1688 (enum dw_line_info_opcode): Add LI_adv_address.
1689 (struct dw_line_info_table): Add view.
1690 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
1691 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
1692 (zero_view_p): New variable.
1693 (ZERO_VIEW_P): New macro.
1694 (output_asm_line_debug_info): New.
1695 (struct var_loc_node): Add view.
1696 (add_AT_view_list, AT_loc_list): New.
1697 (add_var_loc_to_decl): Add view param. Test it against last.
1698 (new_loc_list): Add view params. Record them.
1699 (AT_loc_list_ptr): Handle loc and view lists.
1700 (view_list_to_loc_list_val_node): New.
1701 (print_dw_val): Handle dw_val_class_view_list.
1702 (size_of_die): Likewise.
1703 (value_format): Likewise.
1704 (loc_list_has_views): New.
1705 (gen_llsym): Set vl_symbol too.
1706 (maybe_gen_llsym, skip_loc_list_entry): New.
1707 (dwarf2out_maybe_output_loclist_view_pair): New.
1708 (output_loc_list): Output view list or entries too.
1709 (output_view_list_offset): New.
1710 (output_die): Handle dw_val_class_view_list.
1711 (output_dwarf_version): New.
1712 (output_compilation_unit_header): Use it.
1713 (output_skeleton_debug_sections): Likewise.
1714 (output_rnglists, output_line_info): Likewise.
1715 (output_pubnames, output_aranges): Update version comments.
1716 (output_one_line_info_table): Output view numbers in asm comments.
1717 (dw_loc_list): Determine current endview, pass it to new_loc_list.
1718 Call maybe_gen_llsym.
1719 (loc_list_from_tree_1): Adjust.
1720 (add_AT_location_description): Create view list attribute if
1721 needed, check it's absent otherwise.
1722 (convert_cfa_to_fb_loc_list): Adjust.
1723 (maybe_emit_file): Call output_asm_line_debug_info for test.
1724 (dwarf2out_var_location): Reset views as needed. Precompute
1725 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
1726 attribute. Set view.
1727 (new_line_info_table): Reset next view.
1728 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
1729 (dwarf2out_source_line): Likewise. Output view resets and labels to
1730 the assembler, or select appropriate line info opcodes.
1731 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
1732 (optimize_string_length): Catch it. Adjust.
1733 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
1734 dw_val_class_view_list, and remove it if no longer needed.
1735 (hash_loc_list): Hash view numbers.
1736 (loc_list_hasher::equal): Compare them.
1737 (optimize_location_lists): Check whether a view list symbol is
1738 needed, and whether the locview attribute is present, and
1739 whether they match. Remove the locview attribute if no longer
1741 (index_location_lists): Call skip_loc_list_entry for test.
1742 (dwarf2out_finish): Call output_asm_line_debug_info for test.
1743 Use output_dwarf_version.
1744 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
1745 (struct dw_val_node): Add val_view_list.
1746 * final.c (SEEN_NEXT_VIEW): New.
1747 (set_next_view_needed): New.
1748 (clear_next_view_needed): New.
1749 (maybe_output_next_view): New.
1750 (final_start_function): Rename to...
1751 (final_start_function_1): ... this. Take pointer to FIRST,
1752 add SEEN parameter. Emit param bindings in the initial view.
1753 (final_start_function): Reintroduce SEEN-less interface.
1754 (final): Rename to...
1755 (final_1): ... this. Take SEEN parameter. Output final pending
1756 next view at the end.
1757 (final): Reintroduce seen-less interface.
1758 (final_scan_insn): Output pending next view before switching
1759 sections or ending a block. Mark the next view as needed when
1760 outputting variable locations. Notify debug backend of section
1761 changes, and of location view changes.
1762 (rest_of_handle_final): Adjust.
1763 * toplev.c (process_options): Autodetect value for debug variable
1764 location views option. Warn on incompat5 without -gdwarf-5.
1765 * doc/invoke.texi (gvariable-location-views): New.
1766 (gvariable-location-views=incompat5): New.
1767 (gno-variable-location-views): New.
1769 2018-02-08 David Malcolm <dmalcolm@redhat.com>
1771 PR tree-optimization/84136
1772 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
1773 that the result of find_edge is non-NULL.
1775 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1778 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
1779 storing integer register in SImode. Fix cost of 256 and 512
1780 byte aligned SSE register store.
1782 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
1784 * config/i386/i386.c (ix86_multiplication_cost): Fix
1785 multiplication cost for TARGET_AVX512DQ.
1787 2018-02-08 Marek Polacek <polacek@redhat.com>
1789 PR tree-optimization/84238
1790 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
1793 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1795 PR tree-optimization/84265
1796 * tree-vect-stmts.c (vectorizable_store): Don't treat
1797 VMAT_CONTIGUOUS accesses as grouped.
1798 (vectorizable_load): Likewise.
1800 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1802 PR tree-optimization/81635
1803 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
1804 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
1805 (test_round_for_mask): New functions.
1806 (wide_int_cc_tests): Call test_round_for_mask.
1807 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
1808 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
1809 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
1810 range returned by get_range_info.
1812 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
1815 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
1816 * symtab.c: Include builtins.h
1817 (symtab_node::output_to_lto_symbol_table_p): Move here
1818 from lto-streamer-out.c:output_symbol_p.
1819 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
1820 (output_symbol_p): Move all logic to symtab.c
1821 (produce_symtab): Update.
1823 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
1825 * config/s390/s390-opts.h (enum indirect_branch): Define.
1826 * config/s390/s390-protos.h (s390_return_addr_from_memory)
1827 (s390_indirect_branch_via_thunk)
1828 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
1829 (enum s390_indirect_branch_type): Define.
1830 * config/s390/s390.c (struct s390_frame_layout, struct
1831 machine_function): Remove.
1832 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
1833 (indirect_branch_table_label_no, indirect_branch_table_name):
1835 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
1836 (enum s390_indirect_branch_option): Define.
1837 (s390_return_addr_from_memory): New function.
1838 (s390_handle_string_attribute): New function.
1839 (s390_attribute_table): Add new attribute handler.
1840 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
1841 (s390_indirect_branch_via_thunk): New function.
1842 (s390_indirect_branch_via_inline_thunk): New function.
1843 (s390_function_ok_for_sibcall): When jumping via thunk disallow
1844 sibling call optimization for non z10 compiles.
1845 (s390_emit_call): Force indirect branch target to be a single
1846 register. Add r1 clobber for non-z10 compiles.
1847 (s390_emit_epilogue): Emit return jump via return_use expander.
1848 (s390_reorg): Handle JUMP_INSNs as execute targets.
1849 (s390_option_override_internal): Perform validity checks for the
1850 new command line options.
1851 (s390_indirect_branch_attrvalue): New function.
1852 (s390_indirect_branch_settings): New function.
1853 (s390_set_current_function): Invoke s390_indirect_branch_settings.
1854 (s390_output_indirect_thunk_function): New function.
1855 (s390_code_end): Implement target hook.
1856 (s390_case_values_threshold): Implement target hook.
1857 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
1859 * config/s390/s390.h (struct s390_frame_layout)
1860 (struct machine_function): Move here from s390.c.
1861 (TARGET_INDIRECT_BRANCH_NOBP_RET)
1862 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
1863 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
1864 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
1865 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
1866 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
1867 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
1868 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
1869 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
1870 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
1871 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
1872 (mnemonic attribute): Add values which aren't recognized
1874 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
1875 pattern for branch conversion. Fix mnemonic attribute.
1876 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
1877 indirect branch via thunk if requested.
1878 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
1879 ("*indirect_jump"): Disable for branch conversion using out of
1881 ("indirect_jump_via_thunk<mode>_z10")
1882 ("indirect_jump_via_thunk<mode>")
1883 ("indirect_jump_via_inlinethunk<mode>_z10")
1884 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
1885 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
1886 ("casesi_jump_via_inlinethunk<mode>_z10")
1887 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
1888 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
1889 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
1890 ("*indirect2_jump"): Disable for branch conversion.
1891 ("casesi_jump"): Turn into expander and expand patterns for branch
1893 ("return_use"): New expander.
1894 ("*return"): Emit return via thunk and rename it to ...
1895 ("*return<mode>"): ... this one.
1896 * config/s390/s390.opt: Add new options and and enum for the
1899 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
1901 * lra-constraints.c (match_reload): Unconditionally use
1902 gen_lowpart_SUBREG, rather than selecting between that
1903 and equivalent gen_rtx_SUBREG code.
1905 2018-02-08 Richard Biener <rguenther@suse.de>
1907 PR tree-optimization/84233
1908 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
1909 changed flag instead of boguously re-using phi_inserted.
1911 2018-02-08 Martin Jambor <mjambor@suse.cz>
1913 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
1914 static local variables.
1916 2018-02-08 Richard Biener <rguenther@suse.de>
1918 PR tree-optimization/84278
1919 * tree-vect-stmts.c (vectorizable_store): When looking for
1920 smaller vector types to perform grouped strided loads/stores
1921 make sure the mode is supported by the target.
1922 (vectorizable_load): Likewise.
1924 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
1926 * config/aarch64/aarch64.c (aarch64_components_for_bb):
1927 Increase LDP/STP opportunities by adding adjacent callee-saves.
1929 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
1931 PR rtl-optimization/84068
1932 PR rtl-optimization/83459
1933 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
1935 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
1937 PR tree-optimization/84224
1938 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
1939 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
1942 2018-02-07 Iain Sandoe <iain@codesourcery.com>
1945 * config/rs6000/altivec.md (*restore_world): Remove LR use.
1946 * config/rs6000/predicates.md (restore_world_operation): Adjust op
1947 count, remove one USE.
1949 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
1951 * doc/install.texi (Configuration): Document the
1952 --with-long-double-format={ibm,ieee} PowerPC configuration
1956 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
1957 Convert from define_expand to be define_insn_and_split. Rework
1958 float/double/_Float128 conversions to QI/HI/SImode to work with
1959 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
1960 conversions to QI/HImode types did a store and then a load to
1961 truncate the value. For conversions to VSX registers, don't split
1962 the insn, instead emit the code directly. Use the code iterator
1963 any_fix to combine signed and unsigned conversions.
1964 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
1965 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
1966 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
1967 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
1968 (fix_<mode>di2_hw): Likewise.
1969 (fixuns_<mode>di2_hw): Likewise.
1970 (fix_<mode>si2_hw): Likewise.
1971 (fixuns_<mode>si2_hw): Likewise.
1972 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
1973 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
1974 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
1975 fix<uns>_trunc<SFDF:mode>si2_p8.
1976 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer
1978 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
1979 (fix<uns>_<mode>_mem): Likewise.
1980 (fctiw<u>z_<mode>_mem): Likewise.
1981 (fix<uns>_<mode>_mem): Likewise.
1982 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
1983 the register allocator from doing a direct move to the GPRs to do
1984 a store, and instead use the ISA 3.0 store byte/half-word from
1985 vector register instruction. For IEEE 128-bit floating point,
1986 also optimize stores of 32-bit ints.
1987 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
1989 2018-02-07 Alan Hayward <alan.hayward@arm.com>
1991 * genextract.c (push_pathstr_operand): New function to support
1993 (walk_rtx): Call push_pathstr_operand.
1994 (print_path): Support [a-zA-Z].
1996 2018-02-07 Richard Biener <rguenther@suse.de>
1998 PR tree-optimization/84037
1999 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
2000 (cse_and_gimplify_to_preheader): Declare.
2001 (vect_get_place_in_interleaving_chain): Likewise.
2002 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
2004 (_loop_vec_info::~_loop_vec_info): Delete it.
2005 (cse_and_gimplify_to_preheader): New function.
2006 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
2007 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
2008 (vectorizable_load): Likewise. For grouped stores always base
2009 the IV on the first element.
2010 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
2011 condition before gimplifying.
2013 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2015 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
2016 *DIV_EXPR and *MOD_EXPR.
2018 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
2021 * config/i386/i386.c (ix86_option_override_internal): Mask out
2022 the CF_SET bit when checking -fcf-protection.
2024 2018-02-07 Tom de Vries <tom@codesourcery.com>
2027 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
2030 2018-02-07 Richard Biener <rguenther@suse.de>
2032 PR tree-optimization/84204
2033 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
2036 PR tree-optimization/84205
2037 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
2038 special-case isl_ast_op_zdiv_r.
2040 PR tree-optimization/84223
2041 * graphite-scop-detection.c (gather_bbs::before_dom_children):
2042 Only add conditions from within the region.
2043 (gather_bbs::after_dom_children): Adjust.
2045 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
2048 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
2049 * config/avr/avr.md: Only post-reload split REG-REG moves if
2050 either register is REGERAL_REG_P.
2052 2018-02-07 Jakub Jelinek <jakub@redhat.com>
2054 PR tree-optimization/84235
2055 * tree-ssa-scopedtables.c
2056 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
2057 if the subtraction is performed in floating point type where NaNs are
2058 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
2059 build 1. Formatting fix.
2061 2018-02-06 Jakub Jelinek <jakub@redhat.com>
2064 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
2065 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
2066 and skip it regardless of bb boundaries. Use CALL_P macro,
2067 don't test INSN_P (insn) together with CALL_P or JUMP_P check
2068 unnecessarily, formatting fix.
2070 2018-02-06 Michael Collison <michael.collison@arm.com>
2072 * config/arm/thumb2.md:
2073 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
2074 (*thumb_mov_notscc): Ditto.
2076 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
2079 * config/rs6000/rs6000.md (su code attribute): Use "u" for
2080 unsigned_fix, not "s".
2082 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2084 * configure.ac (gcc_fn_eh_frame_ro): New function.
2085 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
2086 correct .eh_frame permissions.
2087 * configure: Regenerate.
2089 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
2091 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
2094 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2096 * config/rs6000/rs6000.c (rs6000_option_override_internal):
2097 Display warning message for -mno-speculate-indirect-jumps.
2099 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
2101 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
2103 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
2105 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
2107 PR tree-optimization/84225
2108 * tree-eh.c (find_trapping_overflow): Only call
2109 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
2111 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2114 * config/i386/i386.c: Reimplement the check of possible options
2115 -mibt/-mshstk conbination. Change error messages.
2116 * doc/invoke.texi: Fix a typo: remove extra '='.
2118 2018-02-06 Marek Polacek <polacek@redhat.com>
2120 PR tree-optimization/84228
2121 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
2123 2018-02-06 Tamar Christina <tamar.christina@arm.com>
2126 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
2127 emitted arch directives.
2128 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
2129 __ARM_FEATURE_COPROC before changing architectures.
2131 2018-02-06 Richard Biener <rguenther@suse.de>
2133 * config/i386/i386.c (print_reg): Fix typo.
2134 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
2136 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
2138 * configure: Regenerate.
2140 2018-02-05 Martin Sebor <msebor@redhat.com>
2142 PR tree-optimization/83369
2143 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
2146 2018-02-05 Martin Liska <mliska@suse.cz>
2148 * doc/invoke.texi: Cherry-pick upstream r323995.
2150 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
2152 * ira.c (ira_init_register_move_cost): Adjust comment.
2154 2018-02-05 Martin Liska <mliska@suse.cz>
2156 PR gcov-profile/84137
2157 * doc/gcov.texi: Fix typo in documentation.
2159 2018-02-05 Martin Liska <mliska@suse.cz>
2161 PR gcov-profile/83879
2162 * doc/gcov.texi: Document necessity of --dynamic-list-data when
2163 using dlopen functionality.
2165 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
2167 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
2168 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
2169 _mm_maskz_range_ss, _mm_mask_range_round_ss,
2170 _mm_maskz_range_round_ss): New intrinsics.
2171 (__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/i386-builtin.def (__builtin_ia32_rangesd128_round,
2176 __builtin_ia32_rangess128_round): Remove.
2177 (__builtin_ia32_rangesd128_mask_round,
2178 __builtin_ia32_rangess128_mask_round): New builtins.
2179 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
2180 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
2181 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
2182 "<round_saeonly_constraint>")): Changed to ...
2183 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
2184 "<round_saeonly_scalar_constraint>")): ... this.
2185 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
2186 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
2187 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
2188 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
2189 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
2191 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
2193 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
2195 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
2196 Remove all values except native, 8540 and 8548.
2198 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
2200 * config/i386/i386.c (ix86_output_function_return): Pass
2201 INVALID_REGNUM, instead of -1, as invalid register number to
2202 indirect_thunk_name and output_indirect_thunk.
2204 2018-02-02 Julia Koval <julia.koval@intel.com>
2206 * config.gcc: Add -march=icelake.
2207 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
2208 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
2209 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
2210 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
2211 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
2212 (processor_target_table): Add icelake.
2213 (ix86_option_override_internal): Handle new PTAs.
2214 (get_builtin_code_for_version): Handle icelake.
2215 (M_INTEL_COREI7_ICELAKE): New.
2216 (fold_builtin_cpu): Handle icelake.
2217 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
2218 * doc/invoke.texi: Add -march=icelake.
2220 2018-02-02 Julia Koval <julia.koval@intel.com>
2222 * config/i386/i386.c (ix86_option_override_internal): Change flags type
2223 to wide_int_bitmask.
2224 * wide-int-bitmask.h: New.
2226 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
2229 * config/i386/i386.md: Replace Pmode with word_mode in
2230 builtin_setjmp_setup and builtin_longjmp to support x32.
2232 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
2236 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
2238 (rs6000_supported_cpu_names): New static variable.
2239 (linux_cpu_translation_table): Likewise.
2240 (elf_platform) <cpu>: Define new static variable and use it.
2241 Translate kernel AT_PLATFORM name to canonical name if needed.
2242 Error if platform name is unknown.
2244 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
2247 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
2249 2018-02-01 Jeff Law <law@redhat.com>
2252 * config/i386/i386.c (release_scratch_register_on_entry): Add new
2253 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
2254 the scratch if RELEASE_VIA_POP is false.
2255 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
2256 If we have to save a temporary register, decrement SIZE appropriately.
2257 Pass new arguments to release_scratch_register_on_entry.
2258 (ix86_adjust_stack_and_probe): Likewise.
2259 (ix86_emit_probe_stack_range): Pass new arguments to
2260 release_scratch_register_on_entry.
2262 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
2264 PR rtl-optimization/84157
2265 * combine.c (change_zero_ext): Use REG_P predicate in
2266 front of HARD_REGISTER_P predicate.
2268 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
2270 * config/avr/avr.c (avr_option_override): Move disabling of
2271 -fdelete-null-pointer-checks to...
2272 * common/config/avr/avr-common.c (avr_option_optimization_table):
2275 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2277 PR tree-optimization/81635
2278 * tree-data-ref.c (split_constant_offset_1): For types that
2279 wrap on overflow, try to use range info to prove that wrapping
2282 2018-02-01 Renlin Li <renlin.li@arm.com>
2285 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
2287 (aarch64_register_move_cost): Likewise.
2288 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
2290 (REG_CLASS_NAMES): Likewise.
2291 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
2292 TAILCALL_ADDR_REGS. Remove IP registers.
2293 * config/aarch64/aarch64.md (Ucs): Update register constraint.
2295 2018-02-01 Richard Biener <rguenther@suse.de>
2297 * domwalk.h (dom_walker::dom_walker): Add additional constructor
2298 for specifying RPO order and allow NULL for that.
2299 * domwalk.c (dom_walker::dom_walker): Likewise.
2300 (dom_walker::walk): Handle NULL RPO order.
2301 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
2303 (rewrite_update_dom_walker): Likewise.
2304 (mark_def_dom_walker): Likewise.
2306 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2308 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
2309 (aarch64_maybe_expand_sve_subreg_move): Declare.
2310 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
2311 * config/aarch64/predicates.md (aarch64_any_register_operand): New
2313 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
2314 that are semantically a reverse operation.
2315 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
2316 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
2317 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
2319 (aarch64_can_change_mode_class): For big-endian, forbid changes
2320 between two SVE modes if they have different element sizes.
2322 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2324 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
2325 the TImode handling for big-endian targets.
2327 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2329 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
2330 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
2332 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
2333 Remove BSWAP handing for big-endian targets and use the form of
2334 LD1RQ appropariate for the mode.
2336 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2338 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
2339 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
2342 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
2345 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
2346 check for operands that need to go through aarch64_sve_reload_be.
2348 2018-02-01 Jakub Jelinek <jakub@redhat.com>
2350 PR tree-optimization/81661
2351 PR tree-optimization/84117
2352 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
2353 * tree-eh.c: Include gimplify.h.
2354 (find_trapping_overflow, replace_trapping_overflow,
2355 rewrite_to_non_trapping_overflow): New functions.
2356 * tree-vect-loop.c: Include tree-eh.h.
2357 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
2358 * tree-data-ref.c: Include tree-eh.h.
2359 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
2361 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
2363 PR rtl-optimization/84123
2364 * combine.c (change_zero_ext): Check if hard register satisfies
2365 can_change_dest_mode before calling gen_lowpart_SUBREG.
2367 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
2370 * ira.c (ira_init_register_move_cost): Remove assert.
2372 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2374 PR rtl-optimization/84071
2375 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
2376 * doc/tm.texi: Regenerate.
2378 2018-01-31 Richard Biener <rguenther@suse.de>
2380 PR tree-optimization/84132
2381 * tree-data-ref.c (analyze_miv_subscript): Properly
2382 check whether evolution_function_is_affine_multivariate_p
2383 before calling gcd_of_steps_may_divide_p.
2385 2018-01-31 Julia Koval <julia.koval@intel.com>
2388 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
2389 * config/i386/i386.md (rdpid_rex64) New.
2390 (rdpid): Make 32bit only.
2392 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
2395 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
2396 an IDENTIFIER_NODE for FUNCTION_TYPE's.
2398 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2401 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
2403 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
2405 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
2407 PR rtl-optimization/84071
2408 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
2409 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
2411 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2413 * config/arc/arc.c (arc_handle_aux_attribute): New function.
2414 (arc_attribute_table): Add 'aux' attribute.
2415 (arc_in_small_data_p): Consider aux like variables.
2416 (arc_is_aux_reg_p): New function.
2417 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
2418 (arc_get_aux_arg): New function.
2419 (prepare_move_operands): Handle aux-register access.
2420 (arc_handle_aux_attribute): New function.
2421 * doc/extend.texi (ARC Variable attributes): Add subsection.
2423 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
2425 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
2426 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
2427 (arc_attribute_table): Add 'uncached' attribute.
2428 (arc_print_operand): Print '.di' flag for uncached memory
2430 (arc_in_small_data_p): Do not consider for small data the uncached
2432 (arc_is_uncached_mem_p): New function.
2433 * config/arc/predicates.md (compact_store_memory_operand): Check
2434 for uncached memory accesses.
2435 (nonvol_nonimm_operand): Likewise.
2436 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
2438 2018-01-31 Jakub Jelinek <jakub@redhat.com>
2441 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
2442 falign-loops=): Add Optimization flag.
2444 2018-01-30 Jeff Law <law@redhat.com>
2447 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
2448 INT_REGISTERS_SAVED. Check it prior to calling
2449 get_scratch_register_on_entry.
2450 (ix86_adjust_stack_and_probe): Similarly.
2451 (ix86_emit_probe_stack_range): Similarly.
2452 (ix86_expand_prologue): Corresponding changes.
2454 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2457 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
2458 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
2460 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
2463 * lra-constraints.c (curr_insn_transform): Process AND in the
2466 2018-01-30 Jakub Jelinek <jakub@redhat.com>
2468 PR rtl-optimization/83986
2469 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
2470 dependence against last_pending_memory_flush in addition to
2473 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
2475 PR tree-optimization/81611
2476 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
2479 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
2482 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
2485 2018-01-30 Richard Biener <rguenther@suse.de>
2486 Jakub Jelinek <jakub@redhat.com>
2488 PR tree-optimization/84111
2489 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
2490 inner loops added during recursion, as they don't have up-to-date
2493 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2496 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
2497 (can_inline_edge_by_limits_p): ... here.
2498 (can_early_inline_edge_p, check_callers,
2499 update_caller_keys, update_callee_keys, recursive_inlining,
2500 add_new_edges_to_heap, speculation_useful_p,
2501 inline_small_functions,
2502 inline_small_functions, flatten_function,
2503 inline_to_all_callers_1): Update.
2505 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
2507 * profile-count.c (profile_count::combine_with_ipa_count): Handle
2510 2018-01-30 Richard Biener <rguenther@suse.de>
2512 PR tree-optimization/83008
2513 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
2514 invariant and constant vector uses in stmts when they need
2517 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2520 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
2521 * configure: Regenerate.
2523 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2525 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
2527 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
2528 Use gen_rtx_REG rather than gen_lowpart.
2530 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2532 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
2533 rather than 0 when creating partial subregs.
2535 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
2537 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
2540 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
2543 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
2544 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
2545 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
2546 flags. This restores the settings used before the 2017-07-24.
2547 Turning off pre increment/decrement/modify allows IVOPTS to
2548 optimize DF/SF loops where the index is an int.
2550 2018-01-29 Richard Biener <rguenther@suse.de>
2551 Kelvin Nilsen <kelvin@gcc.gnu.org>
2554 * tree-vect-stmts.c (vectorizable_call): Don't call
2555 targetm.vectorize_builtin_md_vectorized_function if callee is
2558 2018-01-22 Carl Love <cel@us.ibm.com>
2560 * doc/extend.tex: Fix typo in second arg in
2561 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
2563 2018-01-29 Richard Biener <rguenther@suse.de>
2565 PR tree-optimization/84086
2566 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
2567 (flush_ssaname_freelist): When SSA names were released reset
2568 the SCEV hash table.
2570 2018-01-29 Richard Biener <rguenther@suse.de>
2572 PR tree-optimization/84057
2573 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
2574 removed paths when removing edges.
2576 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
2578 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
2579 -mfunction-return=@var{choice}.
2581 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
2584 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
2586 (layout::print_source_line): Likewise.
2587 (test_get_line_width_without_trailing_whitespace): Add test cases.
2589 2018-01-27 Jakub Jelinek <jakub@redhat.com>
2592 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
2595 2018-01-26 Jim Wilson <jimw@sifive.com>
2597 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
2599 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
2602 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2604 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
2605 and CMP + SUB-immediate -> SUBS.
2607 2018-01-26 Martin Sebor <msebor@redhat.com>
2609 PR tree-optimization/83896
2610 * tree-ssa-strlen.c (get_string_len): Rename...
2611 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
2612 Avoid assuming length is constant.
2613 (handle_char_store): Use HOST_WIDE_INT for string length.
2615 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
2618 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
2619 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
2621 2018-01-26 Richard Biener <rguenther@suse.de>
2623 PR rtl-optimization/84003
2624 * dse.c (record_store): Only record redundant stores when
2625 the earlier store aliases at least all accesses the later one does.
2627 2018-01-26 Jakub Jelinek <jakub@redhat.com>
2629 PR rtl-optimization/83985
2630 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
2631 REG_CFA_RESTORE insns.
2632 (delete_unmarked_insns): Don't ignore separate shrink wrapping
2633 REG_CFA_RESTORE insns here.
2636 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
2637 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
2639 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2641 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
2642 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
2643 (arc_init): Likewise.
2644 (arc_override_options): Likewise.
2645 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
2647 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
2649 * config/arc/arc.h (TARGET_DBNZ): Define.
2650 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
2651 properly set the tune attribute.
2652 (dbnz): Use TARGET_DBNZ guard.
2653 * config/arc/arc.opt (mtune): Add core3 option.
2655 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2657 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
2658 recognize new pic like addresses.
2659 (arc_delegitimize_address): Clean up.
2661 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2663 * config/arc/arc-arches.def: Option mrf16 valid for all
2665 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
2666 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
2667 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
2668 * config/arc/arc-tables.opt: Regenerate.
2669 * config/arc/arc.c (arc_conditional_register_usage): Handle
2670 reduced register file case.
2671 (arc_file_start): Set must have build attributes.
2672 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
2674 * config/arc/arc.opt (mrf16): Add new option.
2675 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
2676 * config/arc/genmultilib.awk: Handle new mrf16 option.
2677 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
2678 * config/arc/t-multilib: Regenerate.
2679 * doc/invoke.texi (ARC Options): Document mrf16 option.
2681 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2683 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
2684 * config/arc/arc.c (arc_handle_secure_attribute): New function.
2685 (arc_attribute_table): Add 'secure_call' attribute.
2686 (arc_print_operand): Print secure call operand.
2687 (arc_function_ok_for_sibcall): Don't optimize tail calls when
2689 (arc_is_secure_call_p): New function. * config/arc/arc.md
2690 (call_i): Add support for sjli instruction.
2691 (call_value_i): Likewise.
2692 * config/arc/constraints.md (Csc): New constraint.
2694 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
2695 John Eric Martin <John.Martin@emmicro-us.com>
2697 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
2698 * config/arc/arc.c (_arc_jli_section): New struct.
2699 (arc_jli_section): New type.
2700 (rc_jli_sections): New static variable.
2701 (arc_handle_jli_attribute): New function.
2702 (arc_attribute_table): Add jli_always and jli_fixed attribute.
2703 (arc_file_end): New function.
2704 (TARGET_ASM_FILE_END): Define.
2705 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
2706 (arc_add_jli_section): New function.
2707 (jli_call_scan): Likewise.
2708 (arc_reorg): Call jli_call_scan.
2709 (arc_output_addsi): Remove 'S' from printing asm operand.
2710 (arc_is_jli_call_p): New function.
2711 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
2713 (movhi_insn): Likewise.
2714 (movsi_insn): Likewise.
2715 (movsi_set_cc_insn): Likewise.
2716 (loadqi_update): Likewise.
2717 (load_zeroextendqisi_update): Likewise.
2718 (load_signextendqisi_update): Likewise.
2719 (loadhi_update): Likewise.
2720 (load_zeroextendhisi_update): Likewise.
2721 (load_signextendhisi_update): Likewise.
2722 (loadsi_update): Likewise.
2723 (loadsf_update): Likewise.
2724 (movsicc_insn): Likewise.
2725 (bset_insn): Likewise.
2726 (bxor_insn): Likewise.
2727 (bclr_insn): Likewise.
2728 (bmsk_insn): Likewise.
2729 (bicsi3_insn): Likewise.
2730 (cmpsi_cc_c_insn): Likewise.
2731 (movsi_ne): Likewise.
2732 (movsi_cond_exec): Likewise.
2733 (clrsbsi2): Likewise.
2744 (call_i): Remove 'S' asm letter, add jli instruction.
2745 (call_value_i): Likewise.
2746 * config/arc/arc.op (mjli-always): New option.
2747 * config/arc/constraints.md (Cji): New constraint.
2748 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
2750 (subsf3_fpx): Likewise.
2751 (mulsf3_fpx): Likewise.
2752 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
2754 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
2756 * doc/invoke.texi (ARC): Document mjli-always option.
2758 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
2760 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
2761 avoid addition with 0 and use incw and decw where possible.
2763 2018-01-26 Richard Biener <rguenther@suse.de>
2765 PR tree-optimization/81082
2766 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
2767 association if it requires casting to unsigned.
2768 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
2769 from fold_plusminus_mult_expr to catch important cases late when
2770 range info is available.
2772 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2774 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
2775 * configure.ac (hidden_linkonce): New test.
2776 * configure: Regenerate.
2777 * config.in: Regenerate.
2779 2018-01-26 Julia Koval <julia.koval@intel.com>
2781 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
2782 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
2783 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
2784 _mm_mask_bitshuffle_epi64_mask): Fix type.
2785 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
2786 USI_FTYPE_V4DI_V4DI_USI): Remove.
2787 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
2788 __builtin_ia32_vpshufbitqmb256_mask,
2789 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
2790 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
2791 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
2793 2018-01-26 Alan Modra <amodra@gmail.com>
2796 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
2797 UNSPEC_VBPERMQ. Sort other unspecs.
2799 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
2801 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
2803 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
2806 * predict.c (drop_profile): Do not push/pop cfun; update also
2808 (handle_missing_profiles): Fix logic looking for zero profiles.
2810 2018-01-25 Jakub Jelinek <jakub@redhat.com>
2813 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
2814 on functions with #pragma omp declare simd or functions with simd
2816 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
2817 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
2818 Remove trailing \n from warning_at calls.
2820 2018-01-25 Tom de Vries <tom@codesourcery.com>
2823 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
2824 for neutered workers.
2826 2018-01-24 Joseph Myers <joseph@codesourcery.com>
2829 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
2830 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
2832 2018-01-24 Jeff Law <law@redhat.com>
2835 * i386.c (get_probe_interval): Move to earlier point.
2836 (ix86_compute_frame_layout): If -fstack-clash-protection and
2837 the frame is larger than the probe interval, then use pushes
2838 to save registers rather than reg->mem moves.
2839 (ix86_expand_prologue): Remove conditional for int_registers_saved
2842 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
2845 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
2846 min/max for never referenced object.
2848 2018-01-24 Jakub Jelinek <jakub@redhat.com>
2851 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
2853 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
2854 attributes from DECL_ATTRIBUTES (decl) without affecting
2855 DECL_ATTRIBUTES (current_function_decl).
2856 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
2857 functions with non-NULL DECL_ABSTRACT_ORIGIN.
2859 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
2861 PR tree-optimization/83979
2862 * fold-const.c (fold_comparison): Use constant_boolean_node
2863 instead of boolean_{true,false}_node.
2865 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
2867 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
2870 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
2872 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
2873 Simplify the clause that sets the length attribute.
2874 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
2875 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
2876 clause that sets the length attribute.
2877 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
2879 2018-01-24 Tom de Vries <tom@codesourcery.com>
2882 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
2883 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
2884 Add strict parameter.
2885 (prevent_branch_around_nothing): Insert dummy insn between branch to
2886 label and label with no ptx insn inbetween.
2887 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
2889 2018-01-24 Tom de Vries <tom@codesourcery.com>
2892 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
2893 for neutered threads in warp.
2894 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
2896 2018-01-24 Richard Biener <rguenther@suse.de>
2898 PR tree-optimization/83176
2899 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
2902 2018-01-24 Richard Biener <rguenther@suse.de>
2904 PR tree-optimization/82819
2905 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
2906 code generating pluses that are no-ops in the target precision.
2908 2018-01-24 Richard Biener <rguenther@suse.de>
2911 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
2913 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
2915 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
2916 to merge probabilities.
2917 * predict.c (probably_never_executed): Also mark as cold functions
2918 with global 0 profile and guessed local profile.
2919 * profile-count.c (profile_probability::combine_with_count): New
2921 * profile-count.h (profile_probability::operator*,
2922 profile_probability::operator*=, profile_probability::operator/,
2923 profile_probability::operator/=): Reduce precision to adjusted
2924 and set value to guessed on contradictory divisions.
2925 (profile_probability::combine_with_freq): Remove.
2926 (profile_probability::combine_wiht_count): Declare.
2927 (profile_count::force_nonzero):: Set to adjusted.
2928 (profile_count::probability_in):: Set quality to adjusted.
2929 * tree-ssa-tail-merge.c (replace_block_by): Use
2932 2018-01-23 Andrew Waterman <andrew@sifive.com>
2933 Jim Wilson <jimw@sifive.com>
2935 * config/riscv/riscv.c (riscv_stack_boundary): New.
2936 (riscv_option_override): Set riscv_stack_boundary. Handle
2937 riscv_preferred_stack_boundary_arg.
2938 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
2939 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
2940 (STACK_BOUNDARY): Set to riscv_stack_boundary.
2941 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
2942 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
2943 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
2945 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
2948 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
2949 of struct ix86_frame.
2950 (ix86_expand_epilogue): Likewise. Add a local variable for
2951 the reg_save_offset field in struct ix86_frame.
2953 2018-01-23 Bin Cheng <bin.cheng@arm.com>
2955 PR tree-optimization/82604
2956 * tree-loop-distribution.c (enum partition_kind): New enum item
2957 PKIND_PARTIAL_MEMSET.
2958 (partition_builtin_p): Support above new enum item.
2959 (generate_code_for_partition): Ditto.
2960 (compute_access_range): Differentiate cases that equality can be
2961 proven at all loops, the innermost loops or no loops.
2962 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
2963 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
2964 (finalize_partitions, distribute_loop): Don't fuse partition of
2965 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
2966 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
2969 2018-01-23 Martin Liska <mliska@suse.cz>
2971 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
2972 order to ignore the predictor.
2973 (PRED_POLYMORPHIC_CALL): Likewise.
2974 (PRED_RECURSIVE_CALL): Likewise.
2976 2018-01-23 Martin Liska <mliska@suse.cz>
2978 * tree-profile.c (tree_profiling): Print function header to
2979 aware reader which function we are working on.
2980 * value-prof.c (gimple_find_values_to_profile): Do not print
2981 not interesting value histograms.
2983 2018-01-23 Martin Liska <mliska@suse.cz>
2985 * profile-count.h (enum profile_quality): Add
2986 profile_uninitialized as the first value. Do not number values
2987 as they are zero based.
2988 (profile_count::verify): Update sanity check.
2989 (profile_probability::verify): Likewise.
2991 2018-01-23 Nathan Sidwell <nathan@acm.org>
2993 * doc/invoke.texi (ffor-scope): Deprecate.
2995 2018-01-23 David Malcolm <dmalcolm@redhat.com>
2997 PR tree-optimization/83510
2998 * domwalk.c (set_all_edges_as_executable): New function.
2999 (dom_walker::dom_walker): Convert bool param
3000 "skip_unreachable_blocks" to enum reachability. Move setup of
3001 edge flags to set_all_edges_as_executable and only do it when
3002 reachability is REACHABLE_BLOCKS.
3003 * domwalk.h (enum dom_walker::reachability): New enum.
3004 (dom_walker::dom_walker): Convert bool param
3005 "skip_unreachable_blocks" to enum reachability.
3006 (set_all_edges_as_executable): New decl.
3007 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
3008 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
3010 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
3011 but converting true to REACHABLE_BLOCKS.
3012 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
3014 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
3015 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
3016 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
3018 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
3019 if check_all_array_refs will be called.
3021 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3023 * tree.c (selftest::test_location_wrappers): Add more test
3026 2018-01-23 David Malcolm <dmalcolm@redhat.com>
3028 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
3029 (selftest::test_bit_in_range): Likewise.
3031 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3034 * doc/sourcebuild.texi (vect_float): Say that the selector
3035 only describes the situation when -funsafe-math-optimizations is on.
3036 (vect_float_strict): Document.
3038 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
3040 PR tree-optimization/83965
3041 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
3042 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
3043 instead of checking only for a reduction.
3044 (vect_recog_widen_sum_pattern): Likewise.
3046 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3048 * predict.c (probably_never_executed): Only use precise profile info.
3049 (compute_function_frequency): Skip after inlining hack since we now
3050 have quality checking.
3052 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
3054 * profile-count.h (profile_probability::very_unlikely,
3055 profile_probability::unlikely, profile_probability::even): Set
3056 precision to guessed.
3058 2018-01-23 Richard Biener <rguenther@suse.de>
3060 PR tree-optimization/83963
3061 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
3062 Properly terminate dominator walk when crossing the exit edge not
3063 when visiting its source block.
3065 2018-01-23 Jakub Jelinek <jakub@redhat.com>
3068 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
3069 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
3071 2018-01-22 Jakub Jelinek <jakub@redhat.com>
3073 PR tree-optimization/83957
3074 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
3075 semicolon after for body surrounded by braces.
3077 PR tree-optimization/83081
3078 * profile-count.h (profile_probability::split): New method.
3079 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
3080 Use profile_probability::split.
3081 (do_compare_rtx_and_jump): Fix adjustment of probabilities
3082 when splitting a single conditional jump into 2.
3084 2018-01-22 David Malcolm <dmalcolm@redhat.com>
3086 PR tree-optimization/69452
3087 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
3090 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3092 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
3093 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
3094 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
3096 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3098 * config/rl78/rl78-protos.h (rl78_split_movdi): New function declaration.
3099 * config/rl78/rl78.md (movdi): New define_expand.
3100 * config/rl78/rl78.c (rl78_split_movdi): New function.
3102 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
3105 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
3107 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
3108 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
3109 128-bit to produce an UNSPEC move to get the double word with the
3110 signbit and then a shift directly to do signbit.
3111 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
3112 implementation with a new version that just does either a direct
3113 move or a regular move. Move memory interface to separate insns.
3114 Move insns so they are next to the expander.
3115 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
3116 with signbit move. Split big and little endian case.
3117 (signbit<mode>2_dm_mem_le): Likewise.
3118 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
3119 (signbit<mode>2_dm2): Likewise.
3121 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3123 * config/rl78/rl78.md (anddi3): New define_expand.
3125 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3127 * config/rl78/rl78.md (umindi3): New define_expand.
3129 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3131 * config/rl78/rl78.md (smindi3): New define_expand.
3133 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3135 * config/rl78/rl78.md (smaxdi3): New define_expand.
3137 2018-01-22 Carl Love <cel@us.ibm.com>
3139 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
3140 LVX_V1TI): Add macro expansion.
3141 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
3142 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
3143 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
3144 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
3145 Change check to determine if the instruction is a byte reversing
3146 entry. Fix typo in comment.
3147 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
3148 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
3149 Add def_builtin calls for new builtins.
3150 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
3151 Add define_insn expansion.
3153 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3155 * config/rl78/rl78.md (umaxdi3): New define_expand.
3157 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
3159 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
3160 for non-QImode registers.
3162 2018-01-22 Richard Biener <rguenther@suse.de>
3164 PR tree-optimization/83963
3165 * graphite-scop-detection.c (scop_detection::get_sese): Delay
3166 including the loop exit block.
3167 (scop_detection::merge_sese): Likewise.
3168 (scop_detection::add_scop): Do it here instead.
3170 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3172 * doc/sourcebuild.texi (arm_softfloat): Document.
3174 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
3177 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
3178 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
3179 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
3181 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3182 David Edelsohn <dje.gcc@gmail.com>
3185 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3186 Change "crset eq" to "crset 2".
3187 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3188 (*call_indirect_aix<mode>_nospec): Likewise.
3189 (*call_value_indirect_aix<mode>_nospec): Likewise.
3190 (*call_indirect_elfv2<mode>_nospec): Likewise.
3191 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
3192 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
3193 change assembly output from . to $.
3194 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3195 (indirect_jump<mode>_nospec): Change assembly output from . to $.
3196 (*tablejump<mode>_internal1_nospec): Likewise.
3198 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
3201 * config/sh/sh_optimize_sett_clrt.cc:
3202 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
3204 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
3206 PR tree-optimization/83940
3207 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
3208 offset_dt to vect_constant_def rather than vect_unknown_def_type.
3209 (vect_check_load_store_mask): Add a mask_dt_out parameter and
3210 use it to pass back the definition type.
3211 (vect_check_store_rhs): Likewise rhs_dt_out.
3212 (vect_build_gather_load_calls): Add a mask_dt argument and use
3213 it instead of a call to vect_is_simple_use.
3214 (vectorizable_store): Update calls to vect_check_load_store_mask
3215 and vect_check_store_rhs. Use the dt returned by the latter instead
3216 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
3217 instead of calls to vect_is_simple_use. Pass the scalar rather
3218 than the vector operand to vect_is_simple_use when handling
3219 second and subsequent copies of an rhs value.
3220 (vectorizable_load): Update calls to vect_check_load_store_mask
3221 and vect_build_gather_load_calls. Use the cached mask_dt and
3222 gs_info.offset_dt instead of calls to vect_is_simple_use.
3224 2018-01-20 Jakub Jelinek <jakub@redhat.com>
3227 * tree-emutls.c: Include gimplify.h.
3228 (lower_emutls_2): New function.
3229 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
3230 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
3231 it before further processing.
3234 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
3235 UINTVAL (trueop1) instead of INTVAL (op1).
3237 2018-01-19 Jakub Jelinek <jakub@redhat.com>
3241 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
3242 INCOMING_FRAME_SP_OFFSET if not defined.
3243 (scan_trace): Add ENTRY argument. If true and
3244 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
3245 emit a note to adjust the CFA offset.
3246 (create_cfi_notes): Adjust scan_trace callers.
3247 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
3248 INCOMING_FRAME_SP_OFFSET in the CIE.
3249 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
3250 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
3252 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
3253 * doc/tm.texi: Regenerated.
3255 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
3257 PR rtl-optimization/83147
3258 * lra-constraints.c (remove_inheritance_pseudos): Use
3259 lra_substitute_pseudo_within_insn.
3261 2018-01-19 Tom de Vries <tom@codesourcery.com>
3262 Cesar Philippidis <cesar@codesourcery.com>
3265 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
3267 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
3270 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
3271 spaces for function labels.
3273 2018-01-19 Martin Liska <mliska@suse.cz>
3275 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
3276 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
3277 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
3278 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
3279 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
3280 (PRED_CONST_RETURN): Change from 69 to 65.
3281 (PRED_NULL_RETURN): Change from 91 to 71.
3282 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
3283 (PRED_LOOP_GUARD): Change from 66 to 73.
3285 2018-01-19 Martin Liska <mliska@suse.cz>
3287 * predict.c (predict_insn_def): Add new assert.
3288 (struct branch_predictor): Change type to signed integer.
3289 (test_prediction_value_range): Amend test to cover
3291 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
3292 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
3293 (PRED_LOOP_ITERATIONS_MAX): Likewise.
3294 (PRED_LOOP_IV_COMPARE): Likewise.
3295 * predict.h (PROB_UNINITIALIZED): Define new constant.
3297 2018-01-19 Martin Liska <mliska@suse.cz>
3299 * predict.c (dump_prediction): Add new format for
3300 analyze_brprob.py script which is enabled with -details
3302 * profile-count.h (precise_p): New function.
3304 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3306 PR tree-optimization/83922
3307 * tree-vect-loop.c (vect_verify_full_masking): Return false if
3308 there are no statements that need masking.
3309 (vect_active_double_reduction_p): New function.
3310 (vect_analyze_loop_operations): Use it when handling phis that
3311 are not in the loop header.
3313 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
3315 PR tree-optimization/83914
3316 * tree-vect-loop.c (vectorizable_induction): Don't convert
3317 init_expr or apply the peeling adjustment for inductions
3318 that are nested within the vectorized loop.
3320 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3322 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
3325 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3329 * function.h (gimplify_parameters): Add gimple_seq * argument.
3330 * function.c: Include gimple.h and options.h.
3331 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
3332 for the added local temporaries if needed.
3333 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
3334 if there are any parameter cleanups, wrap whole body into a
3335 try/finally with the cleanups.
3337 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
3340 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
3341 Use GET_MODE_CLASS for scalar floating point.
3343 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3347 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
3348 Fix call of call_cgraph_insertion_hooks.
3350 2018-01-18 Martin Sebor <msebor@redhat.com>
3352 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
3354 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
3357 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
3360 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
3363 * common.opt: (-ffile-prefix-map): New option.
3364 * opts.c (common_handle_option): Defer it.
3365 * opts-global.c (handle_common_deferred_options): Handle it.
3366 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
3367 * file-prefix-map.h: New file.
3368 (remap_debug_filename, add_debug_prefix_map): ...here.
3369 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
3370 * final.c (debug_prefix_map, add_debug_prefix_map
3371 remap_debug_filename): Move to...
3372 * file-prefix-map.c: New file.
3373 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
3374 generalize, get rid of alloca(), use strrchr() instead of strchr().
3375 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
3376 Implement in terms of add_prefix_map().
3377 (remap_macro_filename, remap_debug_filename): Implement in term of
3379 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
3380 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
3381 * dbxout.c: Include file-prefix-map.h.
3382 * varasm.c: Likewise.
3383 * vmsdbgout.c: Likewise.
3384 * xcoffout.c: Likewise.
3385 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
3386 * doc/cppopts.texi (-fmacro-prefix-map): Document.
3387 * doc/invoke.texi (-ffile-prefix-map): Document.
3388 (-fdebug-prefix-map): Update description.
3390 2018-01-18 Martin Liska <mliska@suse.cz>
3392 * config/i386/i386.c (indirect_thunk_name): Document that also
3394 (output_indirect_thunk): Document why both instructions
3395 (pause and lfence) are generated.
3397 2018-01-18 Richard Biener <rguenther@suse.de>
3399 PR tree-optimization/83887
3400 * graphite-scop-detection.c
3401 (scop_detection::get_nearest_dom_with_single_entry): Remove.
3402 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
3403 (scop_detection::merge_sese): Re-implement with a flood-fill
3404 algorithm that properly finds a SESE region if it exists.
3406 2018-01-18 Jakub Jelinek <jakub@redhat.com>
3409 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
3410 pointer_diff optimizations use view_convert instead of convert.
3412 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3414 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
3415 Generate different code for -mno-speculate-indirect-jumps.
3416 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
3417 (*call_indirect_aix<mode>): Disable for
3418 -mno-speculate-indirect-jumps.
3419 (*call_indirect_aix<mode>_nospec): New define_insn.
3420 (*call_value_indirect_aix<mode>): Disable for
3421 -mno-speculate-indirect-jumps.
3422 (*call_value_indirect_aix<mode>_nospec): New define_insn.
3423 (*sibcall_nonlocal_sysv<mode>): Generate different code for
3424 -mno-speculate-indirect-jumps.
3425 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
3427 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
3429 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
3430 long double type, set the flags for noting the default long double
3431 type, even if we don't pass or return a long double type.
3433 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
3436 * ipa-inline.c (flatten_function): Do not overwrite final inlining
3439 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
3441 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
3442 support for merge[hl].
3443 (fold_mergehl_helper): New helper function.
3444 (tree-vector-builder.h): New #include for tree_vector_builder usage.
3445 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
3446 (altivec_vmrglw_direct): Add xxmrglw insn.
3448 2018-01-17 Andrew Waterman <andrew@sifive.com>
3450 * config/riscv/riscv.c (riscv_conditional_register_usage): If
3451 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
3453 2018-01-17 David Malcolm <dmalcolm@redhat.com>
3456 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
3457 call the lto_location_cache before reading the
3458 DECL_SOURCE_LOCATION of the types.
3460 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
3461 Richard Sandiford <richard.sandiford@linaro.org>
3463 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
3464 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
3465 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
3466 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
3467 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
3469 * config/aarch64/constraints.md (aarch64_movti_operand):
3471 * config/aarch64/predicates.md (Uti): Add new constraint.
3473 2018-01-17 Carl Love <cel@us.ibm.com>
3475 * config/rs6000/vsx.md (define_expand xl_len_r,
3476 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
3477 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
3479 (define_expand, define_insn): Move the shift left from the
3480 define_insn to the define_expand for lxvl and stxvl instructions.
3481 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
3482 and XL_LEN_R definitions to PURE.
3484 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
3486 * config/i386/i386.c (indirect_thunk_name): Declare regno
3487 as unsigned int. Compare regno with INVALID_REGNUM.
3488 (output_indirect_thunk): Ditto.
3489 (output_indirect_thunk_function): Ditto.
3490 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
3491 in the call to output_indirect_thunk_function.
3493 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
3496 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
3497 rather than the size of inner_type to determine the stack slot size
3498 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
3500 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
3503 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
3506 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
3508 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
3509 endian Linux systems to optionally enable multilibs for selecting
3510 the long double type if the user configured an explicit type.
3511 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
3512 have no long double multilibs if not defined.
3513 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
3514 warn if the user used -mabi={ieee,ibm}longdouble and we built
3515 multilibs for long double.
3516 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
3517 appropriate multilib option.
3518 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
3520 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
3521 for building long double multilibs.
3522 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
3524 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
3526 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
3529 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
3531 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
3534 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
3537 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
3540 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3542 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
3543 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
3545 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
3547 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
3548 different rtl trees depending on TARGET_64BIT.
3549 (rs6000_gen_lvx): Likewise.
3551 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
3553 * config/visium/visium.md (nop): Tweak comment.
3554 (hazard_nop): Likewise.
3556 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
3558 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
3559 -mspeculate-indirect-jumps.
3560 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
3561 for -mno-speculate-indirect-jumps.
3562 (*call_indirect_elfv2<mode>_nospec): New define_insn.
3563 (*call_value_indirect_elfv2<mode>): Disable for
3564 -mno-speculate-indirect-jumps.
3565 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
3566 (indirect_jump): Emit different RTL for
3567 -mno-speculate-indirect-jumps.
3568 (*indirect_jump<mode>): Disable for
3569 -mno-speculate-indirect-jumps.
3570 (*indirect_jump<mode>_nospec): New define_insn.
3571 (tablejump): Emit different RTL for
3572 -mno-speculate-indirect-jumps.
3573 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
3574 (tablejumpsi_nospec): New define_expand.
3575 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
3576 (tablejumpdi_nospec): New define_expand.
3577 (*tablejump<mode>_internal1): Disable for
3578 -mno-speculate-indirect-jumps.
3579 (*tablejump<mode>_internal1_nospec): New define_insn.
3580 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
3583 2018-01-16 Artyom Skrobov tyomitch@gmail.com
3585 * caller-save.c (insert_save): Drop unnecessary parameter. All
3588 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3589 Richard Biener <rguenth@suse.de>
3592 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
3593 return early, inline manually is_gimple_sizepos. Make sure if we
3594 call gimplify_expr we don't end up with a gimple constant.
3595 * tree.c (variably_modified_type_p): Don't return true for
3596 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
3597 * gimplify.h (is_gimple_sizepos): Remove.
3599 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3601 PR tree-optimization/83857
3602 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
3603 vectorizable_live_operation for pure SLP statements.
3604 (vectorizable_live_operation): Handle PHIs.
3606 2018-01-16 Richard Biener <rguenther@suse.de>
3608 PR tree-optimization/83867
3609 * tree-vect-stmts.c (vect_transform_stmt): Precompute
3610 nested_in_vect_loop_p since the scalar stmt may get invalidated.
3612 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3615 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
3616 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
3617 If off is not INTEGER_CST, issue a may not be aligned warning
3618 rather than isn't aligned. Use isn%'t rather than isn't.
3619 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
3621 <case MULT_EXPR>: Improve the case when bottom and one of the
3622 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
3623 operand, in that case check if the other operand is multiple of
3624 bottom divided by the INTEGER_CST operand.
3626 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3629 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
3630 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
3631 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
3632 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
3633 * config/pa/pa.c (pa_function_arg_advance): Likewise.
3634 (pa_function_arg, pa_arg_partial_bytes): Likewise.
3635 (pa_function_arg_size): New function.
3637 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3639 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
3640 in a separate statement.
3642 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
3644 PR tree-optimization/83847
3645 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
3646 group gathers and scatters.
3648 2018-01-16 Jakub Jelinek <jakub@redhat.com>
3650 PR rtl-optimization/86620
3651 * params.def (max-sched-ready-insns): Bump minimum value to 1.
3653 PR rtl-optimization/83213
3654 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
3655 to last if both are JUMP_INSNs.
3657 PR tree-optimization/83843
3658 * gimple-ssa-store-merging.c
3659 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
3660 store_immediate_info for bswap/nop orig_stores.
3662 2018-01-15 Andrew Waterman <andrew@sifive.com>
3664 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
3666 <UDIV>: Increase cost if !TARGET_DIV.
3668 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
3670 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
3671 (define_attr "cr_logical_3op"): New.
3672 (cceq_ior_compare): Adjust.
3673 (cceq_ior_compare_complement): Adjust.
3674 (*cceq_rev_compare): Adjust.
3675 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
3676 (is_cracked_insn): Adjust.
3677 (insn_must_be_first_in_group): Adjust.
3678 * config/rs6000/40x.md: Adjust.
3679 * config/rs6000/440.md: Adjust.
3680 * config/rs6000/476.md: Adjust.
3681 * config/rs6000/601.md: Adjust.
3682 * config/rs6000/603.md: Adjust.
3683 * config/rs6000/6xx.md: Adjust.
3684 * config/rs6000/7450.md: Adjust.
3685 * config/rs6000/7xx.md: Adjust.
3686 * config/rs6000/8540.md: Adjust.
3687 * config/rs6000/cell.md: Adjust.
3688 * config/rs6000/e300c2c3.md: Adjust.
3689 * config/rs6000/e500mc.md: Adjust.
3690 * config/rs6000/e500mc64.md: Adjust.
3691 * config/rs6000/e5500.md: Adjust.
3692 * config/rs6000/e6500.md: Adjust.
3693 * config/rs6000/mpc.md: Adjust.
3694 * config/rs6000/power4.md: Adjust.
3695 * config/rs6000/power5.md: Adjust.
3696 * config/rs6000/power6.md: Adjust.
3697 * config/rs6000/power7.md: Adjust.
3698 * config/rs6000/power8.md: Adjust.
3699 * config/rs6000/power9.md: Adjust.
3700 * config/rs6000/rs64.md: Adjust.
3701 * config/rs6000/titan.md: Adjust.
3703 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3705 * config/i386/predicates.md (indirect_branch_operand): Rewrite
3706 ix86_indirect_branch_register logic.
3708 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3710 * config/i386/constraints.md (Bs): Update
3711 ix86_indirect_branch_register check. Don't check
3712 ix86_indirect_branch_register with GOT_memory_operand.
3714 * config/i386/predicates.md (GOT_memory_operand): Don't check
3715 ix86_indirect_branch_register here.
3716 (GOT32_symbol_operand): Likewise.
3718 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3720 * config/i386/predicates.md (constant_call_address_operand):
3721 Rewrite ix86_indirect_branch_register logic.
3722 (sibcall_insn_operand): Likewise.
3724 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3726 * config/i386/constraints.md (Bs): Replace
3727 ix86_indirect_branch_thunk_register with
3728 ix86_indirect_branch_register.
3730 * config/i386/i386.md (indirect_jump): Likewise.
3731 (tablejump): Likewise.
3732 (*sibcall_memory): Likewise.
3733 (*sibcall_value_memory): Likewise.
3734 Peepholes of indirect call and jump via memory: Likewise.
3735 * config/i386/i386.opt: Likewise.
3736 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3737 (GOT_memory_operand): Likewise.
3738 (call_insn_operand): Likewise.
3739 (sibcall_insn_operand): Likewise.
3740 (GOT32_symbol_operand): Likewise.
3742 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3745 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
3746 type rather than type addr's type points to.
3747 (expand_omp_atomic_mutex): Likewise.
3748 (expand_omp_atomic): Likewise.
3750 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
3753 * config/i386/i386.c (output_indirect_thunk_function): Use
3754 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
3755 for __x86_return_thunk.
3757 2018-01-15 Richard Biener <rguenther@suse.de>
3760 * expmed.c (extract_bit_field_1): Fix typo.
3762 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3765 * config/arm/iterators.md (VF): New mode iterator.
3766 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
3767 Remove integer-related logic from pattern.
3768 (neon_vabd<mode>_3): Likewise.
3770 2018-01-15 Jakub Jelinek <jakub@redhat.com>
3773 * common.opt (fstrict-overflow): No longer an alias.
3774 (fwrapv-pointer): New option.
3775 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
3776 also for pointer types based on flag_wrapv_pointer.
3777 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
3778 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
3779 opts->x_flag_wrapv got set.
3780 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
3781 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
3782 POINTER_TYPE_OVERFLOW_UNDEFINED.
3783 * match.pd: Likewise in address comparison pattern.
3784 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
3786 2018-01-15 Richard Biener <rguenther@suse.de>
3789 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
3790 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
3791 Reset type names to their identifier if their TYPE_DECL doesn't
3792 have linkage (and thus is used for ODR and devirt).
3793 (save_debug_info_for_decl): Remove.
3794 (save_debug_info_for_type): Likewise.
3795 (add_tree_to_fld_list): Adjust.
3796 * tree-pretty-print.c (dump_generic_node): Make dumping of
3797 type names more robust.
3799 2018-01-15 Richard Biener <rguenther@suse.de>
3801 * BASE-VER: Bump to 8.0.1.
3803 2018-01-14 Martin Sebor <msebor@redhat.com>
3806 * builtins.c (check_access): Avoid warning when the no-warning bit
3809 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
3811 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
3812 * ira-color (allocno_hard_regs_compare): Likewise.
3814 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
3817 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
3818 Use .pushsection/.popsection.
3820 2018-01-14 Martin Sebor <msebor@redhat.com>
3823 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
3825 2018-01-14 Jakub Jelinek <jakub@redhat.com>
3827 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
3828 entry from extra_headers.
3829 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
3830 extra_headers, make the list bitwise identical to the i?86-*-* one.
3832 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3834 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
3835 -mcmodel=large with -mindirect-branch=thunk,
3836 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
3837 -mfunction-return=thunk-extern.
3838 * doc/invoke.texi: Document -mcmodel=large is incompatible with
3839 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
3840 -mfunction-return=thunk and -mfunction-return=thunk-extern.
3842 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3844 * config/i386/i386.c (print_reg): Print the name of the full
3845 integer register without '%'.
3846 (ix86_print_operand): Handle 'V'.
3847 * doc/extend.texi: Document 'V' modifier.
3849 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3851 * config/i386/constraints.md (Bs): Disallow memory operand for
3852 -mindirect-branch-register.
3854 * config/i386/predicates.md (indirect_branch_operand): Likewise.
3855 (GOT_memory_operand): Likewise.
3856 (call_insn_operand): Likewise.
3857 (sibcall_insn_operand): Likewise.
3858 (GOT32_symbol_operand): Likewise.
3859 * config/i386/i386.md (indirect_jump): Call convert_memory_address
3860 for -mindirect-branch-register.
3861 (tablejump): Likewise.
3862 (*sibcall_memory): Likewise.
3863 (*sibcall_value_memory): Likewise.
3864 Disallow peepholes of indirect call and jump via memory for
3865 -mindirect-branch-register.
3866 (*call_pop): Replace m with Bw.
3867 (*call_value_pop): Likewise.
3868 (*sibcall_pop_memory): Replace m with Bs.
3869 * config/i386/i386.opt (mindirect-branch-register): New option.
3870 * doc/invoke.texi: Document -mindirect-branch-register option.
3872 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3874 * config/i386/i386-protos.h (ix86_output_function_return): New.
3875 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
3876 set function_return_type.
3877 (indirect_thunk_name): Add ret_p to indicate thunk for function
3879 (output_indirect_thunk_function): Pass false to
3880 indirect_thunk_name.
3881 (ix86_output_indirect_branch_via_reg): Likewise.
3882 (ix86_output_indirect_branch_via_push): Likewise.
3883 (output_indirect_thunk_function): Create alias for function
3884 return thunk if regno < 0.
3885 (ix86_output_function_return): New function.
3886 (ix86_handle_fndecl_attribute): Handle function_return.
3887 (ix86_attribute_table): Add function_return.
3888 * config/i386/i386.h (machine_function): Add
3889 function_return_type.
3890 * config/i386/i386.md (simple_return_internal): Use
3891 ix86_output_function_return.
3892 (simple_return_internal_long): Likewise.
3893 * config/i386/i386.opt (mfunction-return=): New option.
3894 (indirect_branch): Mention -mfunction-return=.
3895 * doc/extend.texi: Document function_return function attribute.
3896 * doc/invoke.texi: Document -mfunction-return= option.
3898 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
3900 * config/i386/i386-opts.h (indirect_branch): New.
3901 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
3902 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
3903 with local indirect jump when converting indirect call and jump.
3904 (ix86_set_indirect_branch_type): New.
3905 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
3906 (indirectlabelno): New.
3907 (indirect_thunk_needed): Likewise.
3908 (indirect_thunk_bnd_needed): Likewise.
3909 (indirect_thunks_used): Likewise.
3910 (indirect_thunks_bnd_used): Likewise.
3911 (INDIRECT_LABEL): Likewise.
3912 (indirect_thunk_name): Likewise.
3913 (output_indirect_thunk): Likewise.
3914 (output_indirect_thunk_function): Likewise.
3915 (ix86_output_indirect_branch_via_reg): Likewise.
3916 (ix86_output_indirect_branch_via_push): Likewise.
3917 (ix86_output_indirect_branch): Likewise.
3918 (ix86_output_indirect_jmp): Likewise.
3919 (ix86_code_end): Call output_indirect_thunk_function if needed.
3920 (ix86_output_call_insn): Call ix86_output_indirect_branch if
3922 (ix86_handle_fndecl_attribute): Handle indirect_branch.
3923 (ix86_attribute_table): Add indirect_branch.
3924 * config/i386/i386.h (machine_function): Add indirect_branch_type
3925 and has_local_indirect_jump.
3926 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
3928 (tablejump): Likewise.
3929 (*indirect_jump): Use ix86_output_indirect_jmp.
3930 (*tablejump_1): Likewise.
3931 (simple_return_indirect_internal): Likewise.
3932 * config/i386/i386.opt (mindirect-branch=): New option.
3933 (indirect_branch): New.
3936 (thunk-inline): Likewise.
3937 (thunk-extern): Likewise.
3938 * doc/extend.texi: Document indirect_branch function attribute.
3939 * doc/invoke.texi: Document -mindirect-branch= option.
3941 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
3944 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
3946 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
3948 * ipa-inline.c (want_inline_small_function_p): Return false if
3949 inlining has already failed with CIF_FINAL_ERROR.
3950 (update_caller_keys): Call want_inline_small_function_p before
3952 (update_callee_keys): Likewise.
3954 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
3956 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
3958 (rs6000_quadword_masked_address_p): Likewise.
3959 (quad_aligned_load_p): Likewise.
3960 (quad_aligned_store_p): Likewise.
3961 (const_load_sequence_p): Add comment to describe the outer-most loop.
3962 (mimic_memory_attributes_and_flags): New function.
3963 (rs6000_gen_stvx): Likewise.
3964 (replace_swapped_aligned_store): Likewise.
3965 (rs6000_gen_lvx): Likewise.
3966 (replace_swapped_aligned_load): Likewise.
3967 (replace_swapped_load_constant): Capitalize argument name in
3968 comment describing this function.
3969 (rs6000_analyze_swaps): Add a third pass to search for vector loads
3970 and stores that access quad-word aligned addresses and replace
3971 with stvx or lvx instructions when appropriate.
3972 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
3973 New function prototype.
3974 (rs6000_quadword_masked_address_p): Likewise.
3975 (rs6000_gen_lvx): Likewise.
3976 (rs6000_gen_stvx): Likewise.
3977 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
3978 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
3979 when memory address is aligned.
3980 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
3981 this split to select lvx instruction when memory address is aligned.
3982 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
3983 instruction when memory address is aligned.
3984 (*vsx_le_perm_load_v16qi): Likewise.
3985 (four unnamed splitters): Modify to select the stvx instruction
3986 when memory is aligned.
3988 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
3990 * predict.c (determine_unlikely_bbs): Handle correctly BBs
3991 which appears in the queue multiple times.
3993 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
3994 Alan Hayward <alan.hayward@arm.com>
3995 David Sherwood <david.sherwood@arm.com>
3997 * tree-vectorizer.h (vec_lower_bound): New structure.
3998 (_loop_vec_info): Add check_nonzero and lower_bounds.
3999 (LOOP_VINFO_CHECK_NONZERO): New macro.
4000 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
4001 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
4002 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
4003 fields. Make seg_len the distance travelled, not including the
4005 (dr_direction_indicator): Declare.
4006 (dr_zero_step_indicator): Likewise.
4007 (dr_known_forward_stride_p): Likewise.
4008 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
4010 (runtime_alias_check_p): Allow runtime alias checks with
4012 (operator ==): Compare access_size and align.
4013 (prune_runtime_alias_test_list): Rework for new distinction between
4014 the access_size and seg_len.
4015 (create_intersect_range_checks_index): Likewise. Cope with polynomial
4017 (get_segment_min_max): New function.
4018 (create_intersect_range_checks): Use it.
4019 (dr_step_indicator): New function.
4020 (dr_direction_indicator): Likewise.
4021 (dr_zero_step_indicator): Likewise.
4022 (dr_known_forward_stride_p): Likewise.
4023 * tree-loop-distribution.c (data_ref_segment_size): Return
4024 DR_STEP * (niters - 1).
4025 (compute_alias_check_pairs): Update call to the dr_with_seg_len
4027 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
4028 (vect_preserves_scalar_order_p): New function, split out from...
4029 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
4030 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
4031 (vect_vfa_access_size): New function.
4032 (vect_vfa_align): Likewise.
4033 (vect_compile_time_alias): Take access_size_a and access_b arguments.
4034 (dump_lower_bound): New function.
4035 (vect_check_lower_bound): Likewise.
4036 (vect_small_gap_p): Likewise.
4037 (vectorizable_with_step_bound_p): Likewise.
4038 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
4039 depencies if the vectorization factor is 1. Convert the checks
4040 for nonzero steps into checks on the bounds of DR_STEP. Try using
4041 a bunds check for variable steps if the minimum required step is
4042 relatively small. Update calls to the dr_with_seg_len
4043 constructor and to vect_compile_time_alias.
4044 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
4046 (vect_loop_versioning): Call it.
4047 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
4049 (vect_estimate_min_profitable_iters): Account for any bounds checks.
4051 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4052 Alan Hayward <alan.hayward@arm.com>
4053 David Sherwood <david.sherwood@arm.com>
4055 * doc/sourcebuild.texi (vect_scatter_store): Document.
4056 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
4058 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
4060 * genopinit.c (main): Add supports_vec_scatter_store and
4061 supports_vec_scatter_store_cached to target_optabs.
4062 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
4063 IFN_MASK_SCATTER_STORE.
4064 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
4066 * internal-fn.h (internal_store_fn_p): Declare.
4067 (internal_fn_stored_value_index): Likewise.
4068 * internal-fn.c (scatter_store_direct): New macro.
4069 (expand_scatter_store_optab_fn): New function.
4070 (direct_scatter_store_optab_supported_p): New macro.
4071 (internal_store_fn_p): New function.
4072 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
4073 IFN_MASK_SCATTER_STORE.
4074 (internal_fn_mask_index): Likewise.
4075 (internal_fn_stored_value_index): New function.
4076 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
4078 * optabs-query.h (supports_vec_scatter_store_p): Declare.
4079 * optabs-query.c (supports_vec_scatter_store_p): New function.
4080 * tree-vectorizer.h (vect_get_store_rhs): Declare.
4081 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
4082 true for scatter stores.
4083 (vect_gather_scatter_fn_p): Handle scatter stores too.
4084 (vect_check_gather_scatter): Consider using scatter stores if
4085 supports_vec_scatter_store_p.
4086 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
4088 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4089 internal_fn_stored_value_index.
4090 (check_load_store_masking): Handle scatter stores too.
4091 (vect_get_store_rhs): Make public.
4092 (vectorizable_call): Use internal_store_fn_p.
4093 (vectorizable_store): Handle scatter store internal functions.
4094 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
4095 when deciding whether the end of the group has been reached.
4096 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
4097 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
4098 (mask_scatter_store<mode>): New insns.
4100 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4101 Alan Hayward <alan.hayward@arm.com>
4102 David Sherwood <david.sherwood@arm.com>
4104 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
4105 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
4106 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
4108 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
4109 Use vect_truncate_gather_scatter_offset if we can't treat the
4110 operation as a normal gather load or scatter store.
4111 (get_group_load_store_type): Take the gather_scatter_info
4112 as argument. Try using a gather load or scatter store for
4113 single-element groups.
4114 (get_load_store_type): Update calls to get_group_load_store_type
4115 and vect_use_strided_gather_scatters_p.
4117 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4118 Alan Hayward <alan.hayward@arm.com>
4119 David Sherwood <david.sherwood@arm.com>
4121 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
4122 optional tree argument.
4123 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
4125 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
4126 but continue to use the current value as a fallback.
4127 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
4128 to compare the updates.
4129 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
4130 (get_load_store_type): Use it when handling a strided access.
4131 (vect_get_strided_load_store_ops): New function.
4132 (vect_get_data_ptr_increment): Likewise.
4133 (vectorizable_load): Handle strided gather loads. Always pass
4134 a step to vect_create_data_ref_ptr and bump_vector_ptr.
4136 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4137 Alan Hayward <alan.hayward@arm.com>
4138 David Sherwood <david.sherwood@arm.com>
4140 * doc/md.texi (gather_load@var{m}): Document.
4141 (mask_gather_load@var{m}): Likewise.
4142 * genopinit.c (main): Add supports_vec_gather_load and
4143 supports_vec_gather_load_cached to target_optabs.
4144 * optabs-tree.c (init_tree_optimization_optabs): Use
4145 ggc_cleared_alloc to allocate target_optabs.
4146 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
4147 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
4149 * internal-fn.h (internal_load_fn_p): Declare.
4150 (internal_gather_scatter_fn_p): Likewise.
4151 (internal_fn_mask_index): Likewise.
4152 (internal_gather_scatter_fn_supported_p): Likewise.
4153 * internal-fn.c (gather_load_direct): New macro.
4154 (expand_gather_load_optab_fn): New function.
4155 (direct_gather_load_optab_supported_p): New macro.
4156 (direct_internal_fn_optab): New function.
4157 (internal_load_fn_p): Likewise.
4158 (internal_gather_scatter_fn_p): Likewise.
4159 (internal_fn_mask_index): Likewise.
4160 (internal_gather_scatter_fn_supported_p): Likewise.
4161 * optabs-query.c (supports_at_least_one_mode_p): New function.
4162 (supports_vec_gather_load_p): Likewise.
4163 * optabs-query.h (supports_vec_gather_load_p): Declare.
4164 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
4165 and memory_type field.
4166 (NUM_PATTERNS): Bump to 15.
4167 * tree-vect-data-refs.c: Include internal-fn.h.
4168 (vect_gather_scatter_fn_p): New function.
4169 (vect_describe_gather_scatter_call): Likewise.
4170 (vect_check_gather_scatter): Try using internal functions for
4171 gather loads. Recognize existing calls to a gather load function.
4172 (vect_analyze_data_refs): Consider using gather loads if
4173 supports_vec_gather_load_p.
4174 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
4175 (vect_get_gather_scatter_offset_type): Likewise.
4176 (vect_convert_mask_for_vectype): Likewise.
4177 (vect_add_conversion_to_patterm): Likewise.
4178 (vect_try_gather_scatter_pattern): Likewise.
4179 (vect_recog_gather_scatter_pattern): New pattern recognizer.
4180 (vect_vect_recog_func_ptrs): Add it.
4181 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
4182 internal_fn_mask_index and internal_gather_scatter_fn_p.
4183 (check_load_store_masking): Take the gather_scatter_info as an
4184 argument and handle gather loads.
4185 (vect_get_gather_scatter_ops): New function.
4186 (vectorizable_call): Check internal_load_fn_p.
4187 (vectorizable_load): Likewise. Handle gather load internal
4189 (vectorizable_store): Update call to check_load_store_masking.
4190 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
4191 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
4192 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
4193 (aarch64_gather_scale_operand_d): New predicates.
4194 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
4195 (mask_gather_load<mode>): New insns.
4197 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4198 Alan Hayward <alan.hayward@arm.com>
4199 David Sherwood <david.sherwood@arm.com>
4201 * optabs.def (fold_left_plus_optab): New optab.
4202 * doc/md.texi (fold_left_plus_@var{m}): Document.
4203 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
4204 * internal-fn.c (fold_left_direct): Define.
4205 (expand_fold_left_optab_fn): Likewise.
4206 (direct_fold_left_optab_supported_p): Likewise.
4207 * fold-const-call.c (fold_const_fold_left): New function.
4208 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
4209 * tree-parloops.c (valid_reduction_p): New function.
4210 (gather_scalar_reductions): Use it.
4211 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
4212 (vect_finish_replace_stmt): Declare.
4213 * tree-vect-loop.c (fold_left_reduction_fn): New function.
4214 (needs_fold_left_reduction_p): New function, split out from...
4215 (vect_is_simple_reduction): ...here. Accept reductions that
4216 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
4217 (vect_force_simple_reduction): Also store the reduction type in
4218 the assignment's STMT_VINFO_REDUC_TYPE.
4219 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
4220 (merge_with_identity): New function.
4221 (vect_expand_fold_left): Likewise.
4222 (vectorize_fold_left_reduction): Likewise.
4223 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
4224 scalar phi in place for it. Check for target support and reject
4225 cases that would reassociate the operation. Defer the transform
4226 phase to vectorize_fold_left_reduction.
4227 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
4228 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
4229 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
4231 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4233 * tree-if-conv.c (predicate_mem_writes): Remove redundant
4234 call to ifc_temp_var.
4236 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4237 Alan Hayward <alan.hayward@arm.com>
4238 David Sherwood <david.sherwood@arm.com>
4240 * target.def (legitimize_address_displacement): Take the original
4241 offset as a poly_int.
4242 * targhooks.h (default_legitimize_address_displacement): Update
4244 * targhooks.c (default_legitimize_address_displacement): Likewise.
4245 * doc/tm.texi: Regenerate.
4246 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
4247 as an argument, moving assert of ad->disp == ad->disp_term to...
4248 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
4249 Try calling targetm.legitimize_address_displacement before expanding
4250 the address rather than afterwards, and adjust for the new interface.
4251 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
4252 Match the new hook interface. Handle SVE addresses.
4253 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
4256 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4258 * Makefile.in (OBJS): Add early-remat.o.
4259 * target.def (select_early_remat_modes): New hook.
4260 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
4261 * doc/tm.texi: Regenerate.
4262 * targhooks.h (default_select_early_remat_modes): Declare.
4263 * targhooks.c (default_select_early_remat_modes): New function.
4264 * timevar.def (TV_EARLY_REMAT): New timevar.
4265 * passes.def (pass_early_remat): New pass.
4266 * tree-pass.h (make_pass_early_remat): Declare.
4267 * early-remat.c: New file.
4268 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
4270 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
4272 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4273 Alan Hayward <alan.hayward@arm.com>
4274 David Sherwood <david.sherwood@arm.com>
4276 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
4277 vfm1 with a bound_epilog parameter.
4278 (vect_do_peeling): Update calls accordingly, and move the prologue
4279 call earlier in the function. Treat the base bound_epilog as 0 for
4280 fully-masked loops and retain vf - 1 for other loops. Add 1 to
4281 this base when peeling for gaps.
4282 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
4283 with fully-masked loops.
4284 (vect_estimate_min_profitable_iters): Handle the single peeled
4285 iteration in that case.
4287 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4288 Alan Hayward <alan.hayward@arm.com>
4289 David Sherwood <david.sherwood@arm.com>
4291 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
4292 single-element interleaving even if the size is not a power of 2.
4293 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
4294 accesses for single-element interleaving if the group size is
4297 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4298 Alan Hayward <alan.hayward@arm.com>
4299 David Sherwood <david.sherwood@arm.com>
4301 * doc/md.texi (fold_extract_last_@var{m}): Document.
4302 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
4303 * optabs.def (fold_extract_last_optab): New optab.
4304 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
4305 * internal-fn.c (fold_extract_direct): New macro.
4306 (expand_fold_extract_optab_fn): Likewise.
4307 (direct_fold_extract_optab_supported_p): Likewise.
4308 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
4309 * tree-vect-loop.c (vect_model_reduction_cost): Handle
4310 EXTRACT_LAST_REDUCTION.
4311 (get_initial_def_for_reduction): Do not create an initial vector
4312 for EXTRACT_LAST_REDUCTION reductions.
4313 (vectorizable_reduction): Leave the scalar phi in place for
4314 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
4315 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
4316 epilogue code for EXTRACT_LAST_REDUCTION and defer the
4317 transform phase to vectorizable_condition.
4318 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
4320 (vect_finish_stmt_generation): ...here.
4321 (vect_finish_replace_stmt): New function.
4322 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
4323 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
4325 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
4327 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4328 Alan Hayward <alan.hayward@arm.com>
4329 David Sherwood <david.sherwood@arm.com>
4331 * doc/md.texi (extract_last_@var{m}): Document.
4332 * optabs.def (extract_last_optab): New optab.
4333 * internal-fn.def (EXTRACT_LAST): New internal function.
4334 * internal-fn.c (cond_unary_direct): New macro.
4335 (expand_cond_unary_optab_fn): Likewise.
4336 (direct_cond_unary_optab_supported_p): Likewise.
4337 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
4338 loops using EXTRACT_LAST.
4339 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
4340 (extract_last_<mode>): ...this optab.
4341 (vec_extract<mode><Vel>): Update accordingly.
4343 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4344 Alan Hayward <alan.hayward@arm.com>
4345 David Sherwood <david.sherwood@arm.com>
4347 * target.def (empty_mask_is_expensive): New hook.
4348 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
4349 * doc/tm.texi: Regenerate.
4350 * targhooks.h (default_empty_mask_is_expensive): Declare.
4351 * targhooks.c (default_empty_mask_is_expensive): New function.
4352 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
4353 if the target says that empty masks are expensive.
4354 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
4356 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
4358 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4359 Alan Hayward <alan.hayward@arm.com>
4360 David Sherwood <david.sherwood@arm.com>
4362 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
4363 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
4364 (vect_use_loop_mask_for_alignment_p): New function.
4365 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
4366 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
4367 niters_skip argument. Make sure that the first niters_skip elements
4368 of the first iteration are inactive.
4369 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
4370 Update call to vect_set_loop_masks_directly.
4371 (get_misalign_in_elems): New function, split out from...
4372 (vect_gen_prolog_loop_niters): ...here.
4373 (vect_update_init_of_dr): Take a code argument that specifies whether
4374 the adjustment should be added or subtracted.
4375 (vect_update_init_of_drs): Likewise.
4376 (vect_prepare_for_masked_peels): New function.
4377 (vect_do_peeling): Skip prologue peeling if we're using a mask
4378 instead. Update call to vect_update_inits_of_drs.
4379 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4381 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
4382 alignment. Do not include the number of peeled iterations in
4383 the minimum threshold in that case.
4384 (vectorizable_induction): Adjust the start value down by
4385 LOOP_VINFO_MASK_SKIP_NITERS iterations.
4386 (vect_transform_loop): Call vect_prepare_for_masked_peels.
4387 Take the number of skipped iterations into account when calculating
4389 * tree-vect-stmts.c (vect_gen_while_not): New function.
4391 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4392 Alan Hayward <alan.hayward@arm.com>
4393 David Sherwood <david.sherwood@arm.com>
4395 * doc/sourcebuild.texi (vect_fully_masked): Document.
4396 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
4398 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
4400 (vect_analyze_loop_2): ...here. Don't check the vectorization
4401 factor against the number of loop iterations if the loop is
4404 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4405 Alan Hayward <alan.hayward@arm.com>
4406 David Sherwood <david.sherwood@arm.com>
4408 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
4409 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
4410 (dump_groups): Update accordingly.
4411 (iv_use::mem_type): New member variable.
4412 (address_p): New function.
4413 (record_use): Add a mem_type argument and initialize the new
4415 (record_group_use): Add a mem_type argument. Use address_p.
4416 Remove obsolete null checks of base_object. Update call to record_use.
4417 (find_interesting_uses_op): Update call to record_group_use.
4418 (find_interesting_uses_cond): Likewise.
4419 (find_interesting_uses_address): Likewise.
4420 (get_mem_type_for_internal_fn): New function.
4421 (find_address_like_use): Likewise.
4422 (find_interesting_uses_stmt): Try find_address_like_use before
4423 calling find_interesting_uses_op.
4424 (addr_offset_valid_p): Use the iv mem_type field as the type
4425 of the addressed memory.
4426 (add_autoinc_candidates): Likewise.
4427 (get_address_cost): Likewise.
4428 (split_small_address_groups_p): Use address_p.
4429 (split_address_groups): Likewise.
4430 (add_iv_candidate_for_use): Likewise.
4431 (autoinc_possible_for_pair): Likewise.
4432 (rewrite_groups): Likewise.
4433 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
4434 (determine_group_iv_cost): Update after split of USE_ADDRESS.
4435 (get_alias_ptr_type_for_ptr_address): New function.
4436 (rewrite_use_address): Rewrite address uses in calls that were
4437 identified by find_address_like_use.
4439 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4440 Alan Hayward <alan.hayward@arm.com>
4441 David Sherwood <david.sherwood@arm.com>
4443 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
4445 * gimple-expr.h (is_gimple_addressable: Likewise.
4446 * gimple-expr.c (is_gimple_address): Likewise.
4447 * internal-fn.c (expand_call_mem_ref): New function.
4448 (expand_mask_load_optab_fn): Use it.
4449 (expand_mask_store_optab_fn): Likewise.
4451 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4452 Alan Hayward <alan.hayward@arm.com>
4453 David Sherwood <david.sherwood@arm.com>
4455 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
4456 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
4457 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
4458 (cond_umax@var{mode}): Document.
4459 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
4460 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
4461 (cond_umin_optab, cond_umax_optab): New optabs.
4462 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
4463 (COND_IOR, COND_XOR): New internal functions.
4464 * internal-fn.h (get_conditional_internal_fn): Declare.
4465 * internal-fn.c (cond_binary_direct): New macro.
4466 (expand_cond_binary_optab_fn): Likewise.
4467 (direct_cond_binary_optab_supported_p): Likewise.
4468 (get_conditional_internal_fn): New function.
4469 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
4470 Cope with reduction statements that are vectorized as calls rather
4472 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
4473 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
4474 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
4475 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
4476 (UNSPEC_COND_EOR): New unspecs.
4477 (optab): Add mappings for them.
4478 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
4479 (sve_int_op, sve_fp_op): New int attributes.
4481 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4482 Alan Hayward <alan.hayward@arm.com>
4483 David Sherwood <david.sherwood@arm.com>
4485 * optabs.def (while_ult_optab): New optab.
4486 * doc/md.texi (while_ult@var{m}@var{n}): Document.
4487 * internal-fn.def (WHILE_ULT): New internal function.
4488 * internal-fn.h (direct_internal_fn_supported_p): New override
4489 that takes two types as argument.
4490 * internal-fn.c (while_direct): New macro.
4491 (expand_while_optab_fn): New function.
4492 (convert_optab_supported_p): Likewise.
4493 (direct_while_optab_supported_p): New macro.
4494 * wide-int.h (wi::udiv_ceil): New function.
4495 * tree-vectorizer.h (rgroup_masks): New structure.
4496 (vec_loop_masks): New typedef.
4497 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
4499 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
4500 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
4501 (vect_max_vf): New function.
4502 (slpeel_make_loop_iterate_ntimes): Delete.
4503 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
4504 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
4505 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
4506 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
4507 internal-fn.h, stor-layout.h and optabs-query.h.
4508 (vect_set_loop_mask): New function.
4509 (add_preheader_seq): Likewise.
4510 (add_header_seq): Likewise.
4511 (interleave_supported_p): Likewise.
4512 (vect_maybe_permute_loop_masks): Likewise.
4513 (vect_set_loop_masks_directly): Likewise.
4514 (vect_set_loop_condition_masked): Likewise.
4515 (vect_set_loop_condition_unmasked): New function, split out from
4516 slpeel_make_loop_iterate_ntimes.
4517 (slpeel_make_loop_iterate_ntimes): Rename to..
4518 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
4519 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
4520 (vect_do_peeling): Update call accordingly.
4521 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
4523 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
4524 mask_compare_type, can_fully_mask_p and fully_masked_p.
4525 (release_vec_loop_masks): New function.
4526 (_loop_vec_info): Use it to free the loop masks.
4527 (can_produce_all_loop_masks_p): New function.
4528 (vect_get_max_nscalars_per_iter): Likewise.
4529 (vect_verify_full_masking): Likewise.
4530 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
4531 retries, and free the mask rgroups before retrying. Check loop-wide
4532 reasons for disallowing fully-masked loops. Make the final decision
4533 about whether use a fully-masked loop or not.
4534 (vect_estimate_min_profitable_iters): Do not assume that peeling
4535 for the number of iterations will be needed for fully-masked loops.
4536 (vectorizable_reduction): Disable fully-masked loops.
4537 (vectorizable_live_operation): Likewise.
4538 (vect_halve_mask_nunits): New function.
4539 (vect_double_mask_nunits): Likewise.
4540 (vect_record_loop_mask): Likewise.
4541 (vect_get_loop_mask): Likewise.
4542 (vect_transform_loop): Handle the case in which the final loop
4543 iteration might handle a partial vector. Call vect_set_loop_condition
4544 instead of slpeel_make_loop_iterate_ntimes.
4545 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
4546 (check_load_store_masking): New function.
4547 (prepare_load_store_mask): Likewise.
4548 (vectorizable_store): Handle fully-masked loops.
4549 (vectorizable_load): Likewise.
4550 (supportable_widening_operation): Use vect_halve_mask_nunits for
4552 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
4553 (vect_gen_while): New function.
4554 * config/aarch64/aarch64.md (umax<mode>3): New expander.
4555 (aarch64_uqdec<mode>): New insn.
4557 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4558 Alan Hayward <alan.hayward@arm.com>
4559 David Sherwood <david.sherwood@arm.com>
4561 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
4562 (reduc_xor_scal_optab): New optabs.
4563 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
4564 (reduc_xor_scal_@var{m}): Document.
4565 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
4566 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
4568 * fold-const-call.c (fold_const_call): Handle them.
4569 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
4570 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
4571 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
4572 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
4573 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
4574 (UNSPEC_XORV): New unspecs.
4575 (optab): Add entries for them.
4576 (BITWISEV): New int iterator.
4577 (bit_reduc_op): New int attributes.
4579 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4580 Alan Hayward <alan.hayward@arm.com>
4581 David Sherwood <david.sherwood@arm.com>
4583 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
4584 * internal-fn.def (VEC_SHL_INSERT): New internal function.
4585 * optabs.def (vec_shl_insert_optab): New optab.
4586 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
4587 (duplicate_and_interleave): Likewise.
4588 * tree-vect-loop.c: Include internal-fn.h.
4589 (neutral_op_for_slp_reduction): New function, split out from
4590 get_initial_defs_for_reduction.
4591 (get_initial_def_for_reduction): Handle option 2 for variable-length
4592 vectors by loading the neutral value into a vector and then shifting
4593 the initial value into element 0.
4594 (get_initial_defs_for_reduction): Replace the code argument with
4595 the neutral value calculated by neutral_op_for_slp_reduction.
4596 Use gimple_build_vector for constant-length vectors.
4597 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
4598 but the first group_size elements have a neutral value.
4599 Use duplicate_and_interleave otherwise.
4600 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
4601 Update call to get_initial_defs_for_reduction. Handle SLP
4602 reductions for variable-length vectors by creating one vector
4603 result for each scalar result, with the elements associated
4604 with other scalar results stubbed out with the neutral value.
4605 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
4606 Require IFN_VEC_SHL_INSERT for double reductions on
4607 variable-length vectors, or SLP reductions that have
4608 a neutral value. Require can_duplicate_and_interleave_p
4609 support for variable-length unchained SLP reductions if there
4610 is no neutral value, such as for MIN/MAX reductions. Also require
4611 the number of vector elements to be a multiple of the number of
4612 SLP statements when doing variable-length unchained SLP reductions.
4613 Update call to vect_create_epilog_for_reduction.
4614 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
4615 and remove initial values.
4616 (duplicate_and_interleave): Make public.
4617 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
4618 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
4620 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4621 Alan Hayward <alan.hayward@arm.com>
4622 David Sherwood <david.sherwood@arm.com>
4624 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
4625 (can_duplicate_and_interleave_p): New function.
4626 (vect_get_and_check_slp_defs): Take the vector of statements
4627 rather than just the current one. Remove excess parentheses.
4628 Restriction rejectinon of vect_constant_def and vect_external_def
4629 for variable-length vectors to boolean types, or types for which
4630 can_duplicate_and_interleave_p is false.
4631 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
4632 (duplicate_and_interleave): New function.
4633 (vect_get_constant_vectors): Use gimple_build_vector for
4634 constant-length vectors and suitable variable-length constant
4635 vectors. Use duplicate_and_interleave for other variable-length
4636 vectors. Don't defer the update when inserting new statements.
4638 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4639 Alan Hayward <alan.hayward@arm.com>
4640 David Sherwood <david.sherwood@arm.com>
4642 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
4643 min_profitable_iters doesn't go negative.
4645 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4646 Alan Hayward <alan.hayward@arm.com>
4647 David Sherwood <david.sherwood@arm.com>
4649 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
4650 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
4651 * optabs.def (vec_mask_load_lanes_optab): New optab.
4652 (vec_mask_store_lanes_optab): Likewise.
4653 * internal-fn.def (MASK_LOAD_LANES): New internal function.
4654 (MASK_STORE_LANES): Likewise.
4655 * internal-fn.c (mask_load_lanes_direct): New macro.
4656 (mask_store_lanes_direct): Likewise.
4657 (expand_mask_load_optab_fn): Handle masked operations.
4658 (expand_mask_load_lanes_optab_fn): New macro.
4659 (expand_mask_store_optab_fn): Handle masked operations.
4660 (expand_mask_store_lanes_optab_fn): New macro.
4661 (direct_mask_load_lanes_optab_supported_p): Likewise.
4662 (direct_mask_store_lanes_optab_supported_p): Likewise.
4663 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
4665 (vect_load_lanes_supported): Likewise.
4666 * tree-vect-data-refs.c (strip_conversion): New function.
4667 (can_group_stmts_p): Likewise.
4668 (vect_analyze_data_ref_accesses): Use it instead of checking
4669 for a pair of assignments.
4670 (vect_store_lanes_supported): Take a masked_p parameter.
4671 (vect_load_lanes_supported): Likewise.
4672 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
4673 vect_store_lanes_supported and vect_load_lanes_supported.
4674 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
4675 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
4676 parameter. Don't allow gaps for masked accesses.
4677 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
4678 and vect_load_lanes_supported.
4679 (get_load_store_type): Take a masked_p parameter and update
4680 call to get_group_load_store_type.
4681 (vectorizable_store): Update call to get_load_store_type.
4682 Handle IFN_MASK_STORE_LANES.
4683 (vectorizable_load): Update call to get_load_store_type.
4684 Handle IFN_MASK_LOAD_LANES.
4686 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4687 Alan Hayward <alan.hayward@arm.com>
4688 David Sherwood <david.sherwood@arm.com>
4690 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
4692 * config/aarch64/aarch64-protos.h
4693 (aarch64_sve_struct_memory_operand_p): Declare.
4694 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
4695 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
4696 (VPRED, vpred): Handle SVE structure modes.
4697 * config/aarch64/constraints.md (Utx): New constraint.
4698 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
4699 (aarch64_sve_struct_nonimmediate_operand): New predicates.
4700 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
4701 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
4702 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
4703 structure modes. Split into pieces after RA.
4704 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
4705 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
4707 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
4708 SVE structure modes.
4709 (aarch64_classify_address): Likewise.
4710 (sizetochar): Move earlier in file.
4711 (aarch64_print_operand): Handle SVE register lists.
4712 (aarch64_array_mode): New function.
4713 (aarch64_sve_struct_memory_operand_p): Likewise.
4714 (TARGET_ARRAY_MODE): Redefine.
4716 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4717 Alan Hayward <alan.hayward@arm.com>
4718 David Sherwood <david.sherwood@arm.com>
4720 * target.def (array_mode): New target hook.
4721 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
4722 * doc/tm.texi: Regenerate.
4723 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
4724 * hooks.c (hook_optmode_mode_uhwi_none): New function.
4725 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
4727 * stor-layout.c (mode_for_array): Likewise. Support polynomial
4730 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4731 Alan Hayward <alan.hayward@arm.com>
4732 David Sherwood <david.sherwood@arm.com>
4734 * fold-const.c (fold_binary_loc): Check the argument types
4735 rather than the result type when testing for a vector operation.
4737 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4739 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
4740 * doc/tm.texi: Regenerate.
4742 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
4743 Alan Hayward <alan.hayward@arm.com>
4744 David Sherwood <david.sherwood@arm.com>
4746 * doc/invoke.texi (-msve-vector-bits=): Document new option.
4747 (sve): Document new AArch64 extension.
4748 * doc/md.texi (w): Extend the description of the AArch64
4749 constraint to include SVE vectors.
4750 (Upl, Upa): Document new AArch64 predicate constraints.
4751 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
4753 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
4754 (msve-vector-bits=): New option.
4755 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
4756 SVE when these are disabled.
4757 (sve): New extension.
4758 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
4759 modes. Adjust their number of units based on aarch64_sve_vg.
4760 (MAX_BITSIZE_MODE_ANY_MODE): Define.
4761 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
4762 aarch64_addr_query_type.
4763 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
4764 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
4765 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
4766 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
4767 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
4768 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
4769 (aarch64_simd_imm_zero_p): Delete.
4770 (aarch64_check_zero_based_sve_index_immediate): Declare.
4771 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4772 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4773 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4774 (aarch64_sve_float_mul_immediate_p): Likewise.
4775 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4777 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
4778 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
4779 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
4780 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
4781 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
4782 (aarch64_regmode_natural_size): Likewise.
4783 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
4784 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
4786 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
4787 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
4788 for VG and the SVE predicate registers.
4789 (V_ALIASES): Add a "z"-prefixed alias.
4790 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
4791 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
4792 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
4793 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
4794 (REG_CLASS_NAMES): Add entries for them.
4795 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
4796 and the predicate registers.
4797 (aarch64_sve_vg): Declare.
4798 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
4799 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
4800 (REGMODE_NATURAL_SIZE): Define.
4801 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
4803 * config/aarch64/aarch64.c: Include cfgrtl.h.
4804 (simd_immediate_info): Add a constructor for series vectors,
4805 and an associated step field.
4806 (aarch64_sve_vg): New variable.
4807 (aarch64_dbx_register_number): Handle VG and the predicate registers.
4808 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
4809 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
4810 (VEC_ANY_DATA, VEC_STRUCT): New constants.
4811 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
4812 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
4813 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
4814 (aarch64_get_mask_mode): New functions.
4815 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
4816 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
4817 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
4818 predicate modes and predicate registers. Explicitly restrict
4819 GPRs to modes of 16 bytes or smaller. Only allow FP registers
4820 to store a vector mode if it is recognized by
4821 aarch64_classify_vector_mode.
4822 (aarch64_regmode_natural_size): New function.
4823 (aarch64_hard_regno_caller_save_mode): Return the original mode
4825 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
4826 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
4827 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
4828 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
4830 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
4831 does not overlap dest if the function is frame-related. Handle
4833 (aarch64_split_add_offset): New function.
4834 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
4835 them aarch64_add_offset.
4836 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
4837 and update call to aarch64_sub_sp.
4838 (aarch64_add_cfa_expression): New function.
4839 (aarch64_expand_prologue): Pass extra temporary registers to the
4840 functions above. Handle the case in which we need to emit new
4841 DW_CFA_expressions for registers that were originally saved
4842 relative to the stack pointer, but now have to be expressed
4843 relative to the frame pointer.
4844 (aarch64_output_mi_thunk): Pass extra temporary registers to the
4846 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
4847 IP0 and IP1 values for SVE frames.
4848 (aarch64_expand_vec_series): New function.
4849 (aarch64_expand_sve_widened_duplicate): Likewise.
4850 (aarch64_expand_sve_const_vector): Likewise.
4851 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
4852 Handle SVE constants. Use emit_move_insn to move a force_const_mem
4853 into the register, rather than emitting a SET directly.
4854 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
4855 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
4856 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
4857 (offset_9bit_signed_scaled_p): New functions.
4858 (aarch64_replicate_bitmask_imm): New function.
4859 (aarch64_bitmask_imm): Use it.
4860 (aarch64_cannot_force_const_mem): Reject expressions involving
4861 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
4862 (aarch64_classify_index): Handle SVE indices, by requiring
4863 a plain register index with a scale that matches the element size.
4864 (aarch64_classify_address): Handle SVE addresses. Assert that
4865 the mode of the address is VOIDmode or an integer mode.
4866 Update call to aarch64_classify_symbol.
4867 (aarch64_classify_symbolic_expression): Update call to
4868 aarch64_classify_symbol.
4869 (aarch64_const_vec_all_in_range_p): New function.
4870 (aarch64_print_vector_float_operand): Likewise.
4871 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
4872 "vN" for FP registers with SVE modes. Handle (const ...) vectors
4873 and the FP immediates 1.0 and 0.5.
4874 (aarch64_print_address_internal): Handle SVE addresses.
4875 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
4876 (aarch64_regno_regclass): Handle predicate registers.
4877 (aarch64_secondary_reload): Handle big-endian reloads of SVE
4879 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
4880 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
4881 (aarch64_convert_sve_vector_bits): New function.
4882 (aarch64_override_options): Use it to handle -msve-vector-bits=.
4883 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
4885 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
4886 Handle SVE vector and predicate modes. Accept VL-based constants
4887 that need only one temporary register, and VL offsets that require
4888 no temporary registers.
4889 (aarch64_conditional_register_usage): Mark the predicate registers
4890 as fixed if SVE isn't available.
4891 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
4892 Return true for SVE vector and predicate modes.
4893 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
4894 rather than an unsigned int. Handle SVE modes.
4895 (aarch64_preferred_simd_mode): Update call accordingly. Handle
4897 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
4899 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
4900 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
4901 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
4902 (aarch64_sve_float_mul_immediate_p): New functions.
4903 (aarch64_sve_valid_immediate): New function.
4904 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
4905 Explicitly reject structure modes. Check for INDEX constants.
4906 Handle PTRUE and PFALSE constants.
4907 (aarch64_check_zero_based_sve_index_immediate): New function.
4908 (aarch64_simd_imm_zero_p): Delete.
4909 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
4910 vector modes. Accept constants in the range of CNT[BHWD].
4911 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
4912 ask for an Advanced SIMD mode.
4913 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
4914 (aarch64_simd_vector_alignment): Handle SVE predicates.
4915 (aarch64_vectorize_preferred_vector_alignment): New function.
4916 (aarch64_simd_vector_alignment_reachable): Use it instead of
4918 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
4919 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
4921 (MAX_VECT_LEN): Delete.
4922 (expand_vec_perm_d): Add a vec_flags field.
4923 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
4924 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
4925 (aarch64_evpc_ext): Don't apply a big-endian lane correction
4927 (aarch64_evpc_rev): Rename to...
4928 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
4929 (aarch64_evpc_rev_global): New function.
4930 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
4931 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
4933 (aarch64_evpc_sve_tbl): New function.
4934 (aarch64_expand_vec_perm_const_1): Update after rename of
4935 aarch64_evpc_rev. Handle SVE permutes too, trying
4936 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
4937 than aarch64_evpc_tbl.
4938 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
4939 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
4940 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
4941 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
4942 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
4943 (aarch64_expand_sve_vcond): New functions.
4944 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
4945 of aarch64_vector_mode_p.
4946 (aarch64_dwarf_poly_indeterminate_value): New function.
4947 (aarch64_compute_pressure_classes): Likewise.
4948 (aarch64_can_change_mode_class): Likewise.
4949 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
4950 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
4951 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
4952 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
4953 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
4954 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
4955 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
4956 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
4958 (Dn, Dl, Dr): Accept const as well as const_vector.
4959 (Dz): Likewise. Compare against CONST0_RTX.
4960 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
4961 of "vector" where appropriate.
4962 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
4963 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
4964 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
4965 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
4966 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
4967 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
4968 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
4969 (v_int_equiv): Extend to SVE modes.
4970 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
4972 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
4973 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
4974 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
4975 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
4976 (SVE_COND_FP_CMP): New int iterators.
4977 (perm_hilo): Handle the new unpack unspecs.
4978 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
4980 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
4981 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
4982 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
4983 (aarch64_equality_operator, aarch64_constant_vector_operand)
4984 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
4985 (aarch64_sve_nonimmediate_operand): Likewise.
4986 (aarch64_sve_general_operand): Likewise.
4987 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
4988 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
4989 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
4990 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
4991 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
4992 (aarch64_sve_float_arith_immediate): Likewise.
4993 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
4994 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
4995 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
4996 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
4997 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
4998 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
4999 (aarch64_sve_float_arith_operand): Likewise.
5000 (aarch64_sve_float_arith_with_sub_operand): Likewise.
5001 (aarch64_sve_float_mul_operand): Likewise.
5002 (aarch64_sve_vec_perm_operand): Likewise.
5003 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
5004 (aarch64_mov_operand): Accept const_poly_int and const_vector.
5005 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
5006 as well as const_vector.
5007 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
5008 in file. Use CONST0_RTX and CONSTM1_RTX.
5009 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
5010 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
5011 Use aarch64_simd_imm_zero.
5012 * config/aarch64/aarch64-sve.md: New file.
5013 * config/aarch64/aarch64.md: Include it.
5014 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
5015 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
5016 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
5017 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
5018 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
5019 (sve): New attribute.
5020 (enabled): Disable instructions with the sve attribute unless
5022 (movqi, movhi): Pass CONST_POLY_INT operaneds through
5023 aarch64_expand_mov_immediate.
5024 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
5025 CNT[BHSD] immediates.
5026 (movti): Split CONST_POLY_INT moves into two halves.
5027 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
5028 Split additions that need a temporary here if the destination
5029 is the stack pointer.
5030 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
5031 (*add<mode>3_poly_1): New instruction.
5032 (set_clobber_cc): New expander.
5034 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5036 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
5037 parameter and use it instead of GET_MODE_SIZE (innermode). Use
5038 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
5039 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
5040 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
5041 Change innermode from fixed_mode_size to machine_mode.
5042 (simplify_subreg): Update call accordingly. Handle a constant-sized
5043 subreg of a variable-length CONST_VECTOR.
5045 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
5046 Alan Hayward <alan.hayward@arm.com>
5047 David Sherwood <david.sherwood@arm.com>
5049 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
5050 (add_offset_to_base): New function, split out from...
5051 (create_mem_ref): ...here. When handling a scale other than 1,
5052 check first whether the address is valid without the offset.
5053 Add it into the base if so, leaving the index and scale as-is.
5055 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5058 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
5059 fold_for_warn before checking if arg2 is INTEGER_CST.
5061 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
5063 * config/rs6000/predicates.md (load_multiple_operation): Delete.
5064 (store_multiple_operation): Delete.
5065 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
5066 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
5067 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
5068 guarded by TARGET_STRING.
5069 (rs6000_output_load_multiple): Delete.
5070 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
5071 OPTION_MASK_STRING / TARGET_STRING handling.
5072 (print_operand) <'N', 'O'>: Add comment that these are unused now.
5073 (const rs6000_opt_masks) <"string">: Change mask to 0.
5074 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
5075 (MASK_STRING): Delete.
5076 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
5078 (load_multiple): Delete.
5085 (store_multiple): Delete.
5092 (movmemsi_8reg): Delete.
5093 (corresponding unnamed define_insn): Delete.
5094 (movmemsi_6reg): Delete.
5095 (corresponding unnamed define_insn): Delete.
5096 (movmemsi_4reg): Delete.
5097 (corresponding unnamed define_insn): Delete.
5098 (movmemsi_2reg): Delete.
5099 (corresponding unnamed define_insn): Delete.
5100 (movmemsi_1reg): Delete.
5101 (corresponding unnamed define_insn): Delete.
5102 * config/rs6000/rs6000.opt (mno-string): New.
5103 (mstring): Replace by deprecation warning stub.
5104 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
5106 2018-01-12 Jakub Jelinek <jakub@redhat.com>
5108 * regrename.c (regrename_do_replace): If replacing the same
5109 reg multiple times, try to reuse last created gen_raw_REG.
5112 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
5113 main to workaround a bug in GDB.
5115 2018-01-12 Tom de Vries <tom@codesourcery.com>
5118 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
5120 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
5122 PR rtl-optimization/80481
5123 * ira-color.c (get_cap_member): New function.
5124 (allocnos_conflict_by_live_ranges_p): Use it.
5125 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
5126 (setup_slot_coalesced_allocno_live_ranges): Ditto.
5128 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
5131 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
5132 (*saddl_se_1): Ditto.
5134 (*ssubl_se_1): Ditto.
5136 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5138 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
5139 rather than wi::to_widest for DR_INITs.
5140 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
5141 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
5142 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
5144 (vect_analyze_group_access_1): Note that here.
5146 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5148 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
5149 polynomial type sizes.
5151 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
5153 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
5154 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
5155 (gimple_add_tmp_var): Likewise.
5157 2018-01-12 Martin Liska <mliska@suse.cz>
5159 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
5160 (gimple_alloc_sizes): Likewise.
5161 (dump_gimple_statistics): Use PRIu64 in printf format.
5162 * gimple.h: Change uint64_t to int.
5164 2018-01-12 Martin Liska <mliska@suse.cz>
5166 * tree-core.h: Use uint64_t instead of int.
5167 * tree.c (tree_node_counts): Likewise.
5168 (tree_node_sizes): Likewise.
5169 (dump_tree_statistics): Use PRIu64 in printf format.
5171 2018-01-12 Martin Liska <mliska@suse.cz>
5173 * Makefile.in: As qsort_chk is implemented in vec.c, add
5174 vec.o to linkage of gencfn-macros.
5175 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
5176 passing the info to record_node_allocation_statistics.
5177 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
5179 * ggc-common.c (struct ggc_usage): Add operator== and use
5180 it in operator< and compare function.
5181 * mem-stats.h (struct mem_usage): Likewise.
5182 * vec.c (struct vec_usage): Remove operator< and compare
5183 function. Can be simply inherited.
5185 2018-01-12 Martin Jambor <mjambor@suse.cz>
5188 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
5189 * tree-ssa-math-opts.c: Include domwalk.h.
5190 (convert_mult_to_fma_1): New function.
5191 (fma_transformation_info): New type.
5192 (fma_deferring_state): Likewise.
5193 (cancel_fma_deferring): New function.
5194 (result_of_phi): Likewise.
5195 (last_fma_candidate_feeds_initial_phi): Likewise.
5196 (convert_mult_to_fma): Added deferring logic, split actual
5197 transformation to convert_mult_to_fma_1.
5198 (math_opts_dom_walker): New type.
5199 (math_opts_dom_walker::after_dom_children): New method, body moved
5200 here from pass_optimize_widening_mul::execute, added deferring logic
5202 (pass_optimize_widening_mul::execute): Moved most of code to
5203 math_opts_dom_walker::after_dom_children.
5204 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
5205 * config/i386/i386.c (ix86_option_override_internal): Added
5206 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
5208 2018-01-12 Richard Biener <rguenther@suse.de>
5211 * dwarf2out.c (gen_variable_die): Do not reset old_die for
5212 inline instance vars.
5214 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
5217 * config/rx/rx.c (rx_is_restricted_memory_address):
5220 2018-01-12 Richard Biener <rguenther@suse.de>
5222 PR tree-optimization/80846
5223 * target.def (split_reduction): New target hook.
5224 * targhooks.c (default_split_reduction): New function.
5225 * targhooks.h (default_split_reduction): Declare.
5226 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
5227 target requests first reduce vectors by combining low and high
5229 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
5230 (get_vectype_for_scalar_type_and_size): Export.
5231 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
5232 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
5233 * doc/tm.texi: Regenerate.
5234 * config/i386/i386.c (ix86_split_reduction): Implement
5235 TARGET_VECTORIZE_SPLIT_REDUCTION.
5237 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5240 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
5241 in PIC mode except for TARGET_VXWORKS_RTP.
5242 * config/sparc/sparc.c: Include cfgrtl.h.
5243 (TARGET_INIT_PIC_REG): Define.
5244 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
5245 (sparc_pic_register_p): New predicate.
5246 (sparc_legitimate_address_p): Use it.
5247 (sparc_legitimize_pic_address): Likewise.
5248 (sparc_delegitimize_address): Likewise.
5249 (sparc_mode_dependent_address_p): Likewise.
5250 (gen_load_pcrel_sym): Remove 4th parameter.
5251 (load_got_register): Adjust call to above. Remove obsolete stuff.
5252 (sparc_expand_prologue): Do not call load_got_register here.
5253 (sparc_flat_expand_prologue): Likewise.
5254 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
5255 (sparc_use_pseudo_pic_reg): New function.
5256 (sparc_init_pic_reg): Likewise.
5257 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
5258 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
5260 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
5262 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
5263 Add item for branch_cost.
5265 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
5267 PR rtl-optimization/83565
5268 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
5269 not extend the result to a larger mode for rotate operations.
5270 (num_sign_bit_copies1): Likewise.
5272 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
5275 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
5277 Use values-Xc.o for -pedantic.
5278 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
5280 2018-01-12 Martin Liska <mliska@suse.cz>
5283 * ipa-devirt.c (final_warning_record::grow_type_warnings):
5285 (possible_polymorphic_call_targets): Use it.
5286 (ipa_devirt): Likewise.
5288 2018-01-12 Martin Liska <mliska@suse.cz>
5290 * profile-count.h (enum profile_quality): Use 0 as invalid
5291 enum value of profile_quality.
5293 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
5295 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
5296 -mext-string options.
5298 2018-01-12 Richard Biener <rguenther@suse.de>
5300 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
5301 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
5302 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
5304 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
5306 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
5308 * configure.ac (--with-long-double-format): Add support for the
5309 configuration option to change the default long double format on
5311 * config.gcc (powerpc*-linux*-*): Likewise.
5312 * configure: Regenerate.
5313 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
5314 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
5315 used without modification.
5317 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
5319 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
5320 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
5321 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
5322 MISC_BUILTIN_SPEC_BARRIER.
5323 (rs6000_init_builtins): Likewise.
5324 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
5326 (speculation_barrier): New define_insn.
5327 * doc/extend.texi: Document __builtin_speculation_barrier.
5329 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5332 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
5333 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
5334 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
5336 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
5337 integral modes instead of "ss" and "sd".
5338 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
5339 vectors with 32-bit and 64-bit elements.
5340 (vecdupssescalarmodesuffix): New mode attribute.
5341 (vec_dup<mode>): Use it.
5343 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
5346 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
5347 frame if argument is passed on stack.
5349 2018-01-11 Jakub Jelinek <jakub@redhat.com>
5352 * ree.c (combine_reaching_defs): Optimize also
5353 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
5354 reg2=any_extend(exp); reg1=reg2;, formatting fix.
5356 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5359 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
5361 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
5364 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
5365 after they are computed.
5367 2018-01-11 Bin Cheng <bin.cheng@arm.com>
5369 PR tree-optimization/83695
5370 * gimple-loop-linterchange.cc
5371 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
5372 reset cached scev information after interchange.
5373 (pass_linterchange::execute): Remove call to scev_reset_htab.
5375 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5377 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
5378 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
5379 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
5380 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
5381 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
5382 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
5383 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
5384 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
5385 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
5386 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
5387 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
5388 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
5389 (V_lane_reg): Likewise.
5390 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
5392 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
5393 (vfmal_lane_low<mode>_intrinsic,
5394 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
5395 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
5396 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
5397 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
5398 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
5399 vfmsl_lane_high<mode>_intrinsic): New define_insns.
5401 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5403 * config/arm/arm-cpus.in (fp16fml): New feature.
5404 (ALL_SIMD): Add fp16fml.
5405 (armv8.2-a): Add fp16fml as an option.
5406 (armv8.3-a): Likewise.
5407 (armv8.4-a): Add fp16fml as part of fp16.
5408 * config/arm/arm.h (TARGET_FP16FML): Define.
5409 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
5411 * config/arm/arm-modes.def (V2HF): Define.
5412 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
5413 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
5414 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
5415 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
5416 vfmsl_low, vfmsl_high): New set of builtins.
5417 * config/arm/iterators.md (PLUSMINUS): New code iterator.
5418 (vfml_op): New code attribute.
5419 (VFMLHALVES): New int iterator.
5420 (VFML, VFMLSEL): New mode attributes.
5421 (V_reg): Define mapping for V2HF.
5422 (V_hi, V_lo): New mode attributes.
5423 (VF_constraint): Likewise.
5424 (vfml_half, vfml_half_selector): New int attributes.
5425 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
5427 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
5428 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
5430 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
5431 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
5432 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
5433 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
5435 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
5436 Document new effective target and option set.
5438 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5440 * config/arm/arm-cpus.in (armv8_4): New feature.
5441 (ARMv8_4a): New fgroup.
5442 (armv8.4-a): New arch.
5443 * config/arm/arm-tables.opt: Regenerate.
5444 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
5445 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
5446 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
5447 Add matching rules for -march=armv8.4-a and extensions.
5448 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
5450 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
5453 * config/rx/rx.md (BW): New mode attribute.
5454 (sync_lock_test_and_setsi): Add mode suffix to insn output.
5456 2018-01-11 Richard Biener <rguenther@suse.de>
5458 PR tree-optimization/83435
5459 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
5460 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
5461 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
5463 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5464 Alan Hayward <alan.hayward@arm.com>
5465 David Sherwood <david.sherwood@arm.com>
5467 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
5469 (aarch64_classify_address): Initialize it. Track polynomial offsets.
5470 (aarch64_print_address_internal): Use it to check for a zero offset.
5472 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5473 Alan Hayward <alan.hayward@arm.com>
5474 David Sherwood <david.sherwood@arm.com>
5476 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
5477 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
5478 Return a poly_int64 rather than a HOST_WIDE_INT.
5479 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
5480 rather than a HOST_WIDE_INT.
5481 * config/aarch64/aarch64.h (aarch64_frame): Protect with
5482 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
5483 hard_fp_offset, frame_size, initial_adjust, callee_offset and
5484 final_offset from HOST_WIDE_INT to poly_int64.
5485 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
5486 to_constant when getting the number of units in an Advanced SIMD
5488 (aarch64_builtin_vectorized_function): Check for a constant number
5490 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
5492 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
5493 attribute instead of GET_MODE_NUNITS.
5494 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
5495 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
5496 GET_MODE_SIZE for fixed-size registers.
5497 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
5498 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
5499 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
5500 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
5501 (aarch64_print_operand, aarch64_print_address_internal)
5502 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
5503 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
5504 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
5505 Handle polynomial GET_MODE_SIZE.
5506 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
5507 wider than SImode without modification.
5508 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
5509 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
5510 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
5511 passing and returning SVE modes.
5512 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
5513 rather than GEN_INT.
5514 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
5515 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
5516 (aarch64_allocate_and_probe_stack_space): Likewise.
5517 (aarch64_layout_frame): Cope with polynomial offsets.
5518 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
5519 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
5521 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
5522 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
5523 poly_int64 rather than a HOST_WIDE_INT.
5524 (aarch64_get_separate_components, aarch64_process_components)
5525 (aarch64_expand_prologue, aarch64_expand_epilogue)
5526 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
5527 (aarch64_anchor_offset): New function, split out from...
5528 (aarch64_legitimize_address): ...here.
5529 (aarch64_builtin_vectorization_cost): Handle polynomial
5530 TYPE_VECTOR_SUBPARTS.
5531 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
5533 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
5534 number of elements from the PARALLEL rather than the mode.
5535 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
5536 rather than GET_MODE_BITSIZE.
5537 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
5538 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
5539 (aarch64_expand_vec_perm_const_1): Handle polynomial
5540 d->perm.length () and d->perm elements.
5541 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
5542 Apply to_constant to d->perm elements.
5543 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
5544 polynomial CONST_VECTOR_NUNITS.
5545 (aarch64_move_pointer): Take amount as a poly_int64 rather
5547 (aarch64_progress_pointer): Avoid temporary variable.
5548 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
5549 the mode attribute instead of GET_MODE.
5551 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5552 Alan Hayward <alan.hayward@arm.com>
5553 David Sherwood <david.sherwood@arm.com>
5555 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
5556 x exists before using it.
5557 (aarch64_add_constant_internal): Rename to...
5558 (aarch64_add_offset_1): ...this. Replace regnum with separate
5559 src and dest rtxes. Handle the case in which they're different,
5560 including when the offset is zero. Replace scratchreg with an rtx.
5561 Use 2 additions if there is no spare register into which we can
5562 move a 16-bit constant.
5563 (aarch64_add_constant): Delete.
5564 (aarch64_add_offset): Replace reg with separate src and dest
5565 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
5566 Use aarch64_add_offset_1.
5567 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
5568 an rtx rather than an int. Take the delta as a poly_int64
5569 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
5570 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
5571 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
5572 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
5573 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
5575 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
5576 aarch64_add_constant.
5578 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5580 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
5581 Use scalar_float_mode.
5583 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5585 * config/aarch64/aarch64-simd.md
5586 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
5587 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
5588 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
5589 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
5590 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
5591 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
5592 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
5593 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
5594 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
5595 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
5597 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5600 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
5601 targ_options->x_arm_arch_string is non NULL.
5603 2018-01-11 Tamar Christina <tamar.christina@arm.com>
5605 * config/aarch64/aarch64.h
5606 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
5608 2018-01-11 Sudakshina Das <sudi.das@arm.com>
5611 * expmed.c (emit_store_flag_force): Swap if const op0
5612 and change VOIDmode to mode of op0.
5614 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
5616 PR rtl-optimization/83761
5617 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
5618 than bytes to mode_for_size.
5620 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5623 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
5624 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
5627 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
5630 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
5631 when in layout mode.
5632 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
5633 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
5636 2018-01-10 Michael Collison <michael.collison@arm.com>
5638 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
5639 * config/aarch64/aarch64-option-extension.def: Add
5640 AARCH64_OPT_EXTENSION of 'fp16fml'.
5641 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5642 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
5643 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
5644 * config/aarch64/constraints.md (Ui7): New constraint.
5645 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
5646 (VFMLA_SEL_W): Ditto.
5649 (VFMLA16_LOW): New int iterator.
5650 (VFMLA16_HIGH): Ditto.
5651 (UNSPEC_FMLAL): New unspec.
5652 (UNSPEC_FMLSL): Ditto.
5653 (UNSPEC_FMLAL2): Ditto.
5654 (UNSPEC_FMLSL2): Ditto.
5655 (f16mac): New code attribute.
5656 * config/aarch64/aarch64-simd-builtins.def
5657 (aarch64_fmlal_lowv2sf): Ditto.
5658 (aarch64_fmlsl_lowv2sf): Ditto.
5659 (aarch64_fmlalq_lowv4sf): Ditto.
5660 (aarch64_fmlslq_lowv4sf): Ditto.
5661 (aarch64_fmlal_highv2sf): Ditto.
5662 (aarch64_fmlsl_highv2sf): Ditto.
5663 (aarch64_fmlalq_highv4sf): Ditto.
5664 (aarch64_fmlslq_highv4sf): Ditto.
5665 (aarch64_fmlal_lane_lowv2sf): Ditto.
5666 (aarch64_fmlsl_lane_lowv2sf): Ditto.
5667 (aarch64_fmlal_laneq_lowv2sf): Ditto.
5668 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
5669 (aarch64_fmlalq_lane_lowv4sf): Ditto.
5670 (aarch64_fmlsl_lane_lowv4sf): Ditto.
5671 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
5672 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
5673 (aarch64_fmlal_lane_highv2sf): Ditto.
5674 (aarch64_fmlsl_lane_highv2sf): Ditto.
5675 (aarch64_fmlal_laneq_highv2sf): Ditto.
5676 (aarch64_fmlsl_laneq_highv2sf): Ditto.
5677 (aarch64_fmlalq_lane_highv4sf): Ditto.
5678 (aarch64_fmlsl_lane_highv4sf): Ditto.
5679 (aarch64_fmlalq_laneq_highv4sf): Ditto.
5680 (aarch64_fmlsl_laneq_highv4sf): Ditto.
5681 * config/aarch64/aarch64-simd.md:
5682 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
5683 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5684 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
5685 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
5686 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
5687 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
5688 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
5689 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
5690 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
5691 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
5692 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
5693 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
5694 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
5695 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
5696 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
5697 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
5698 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
5699 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
5700 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
5701 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
5702 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
5703 (vfmlsl_low_u32): Ditto.
5704 (vfmlalq_low_u32): Ditto.
5705 (vfmlslq_low_u32): Ditto.
5706 (vfmlal_high_u32): Ditto.
5707 (vfmlsl_high_u32): Ditto.
5708 (vfmlalq_high_u32): Ditto.
5709 (vfmlslq_high_u32): Ditto.
5710 (vfmlal_lane_low_u32): Ditto.
5711 (vfmlsl_lane_low_u32): Ditto.
5712 (vfmlal_laneq_low_u32): Ditto.
5713 (vfmlsl_laneq_low_u32): Ditto.
5714 (vfmlalq_lane_low_u32): Ditto.
5715 (vfmlslq_lane_low_u32): Ditto.
5716 (vfmlalq_laneq_low_u32): Ditto.
5717 (vfmlslq_laneq_low_u32): Ditto.
5718 (vfmlal_lane_high_u32): Ditto.
5719 (vfmlsl_lane_high_u32): Ditto.
5720 (vfmlal_laneq_high_u32): Ditto.
5721 (vfmlsl_laneq_high_u32): Ditto.
5722 (vfmlalq_lane_high_u32): Ditto.
5723 (vfmlslq_lane_high_u32): Ditto.
5724 (vfmlalq_laneq_high_u32): Ditto.
5725 (vfmlslq_laneq_high_u32): Ditto.
5726 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
5727 (AARCH64_FL_FOR_ARCH8_4): New.
5728 (AARCH64_ISA_F16FML): New ISA flag.
5729 (TARGET_F16FML): New feature flag for fp16fml.
5730 (doc/invoke.texi): Document new fp16fml option.
5732 2018-01-10 Michael Collison <michael.collison@arm.com>
5734 * config/aarch64/aarch64-builtins.c:
5735 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
5736 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5737 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
5738 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
5739 (AARCH64_ISA_SHA3): New ISA flag.
5740 (TARGET_SHA3): New feature flag for sha3.
5741 * config/aarch64/iterators.md (sha512_op): New int attribute.
5742 (CRYPTO_SHA512): New int iterator.
5743 (UNSPEC_SHA512H): New unspec.
5744 (UNSPEC_SHA512H2): Ditto.
5745 (UNSPEC_SHA512SU0): Ditto.
5746 (UNSPEC_SHA512SU1): Ditto.
5747 * config/aarch64/aarch64-simd-builtins.def
5748 (aarch64_crypto_sha512hqv2di): New builtin.
5749 (aarch64_crypto_sha512h2qv2di): Ditto.
5750 (aarch64_crypto_sha512su0qv2di): Ditto.
5751 (aarch64_crypto_sha512su1qv2di): Ditto.
5752 (aarch64_eor3qv8hi): Ditto.
5753 (aarch64_rax1qv2di): Ditto.
5754 (aarch64_xarqv2di): Ditto.
5755 (aarch64_bcaxqv8hi): Ditto.
5756 * config/aarch64/aarch64-simd.md:
5757 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
5758 (aarch64_crypto_sha512su0qv2di): Ditto.
5759 (aarch64_crypto_sha512su1qv2di): Ditto.
5760 (aarch64_eor3qv8hi): Ditto.
5761 (aarch64_rax1qv2di): Ditto.
5762 (aarch64_xarqv2di): Ditto.
5763 (aarch64_bcaxqv8hi): Ditto.
5764 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
5765 (vsha512h2q_u64): Ditto.
5766 (vsha512su0q_u64): Ditto.
5767 (vsha512su1q_u64): Ditto.
5768 (veor3q_u16): Ditto.
5769 (vrax1q_u64): Ditto.
5771 (vbcaxq_u16): Ditto.
5772 * config/arm/types.md (crypto_sha512): New type attribute.
5773 (crypto_sha3): Ditto.
5774 (doc/invoke.texi): Document new sha3 option.
5776 2018-01-10 Michael Collison <michael.collison@arm.com>
5778 * config/aarch64/aarch64-builtins.c:
5779 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
5780 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5781 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
5782 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
5783 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
5784 (AARCH64_ISA_SM4): New ISA flag.
5785 (TARGET_SM4): New feature flag for sm4.
5786 * config/aarch64/aarch64-simd-builtins.def
5787 (aarch64_sm3ss1qv4si): Ditto.
5788 (aarch64_sm3tt1aq4si): Ditto.
5789 (aarch64_sm3tt1bq4si): Ditto.
5790 (aarch64_sm3tt2aq4si): Ditto.
5791 (aarch64_sm3tt2bq4si): Ditto.
5792 (aarch64_sm3partw1qv4si): Ditto.
5793 (aarch64_sm3partw2qv4si): Ditto.
5794 (aarch64_sm4eqv4si): Ditto.
5795 (aarch64_sm4ekeyqv4si): Ditto.
5796 * config/aarch64/aarch64-simd.md:
5797 (aarch64_sm3ss1qv4si): Ditto.
5798 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
5799 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
5800 (aarch64_sm4eqv4si): Ditto.
5801 (aarch64_sm4ekeyqv4si): Ditto.
5802 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
5803 (sm3part_op): Ditto.
5804 (CRYPTO_SM3TT): Ditto.
5805 (CRYPTO_SM3PART): Ditto.
5806 (UNSPEC_SM3SS1): New unspec.
5807 (UNSPEC_SM3TT1A): Ditto.
5808 (UNSPEC_SM3TT1B): Ditto.
5809 (UNSPEC_SM3TT2A): Ditto.
5810 (UNSPEC_SM3TT2B): Ditto.
5811 (UNSPEC_SM3PARTW1): Ditto.
5812 (UNSPEC_SM3PARTW2): Ditto.
5813 (UNSPEC_SM4E): Ditto.
5814 (UNSPEC_SM4EKEY): Ditto.
5815 * config/aarch64/constraints.md (Ui2): New constraint.
5816 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
5817 * config/arm/types.md (crypto_sm3): New type attribute.
5818 (crypto_sm4): Ditto.
5819 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
5820 (vsm3tt1aq_u32): Ditto.
5821 (vsm3tt1bq_u32): Ditto.
5822 (vsm3tt2aq_u32): Ditto.
5823 (vsm3tt2bq_u32): Ditto.
5824 (vsm3partw1q_u32): Ditto.
5825 (vsm3partw2q_u32): Ditto.
5826 (vsm4eq_u32): Ditto.
5827 (vsm4ekeyq_u32): Ditto.
5828 (doc/invoke.texi): Document new sm4 option.
5830 2018-01-10 Michael Collison <michael.collison@arm.com>
5832 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
5833 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
5834 (AARCH64_FL_FOR_ARCH8_4): New.
5835 (AARCH64_FL_V8_4): New flag.
5836 (doc/invoke.texi): Document new armv8.4-a option.
5838 2018-01-10 Michael Collison <michael.collison@arm.com>
5840 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
5841 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
5842 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
5843 * config/aarch64/aarch64-option-extension.def: Add
5844 AARCH64_OPT_EXTENSION of 'sha2'.
5845 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
5846 (crypto): Disable sha2 and aes if crypto disabled.
5847 (crypto): Enable aes and sha2 if enabled.
5848 (simd): Disable sha2 and aes if simd disabled.
5849 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
5851 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
5852 (TARGET_SHA2): New feature flag for sha2.
5853 (TARGET_AES): New feature flag for aes.
5854 * config/aarch64/aarch64-simd.md:
5855 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
5856 conditional on TARGET_AES.
5857 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
5858 (aarch64_crypto_sha1hsi): Make pattern conditional
5860 (aarch64_crypto_sha1hv4si): Ditto.
5861 (aarch64_be_crypto_sha1hv4si): Ditto.
5862 (aarch64_crypto_sha1su1v4si): Ditto.
5863 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
5864 (aarch64_crypto_sha1su0v4si): Ditto.
5865 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
5866 (aarch64_crypto_sha256su0v4si): Ditto.
5867 (aarch64_crypto_sha256su1v4si): Ditto.
5868 (doc/invoke.texi): Document new aes and sha2 options.
5870 2018-01-10 Martin Sebor <msebor@redhat.com>
5872 PR tree-optimization/83781
5873 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
5876 2018-01-11 Martin Sebor <msebor@gmail.com>
5877 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5879 PR tree-optimization/83501
5880 PR tree-optimization/81703
5882 * tree-ssa-strlen.c (get_string_cst): Rename...
5883 (get_string_len): ...to this. Handle global constants.
5884 (handle_char_store): Adjust.
5886 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
5887 Jim Wilson <jimw@sifive.com>
5889 * config/riscv/riscv-protos.h (riscv_output_return): New.
5890 * config/riscv/riscv.c (struct machine_function): New naked_p field.
5891 (riscv_attribute_table, riscv_output_return),
5892 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
5893 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
5894 (riscv_compute_frame_info): Only compute frame->mask if not a naked
5896 (riscv_expand_prologue): Add early return for naked function.
5897 (riscv_expand_epilogue): Likewise.
5898 (riscv_function_ok_for_sibcall): Return false for naked function.
5899 (riscv_set_current_function): New.
5900 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
5901 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
5902 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
5903 * doc/extend.texi (RISC-V Function Attributes): New.
5905 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
5907 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
5908 check for 128-bit long double before checking TCmode.
5909 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
5910 128-bit long doubles before checking TFmode or TCmode.
5911 (FLOAT128_IBM_P): Likewise.
5913 2018-01-10 Martin Sebor <msebor@redhat.com>
5915 PR tree-optimization/83671
5916 * builtins.c (c_strlen): Unconditionally return zero for the empty
5918 Use -Warray-bounds for warnings.
5919 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
5920 for non-constant array indices with COMPONENT_REF, arrays of
5921 arrays, and pointers to arrays.
5922 (gimple_fold_builtin_strlen): Determine and set length range for
5923 non-constant character arrays.
5925 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
5928 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
5931 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
5933 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
5935 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
5938 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
5939 VECTOR_MEM_ALTIVEC_OR_VSX_P.
5940 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
5941 indexed_or_indirect_operand predicate.
5942 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
5943 (*vsx_le_perm_load_v8hi): Likewise.
5944 (*vsx_le_perm_load_v16qi): Likewise.
5945 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
5946 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
5947 (*vsx_le_perm_store_v8hi): Likewise.
5948 (*vsx_le_perm_store_v16qi): Likewise.
5949 (eight unnamed splitters): Likewise.
5951 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
5953 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
5954 * config/rs6000/emmintrin.h: Likewise.
5955 * config/rs6000/mmintrin.h: Likewise.
5956 * config/rs6000/xmmintrin.h: Likewise.
5958 2018-01-10 David Malcolm <dmalcolm@redhat.com>
5961 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
5963 * tree.c (tree_nop_conversion): Return true for location wrapper
5965 (maybe_wrap_with_location): New function.
5966 (selftest::check_strip_nops): New function.
5967 (selftest::test_location_wrappers): New function.
5968 (selftest::tree_c_tests): Call it.
5969 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
5970 (maybe_wrap_with_location): New decl.
5971 (EXPR_LOCATION_WRAPPER_P): New macro.
5972 (location_wrapper_p): New inline function.
5973 (tree_strip_any_location_wrapper): New inline function.
5975 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
5978 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
5979 stack_realign_offset for the largest alignment of stack slot
5981 (ix86_find_max_used_stack_alignment): New function.
5982 (ix86_finalize_stack_frame_flags): Use it. Set
5983 max_used_stack_alignment if we don't realign stack.
5984 * config/i386/i386.h (machine_function): Add
5985 max_used_stack_alignment.
5987 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
5989 * config/arm/arm.opt (-mbranch-cost): New option.
5990 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
5993 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
5996 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
5997 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
5999 2018-01-10 Richard Biener <rguenther@suse.de>
6002 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
6003 early out so it also covers the case where we have a non-NULL
6006 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6008 PR tree-optimization/83753
6009 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
6010 for non-strided grouped accesses if the number of elements is 1.
6012 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
6015 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
6016 * i386.h (TARGET_USE_GATHER): Define.
6017 * x86-tune.def (X86_TUNE_USE_GATHER): New.
6019 2018-01-10 Martin Liska <mliska@suse.cz>
6022 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
6023 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
6025 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
6026 CLEANUP_NO_PARTITIONING is not set.
6028 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
6030 * doc/rtl.texi: Remove documentation of (const ...) wrappers
6031 for vectors, as a partial revert of r254296.
6032 * rtl.h (const_vec_p): Delete.
6033 (const_vec_duplicate_p): Don't test for vector CONSTs.
6034 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
6035 * expmed.c (make_tree): Likewise.
6038 * common.md (E, F): Use CONSTANT_P instead of checking for
6040 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
6041 checking for CONST_VECTOR.
6043 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6046 * predict.c (force_edge_cold): Handle in more sane way edges
6049 2018-01-09 Carl Love <cel@us.ibm.com>
6051 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
6053 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
6054 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
6055 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
6056 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
6057 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
6058 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
6059 * config/rs6000/rs6000-protos.h: Add extern defition for
6060 rs6000_generate_float2_double_code.
6061 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
6063 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
6064 (float2_v2df): Add define_expand.
6066 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
6069 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
6070 op_mode in the force_to_mode call.
6072 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6074 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
6075 instead of checking each element individually.
6076 (aarch64_evpc_uzp): Likewise.
6077 (aarch64_evpc_zip): Likewise.
6078 (aarch64_evpc_ext): Likewise.
6079 (aarch64_evpc_rev): Likewise.
6080 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
6081 instead of checking each element individually. Return true without
6083 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
6084 whether all selected elements come from the same input, instead of
6085 checking each element individually. Remove calls to gen_rtx_REG,
6086 start_sequence and end_sequence and instead assert that no rtl is
6089 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6091 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
6092 order of HIGH and CONST checks.
6094 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
6096 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
6097 if the destination isn't an SSA_NAME.
6099 2018-01-09 Richard Biener <rguenther@suse.de>
6101 PR tree-optimization/83668
6102 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
6104 (canonicalize_loop_form): ... here, renamed from ...
6105 (canonicalize_loop_closed_ssa_form): ... this and amended to
6106 swap successor edges for loop exit blocks to make us use
6107 the RPO order we need for initial schedule generation.
6109 2018-01-09 Joseph Myers <joseph@codesourcery.com>
6111 PR tree-optimization/64811
6112 * match.pd: When optimizing comparisons with Inf, avoid
6113 introducing or losing exceptions from comparisons with NaN.
6115 2018-01-09 Martin Liska <mliska@suse.cz>
6118 * asan.c (shadow_mem_size): Add gcc_assert.
6120 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
6122 Don't save registers in main().
6125 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
6126 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
6127 * config/avr/avr.c (avr_set_current_function): Don't error if
6128 naked, OS_task or OS_main are specified at the same time.
6129 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
6131 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
6133 * common/config/avr/avr-common.c (avr_option_optimization_table):
6134 Switch on -mmain-is-OS_task for optimizing compilations.
6136 2018-01-09 Richard Biener <rguenther@suse.de>
6138 PR tree-optimization/83572
6139 * graphite.c: Include cfganal.h.
6140 (graphite_transform_loops): Connect infinite loops to exit
6141 and remove fake edges at the end.
6143 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6145 * ipa-inline.c (edge_badness): Revert accidental checkin.
6147 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
6150 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
6151 symbols; not inline clones.
6153 2018-01-09 Jakub Jelinek <jakub@redhat.com>
6156 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
6157 hard registers. Formatting fixes.
6159 PR preprocessor/83722
6160 * gcc.c (try_generate_repro): Pass
6161 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
6162 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
6165 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
6166 Kito Cheng <kito.cheng@gmail.com>
6168 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
6169 (riscv_leaf_function_p): Delete.
6170 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
6172 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6174 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
6176 (do_ifelse): New function.
6177 (do_isel): New function.
6178 (do_sub3): New function.
6179 (do_add3): New function.
6180 (do_load_mask_compare): New function.
6181 (do_overlap_load_compare): New function.
6182 (expand_compare_loop): New function.
6183 (expand_block_compare): Call expand_compare_loop() when appropriate.
6184 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
6186 (-mblock-compare-inline-loop-limit): New option.
6188 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
6191 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
6192 Reverse order of second and third operands in first alternative.
6193 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
6194 of first and second elements in UNSPEC_VPERMR vector.
6195 (altivec_expand_vec_perm_le): Likewise.
6197 2018-01-08 Jeff Law <law@redhat.com>
6199 PR rtl-optimizatin/81308
6200 * tree-switch-conversion.c (cfg_altered): New file scoped static.
6201 (process_switch): If group_case_labels makes a change, then set
6203 (pass_convert_switch::execute): If a switch is converted, then
6204 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
6206 PR rtl-optimization/81308
6207 * recog.c (split_all_insns): Conditionally cleanup the CFG after
6210 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
6212 PR target/83663 - Revert r255946
6213 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
6214 generation for cases where splatting a value is not useful.
6215 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
6216 across a vec_duplicate and a paradoxical subreg forming a vector
6217 mode to a vec_concat.
6219 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6221 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
6222 -march=armv8.3-a variants.
6223 * config/arm/t-multilib: Likewise.
6224 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
6226 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
6228 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
6230 (cceq_ior_compare_complement): Give it a name so I can use it, and
6231 change boolean_or_operator predicate to boolean_operator so it can
6232 be used to generate a crand.
6233 (eqne): New code iterator.
6234 (bd/bd_neg): New code_attrs.
6235 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
6236 a single define_insn.
6237 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
6238 decrement (bdnzt/bdnzf/bdzt/bdzf).
6239 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
6240 with the new names of the branch decrement patterns, and added the
6241 names of the branch decrement conditional patterns.
6243 2018-01-08 Richard Biener <rguenther@suse.de>
6245 PR tree-optimization/83563
6246 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
6249 2018-01-08 Richard Biener <rguenther@suse.de>
6252 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
6254 2018-01-08 Richard Biener <rguenther@suse.de>
6256 PR tree-optimization/83685
6257 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
6258 references to abnormals.
6260 2018-01-08 Richard Biener <rguenther@suse.de>
6263 * dwarf2out.c (output_indirect_strings): Handle empty
6264 skeleton_debug_str_hash.
6265 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
6267 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6269 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
6270 (emit_store_direct): Likewise.
6271 (arc_trampoline_adjust_address): Likewise.
6272 (arc_asm_trampoline_template): New function.
6273 (arc_initialize_trampoline): Use asm_trampoline_template.
6274 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
6275 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
6276 * config/arc/arc.md (flush_icache): Delete pattern.
6278 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
6280 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
6281 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
6284 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6287 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
6288 by not USED_FOR_TARGET.
6289 (make_pass_resolve_sw_modes): Likewise.
6291 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
6293 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
6296 2018-01-08 Richard Biener <rguenther@suse.de>
6299 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
6301 2018-01-08 Richard Biener <rguenther@suse.de>
6304 * match.pd ((t * 2) / 2) -> t): Add missing :c.
6306 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
6309 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
6310 basic blocks with a small number of successors.
6311 (convert_control_dep_chain_into_preds): Improve handling of
6313 (dump_predicates): Split apart into...
6314 (dump_pred_chain): ...here...
6315 (dump_pred_info): ...and here.
6316 (can_one_predicate_be_invalidated_p): Add debugging printfs.
6317 (can_chain_union_be_invalidated_p): Improve check for invalidation
6319 (uninit_uses_cannot_happen): Avoid unnecessary if
6320 convert_control_dep_chain_into_preds yielded nothing.
6322 2018-01-06 Martin Sebor <msebor@redhat.com>
6324 PR tree-optimization/83640
6325 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
6326 subtracting negative offset from size.
6327 (builtin_access::overlap): Adjust offset bounds of the access to fall
6328 within the size of the object if possible.
6330 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
6332 PR rtl-optimization/83699
6333 * expmed.c (extract_bit_field_1): Restrict the vector usage of
6334 extract_bit_field_as_subreg to cases in which the extracted
6335 value is also a vector.
6337 * lra-constraints.c (process_alt_operands): Test for the equivalence
6338 substitutions when detecting a possible reload cycle.
6340 2018-01-06 Jakub Jelinek <jakub@redhat.com>
6343 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
6344 by default if flag_selective_schedling{,2}. Formatting fixes.
6346 PR rtl-optimization/83682
6347 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
6348 if it has non-VECTOR_MODE element mode.
6349 (vec_duplicate_p): Likewise.
6352 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
6353 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
6355 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6358 * config/i386/i386-builtin.def
6359 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
6360 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
6361 Require also OPTION_MASK_ISA_AVX512F in addition to
6362 OPTION_MASK_ISA_GFNI.
6363 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
6364 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
6365 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
6366 to OPTION_MASK_ISA_GFNI.
6367 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
6368 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
6369 OPTION_MASK_ISA_AVX512BW.
6370 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
6371 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
6372 addition to OPTION_MASK_ISA_GFNI.
6373 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
6374 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
6375 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
6376 to OPTION_MASK_ISA_GFNI.
6377 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
6378 a requirement for all ISAs rather than any of them with a few
6380 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
6382 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
6383 bitmasks to be enabled with 3 exceptions, instead of requiring any
6384 enabled ISA with lots of exceptions.
6385 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
6386 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
6387 Change avx512bw in isa attribute to avx512f.
6388 * config/i386/sgxintrin.h: Add license boilerplate.
6389 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
6390 to __AVX512F__ and __AVX512VL to __AVX512VL__.
6391 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
6392 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
6394 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
6395 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
6396 temporarily sse2 rather than sse if not enabled already.
6399 * config/i386/sse.md (VI248_VLBW): Rename to ...
6400 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
6401 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
6402 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
6403 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
6404 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
6405 mode iterator instead of VI248_VLBW.
6407 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
6409 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
6410 (record_modified): Skip clobbers; add debug output.
6411 (param_change_prob): Use sreal frequencies.
6413 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6415 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
6416 punt for user-aligned variables.
6418 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
6420 * tree-chrec.c (chrec_contains_symbols): Return true for
6423 2018-01-05 Sudakshina Das <sudi.das@arm.com>
6426 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
6427 of (x|y) == x for BICS pattern.
6429 2018-01-05 Jakub Jelinek <jakub@redhat.com>
6431 PR tree-optimization/83605
6432 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
6433 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
6436 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
6438 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
6439 * config/epiphany/rtems.h: New file.
6441 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6442 Uros Bizjak <ubizjak@gmail.com>
6445 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
6446 QIreg_operand instead of register_operand predicate.
6447 * config/i386/i386.c (ix86_rop_should_change_byte_p,
6448 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
6449 comments instead of -fmitigate[-_]rop.
6451 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
6454 * cgraphunit.c (symbol_table::compile): Switch to text_section
6455 before calling assembly_start debug hook.
6456 * run-rtl-passes.c (run_rtl_passes): Likewise.
6459 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6461 * tree-vrp.c (extract_range_from_binary_expr_1): Check
6462 range_int_cst_p rather than !symbolic_range_p before calling
6463 extract_range_from_multiplicative_op_1.
6465 2018-01-04 Jeff Law <law@redhat.com>
6467 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
6468 redundant test in assertion.
6470 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6472 * doc/rtl.texi: Document machine_mode wrapper classes.
6474 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6476 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
6479 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6481 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
6482 the VEC_PERM_EXPR fold to fail.
6484 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6487 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
6488 to switched_sections.
6490 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6493 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
6496 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
6499 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
6500 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
6502 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6505 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
6506 is BLKmode and bitpos not zero or mode change is needed.
6508 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
6511 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
6514 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
6517 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
6518 instead of MULT rtx. Update all corresponding splitters.
6520 (*ssub<modesuffix>): Ditto.
6522 (*cmp_sadd_di): Update split patterns.
6523 (*cmp_sadd_si): Ditto.
6524 (*cmp_sadd_sidi): Ditto.
6525 (*cmp_ssub_di): Ditto.
6526 (*cmp_ssub_si): Ditto.
6527 (*cmp_ssub_sidi): Ditto.
6528 * config/alpha/predicates.md (const23_operand): New predicate.
6529 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
6530 Look for ASHIFT, not MULT inner operand.
6531 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
6533 2018-01-04 Martin Liska <mliska@suse.cz>
6535 PR gcov-profile/83669
6536 * gcov.c (output_intermediate_file): Add version to intermediate
6538 * doc/gcov.texi: Document new field 'version' in intermediate
6539 file format. Fix location of '-k' option of gcov command.
6541 2018-01-04 Martin Liska <mliska@suse.cz>
6544 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
6546 2018-01-04 Jakub Jelinek <jakub@redhat.com>
6548 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
6550 2018-01-03 Martin Sebor <msebor@redhat.com>
6552 PR tree-optimization/83655
6553 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
6554 checking calls with invalid arguments.
6556 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6558 * tree-vect-stmts.c (vect_get_store_rhs): New function.
6559 (vectorizable_mask_load_store): Delete.
6560 (vectorizable_call): Return false for masked loads and stores.
6561 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
6562 instead of gimple_assign_rhs1.
6563 (vectorizable_load): Handle IFN_MASK_LOAD.
6564 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
6566 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6568 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
6570 (vectorizable_mask_load_store): ...here.
6571 (vectorizable_load): ...and here.
6573 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6575 * tree-vect-stmts.c (vect_build_all_ones_mask)
6576 (vect_build_zero_merge_argument): New functions, split out from...
6577 (vectorizable_load): ...here.
6579 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6581 * tree-vect-stmts.c (vect_check_store_rhs): New function,
6583 (vectorizable_mask_load_store): ...here.
6584 (vectorizable_store): ...and here.
6586 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6588 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
6590 (vectorizable_mask_load_store): ...here.
6592 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6594 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
6595 (vect_model_store_cost): Take a vec_load_store_type instead of a
6597 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
6598 (vect_model_store_cost): Take a vec_load_store_type instead of a
6600 (vectorizable_mask_load_store): Update accordingly.
6601 (vectorizable_store): Likewise.
6602 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
6604 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6606 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
6607 IFN_MASK_LOAD calls here rather than...
6608 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
6610 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6611 Alan Hayward <alan.hayward@arm.com>
6612 David Sherwood <david.sherwood@arm.com>
6614 * expmed.c (extract_bit_field_1): For vector extracts,
6615 fall back to extract_bit_field_as_subreg if vec_extract
6618 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6619 Alan Hayward <alan.hayward@arm.com>
6620 David Sherwood <david.sherwood@arm.com>
6622 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
6623 they are variable or constant sized.
6624 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
6625 slots for constant-sized data.
6627 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6628 Alan Hayward <alan.hayward@arm.com>
6629 David Sherwood <david.sherwood@arm.com>
6631 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
6632 handling COND_EXPRs with boolean comparisons, try to find a better
6633 basis for the mask type than the boolean itself.
6635 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6637 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
6638 is calculated and how it can be overridden.
6639 * genmodes.c (max_bitsize_mode_any_mode): New variable.
6640 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
6642 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
6645 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6646 Alan Hayward <alan.hayward@arm.com>
6647 David Sherwood <david.sherwood@arm.com>
6649 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
6650 Remove the mode argument.
6651 (aarch64_simd_valid_immediate): Remove the mode and inverse
6653 * config/aarch64/iterators.md (bitsize): New iterator.
6654 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
6655 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
6656 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
6657 aarch64_simd_valid_immediate.
6658 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
6659 (aarch64_reg_or_bic_imm): Likewise.
6660 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
6661 with an insn_type enum and msl with a modifier_type enum.
6662 Replace element_width with a scalar_mode. Change the shift
6663 to unsigned int. Add constructors for scalar_float_mode and
6664 scalar_int_mode elements.
6665 (aarch64_vect_float_const_representable_p): Delete.
6666 (aarch64_can_const_movi_rtx_p)
6667 (aarch64_simd_scalar_immediate_valid_for_move)
6668 (aarch64_simd_make_constant): Update call to
6669 aarch64_simd_valid_immediate.
6670 (aarch64_advsimd_valid_immediate_hs): New function.
6671 (aarch64_advsimd_valid_immediate): Likewise.
6672 (aarch64_simd_valid_immediate): Remove mode and inverse
6673 arguments. Rewrite to use the above. Use const_vec_duplicate_p
6674 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
6675 and aarch64_float_const_representable_p on the result.
6676 (aarch64_output_simd_mov_immediate): Remove mode argument.
6677 Update call to aarch64_simd_valid_immediate and use of
6678 simd_immediate_info.
6679 (aarch64_output_scalar_simd_mov_immediate): Update call
6682 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6683 Alan Hayward <alan.hayward@arm.com>
6684 David Sherwood <david.sherwood@arm.com>
6686 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
6687 (mode_nunits): Likewise CONST_MODE_NUNITS.
6688 * machmode.def (ADJUST_NUNITS): Document.
6689 * genmodes.c (mode_data::need_nunits_adj): New field.
6690 (blank_mode): Update accordingly.
6691 (adj_nunits): New variable.
6692 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
6694 (emit_mode_size_inline): Set need_bytesize_adj for all modes
6695 listed in adj_nunits.
6696 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
6697 listed in adj_nunits. Don't emit case statements for such modes.
6698 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
6699 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
6700 nothing if adj_nunits is nonnull.
6701 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
6702 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
6703 (emit_mode_fbit): Update use of print_maybe_const_decl.
6704 (emit_move_size): Likewise. Treat the array as non-const
6706 (emit_mode_adjustments): Handle adj_nunits.
6708 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6710 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
6711 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
6712 (VECTOR_MODES): Use it.
6713 (make_vector_modes): Take the prefix as an argument.
6715 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6716 Alan Hayward <alan.hayward@arm.com>
6717 David Sherwood <david.sherwood@arm.com>
6719 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
6720 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
6721 for MODE_VECTOR_BOOL.
6722 * machmode.def (VECTOR_BOOL_MODE): Document.
6723 * genmodes.c (VECTOR_BOOL_MODE): New macro.
6724 (make_vector_bool_mode): New function.
6725 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
6727 * lto-streamer-in.c (lto_input_mode_table): Likewise.
6728 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
6730 * stor-layout.c (int_mode_for_mode): Likewise.
6731 * tree.c (build_vector_type_for_mode): Likewise.
6732 * varasm.c (output_constant_pool_2): Likewise.
6733 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
6734 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
6735 for MODE_VECTOR_BOOL.
6736 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
6737 of mode class checks.
6738 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
6739 instead of a list of mode class checks.
6740 (expand_vector_scalar_condition): Likewise.
6741 (type_for_widest_vector_mode): Handle BImode as an inner mode.
6743 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6744 Alan Hayward <alan.hayward@arm.com>
6745 David Sherwood <david.sherwood@arm.com>
6747 * machmode.h (mode_size): Change from unsigned short to
6749 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
6750 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6751 or if measurement_type is not polynomial.
6752 (fixed_size_mode::includes_p): Check for constant-sized modes.
6753 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
6754 return a poly_uint16 rather than an unsigned short.
6755 (emit_mode_size): Change the type of mode_size from unsigned short
6756 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
6757 (emit_mode_adjustments): Cope with polynomial vector sizes.
6758 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6760 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6762 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
6763 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
6764 * caller-save.c (setup_save_areas): Likewise.
6765 (replace_reg_with_saved_mem): Likewise.
6766 * calls.c (emit_library_call_value_1): Likewise.
6767 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
6768 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
6769 (gen_lowpart_for_combine): Likewise.
6770 * convert.c (convert_to_integer_1): Likewise.
6771 * cse.c (equiv_constant, cse_insn): Likewise.
6772 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
6773 (cselib_subst_to_values): Likewise.
6774 * dce.c (word_dce_process_block): Likewise.
6775 * df-problems.c (df_word_lr_mark_ref): Likewise.
6776 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
6777 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
6778 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
6779 (rtl_for_decl_location): Likewise.
6780 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
6781 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
6782 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
6783 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
6784 (expand_expr_real_1): Likewise.
6785 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
6786 (pad_below): Likewise.
6787 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6788 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
6789 * ira.c (get_subreg_tracking_sizes): Likewise.
6790 * ira-build.c (ira_create_allocno_objects): Likewise.
6791 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
6792 (ira_sort_regnos_for_alter_reg): Likewise.
6793 * ira-costs.c (record_operand_costs): Likewise.
6794 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
6795 (resolve_simple_move): Likewise.
6796 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
6797 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
6798 (lra_constraints): Likewise.
6799 (CONST_POOL_OK_P): Reject variable-sized modes.
6800 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
6801 (add_pseudo_to_slot, lra_spill): Likewise.
6802 * omp-low.c (omp_clause_aligned_alignment): Likewise.
6803 * optabs-query.c (get_best_extraction_insn): Likewise.
6804 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
6805 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
6806 (expand_mult_highpart, valid_multiword_target_p): Likewise.
6807 * recog.c (offsettable_address_addr_space_p): Likewise.
6808 * regcprop.c (maybe_mode_change): Likewise.
6809 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
6810 * regrename.c (build_def_use): Likewise.
6811 * regstat.c (dump_reg_info): Likewise.
6812 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
6813 (find_reloads, find_reloads_subreg_address): Likewise.
6814 * reload1.c (eliminate_regs_1): Likewise.
6815 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
6816 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
6817 (simplify_binary_operation_1, simplify_subreg): Likewise.
6818 * targhooks.c (default_function_arg_padding): Likewise.
6819 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
6820 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
6821 (verify_gimple_assign_ternary): Likewise.
6822 * tree-inline.c (estimate_move_cost): Likewise.
6823 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
6824 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
6825 (get_address_cost_ainc): Likewise.
6826 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
6827 (vect_supportable_dr_alignment): Likewise.
6828 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6829 (vectorizable_reduction): Likewise.
6830 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
6831 (vectorizable_operation, vectorizable_load): Likewise.
6832 * tree.c (build_same_sized_truth_vector_type): Likewise.
6833 * valtrack.c (cleanup_auto_inc_dec): Likewise.
6834 * var-tracking.c (emit_note_insn_var_location): Likewise.
6835 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
6836 (ADDR_VEC_ALIGN): Likewise.
6838 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6839 Alan Hayward <alan.hayward@arm.com>
6840 David Sherwood <david.sherwood@arm.com>
6842 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
6844 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
6845 or if measurement_type is polynomial.
6846 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
6847 * combine.c (make_extraction): Likewise.
6848 * dse.c (find_shift_sequence): Likewise.
6849 * dwarf2out.c (mem_loc_descriptor): Likewise.
6850 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
6851 (extract_bit_field, extract_low_bits): Likewise.
6852 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
6853 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
6854 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
6855 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
6856 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
6857 * reload.c (find_reloads): Likewise.
6858 * reload1.c (alter_reg): Likewise.
6859 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
6860 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
6861 * tree-if-conv.c (predicate_mem_writes): Likewise.
6862 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
6863 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
6864 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
6865 * valtrack.c (dead_debug_insert_temp): Likewise.
6866 * varasm.c (mergeable_constant_section): Likewise.
6867 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
6869 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6870 Alan Hayward <alan.hayward@arm.com>
6871 David Sherwood <david.sherwood@arm.com>
6873 * expr.c (expand_assignment): Cope with polynomial mode sizes
6874 when assigning to a CONCAT.
6876 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6877 Alan Hayward <alan.hayward@arm.com>
6878 David Sherwood <david.sherwood@arm.com>
6880 * machmode.h (mode_precision): Change from unsigned short to
6882 (mode_to_precision): Return a poly_uint16 rather than an unsigned
6884 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
6885 or if measurement_type is not polynomial.
6886 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
6887 in which the mode is already known to be a scalar_int_mode.
6888 * genmodes.c (emit_mode_precision): Change the type of mode_precision
6889 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
6891 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
6892 for GET_MODE_PRECISION.
6893 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
6894 for GET_MODE_PRECISION.
6895 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
6897 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
6898 (expand_field_assignment, make_extraction): Likewise.
6899 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
6900 (get_last_value): Likewise.
6901 * convert.c (convert_to_integer_1): Likewise.
6902 * cse.c (cse_insn): Likewise.
6903 * expr.c (expand_expr_real_1): Likewise.
6904 * lra-constraints.c (simplify_operand_subreg): Likewise.
6905 * optabs-query.c (can_atomic_load_p): Likewise.
6906 * optabs.c (expand_atomic_load): Likewise.
6907 (expand_atomic_store): Likewise.
6908 * ree.c (combine_reaching_defs): Likewise.
6909 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
6910 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
6911 * tree.h (type_has_mode_precision_p): Likewise.
6912 * ubsan.c (instrument_si_overflow): Likewise.
6914 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6915 Alan Hayward <alan.hayward@arm.com>
6916 David Sherwood <david.sherwood@arm.com>
6918 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
6919 polynomial numbers of units.
6920 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
6921 (valid_vector_subparts_p): New function.
6922 (build_vector_type): Remove temporary shim and take the number
6923 of units as a poly_uint64 rather than an int.
6924 (build_opaque_vector_type): Take the number of units as a
6925 poly_uint64 rather than an int.
6926 * tree.c (build_vector_from_ctor): Handle polynomial
6927 TYPE_VECTOR_SUBPARTS.
6928 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
6929 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
6930 (build_vector_from_val): If the number of units is variable,
6931 use build_vec_duplicate_cst for constant operands and
6932 VEC_DUPLICATE_EXPR otherwise.
6933 (make_vector_type): Remove temporary is_constant ().
6934 (build_vector_type, build_opaque_vector_type): Take the number of
6935 units as a poly_uint64 rather than an int.
6936 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
6938 * cfgexpand.c (expand_debug_expr): Likewise.
6939 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
6940 (store_constructor, expand_expr_real_1): Likewise.
6941 (const_scalar_mask_from_tree): Likewise.
6942 * fold-const-call.c (fold_const_reduction): Likewise.
6943 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
6944 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
6945 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
6946 (fold_relational_const): Likewise.
6947 (native_interpret_vector): Likewise. Change the size from an
6948 int to an unsigned int.
6949 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
6950 TYPE_VECTOR_SUBPARTS.
6951 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
6952 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
6953 duplicating a non-constant operand into a variable-length vector.
6954 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
6955 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
6956 * ipa-icf.c (sem_variable::equals): Likewise.
6957 * match.pd: Likewise.
6958 * omp-simd-clone.c (simd_clone_subparts): Likewise.
6959 * print-tree.c (print_node): Likewise.
6960 * stor-layout.c (layout_type): Likewise.
6961 * targhooks.c (default_builtin_vectorization_cost): Likewise.
6962 * tree-cfg.c (verify_gimple_comparison): Likewise.
6963 (verify_gimple_assign_binary): Likewise.
6964 (verify_gimple_assign_ternary): Likewise.
6965 (verify_gimple_assign_single): Likewise.
6966 * tree-pretty-print.c (dump_generic_node): Likewise.
6967 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
6968 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
6969 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
6970 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
6971 (vect_shift_permute_load_chain): Likewise.
6972 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
6973 (expand_vector_condition, optimize_vector_constructor): Likewise.
6974 (lower_vec_perm, get_compute_type): Likewise.
6975 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
6976 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
6977 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
6978 (vect_recog_mask_conversion_pattern): Likewise.
6979 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
6980 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
6981 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
6982 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
6983 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
6984 (vectorizable_shift, vectorizable_operation, vectorizable_store)
6985 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
6986 (supportable_widening_operation): Likewise.
6987 (supportable_narrowing_operation): Likewise.
6988 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
6990 * varasm.c (output_constant): Likewise.
6992 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
6993 Alan Hayward <alan.hayward@arm.com>
6994 David Sherwood <david.sherwood@arm.com>
6996 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
6997 so that both the length == 3 and length != 3 cases set up their
6998 own permute vectors. Add comments explaining why we know the
6999 number of elements is constant.
7000 (vect_permute_load_chain): Likewise.
7002 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7003 Alan Hayward <alan.hayward@arm.com>
7004 David Sherwood <david.sherwood@arm.com>
7006 * machmode.h (mode_nunits): Change from unsigned char to
7008 (ONLY_FIXED_SIZE_MODES): New macro.
7009 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
7010 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
7011 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
7013 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
7014 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
7015 or if measurement_type is not polynomial.
7016 * genmodes.c (ZERO_COEFFS): New macro.
7017 (emit_mode_nunits_inline): Make mode_nunits_inline return a
7019 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
7020 Use ZERO_COEFFS when emitting initializers.
7021 * data-streamer.h (bp_pack_poly_value): New function.
7022 (bp_unpack_poly_value): Likewise.
7023 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
7024 for GET_MODE_NUNITS.
7025 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
7026 for GET_MODE_NUNITS.
7027 * tree.c (make_vector_type): Remove temporary shim and make
7028 the real function take the number of units as a poly_uint64
7030 (build_vector_type_for_mode): Handle polynomial nunits.
7031 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
7032 * emit-rtl.c (const_vec_series_p_1): Likewise.
7033 (gen_rtx_CONST_VECTOR): Likewise.
7034 * fold-const.c (test_vec_duplicate_folding): Likewise.
7035 * genrecog.c (validate_pattern): Likewise.
7036 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
7037 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
7038 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
7039 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
7040 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
7041 * rtlanal.c (subreg_get_info): Likewise.
7042 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7043 (vect_grouped_load_supported): Likewise.
7044 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
7045 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
7046 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
7047 (simplify_const_unary_operation, simplify_binary_operation_1)
7048 (simplify_const_binary_operation, simplify_ternary_operation)
7049 (test_vector_ops_duplicate, test_vector_ops): Likewise.
7050 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
7051 instead of CONST_VECTOR_NUNITS.
7052 * varasm.c (output_constant_pool_2): Likewise.
7053 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
7054 explicit-encoded elements in the XVEC for variable-length vectors.
7056 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7058 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
7060 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7061 Alan Hayward <alan.hayward@arm.com>
7062 David Sherwood <david.sherwood@arm.com>
7064 * coretypes.h (fixed_size_mode): Declare.
7065 (fixed_size_mode_pod): New typedef.
7066 * builtins.h (target_builtins::x_apply_args_mode)
7067 (target_builtins::x_apply_result_mode): Change type to
7068 fixed_size_mode_pod.
7069 * builtins.c (apply_args_size, apply_result_size, result_vector)
7070 (expand_builtin_apply_args_1, expand_builtin_apply)
7071 (expand_builtin_return): Update accordingly.
7073 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7075 * cse.c (hash_rtx_cb): Hash only the encoded elements.
7076 * cselib.c (cselib_hash_rtx): Likewise.
7077 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
7078 CONST_VECTOR encoding.
7080 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7081 Jeff Law <law@redhat.com>
7084 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
7085 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
7086 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
7087 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
7090 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
7091 explicitly probe *sp in a noreturn function if there were any callee
7092 register saves or frame pointer is needed.
7094 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7097 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
7098 BLKmode for ternary, binary or unary expressions.
7101 * var-tracking.c (delete_vta_debug_insn): New inline function.
7102 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
7103 insns from get_insns () to NULL instead of each bb separately.
7104 Use delete_vta_debug_insn. No longer static.
7105 (vt_debug_insns_local, variable_tracking_main_1): Adjust
7106 delete_vta_debug_insns callers.
7107 * rtl.h (delete_vta_debug_insns): Declare.
7108 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
7109 instead of variable_tracking_main.
7111 2018-01-03 Martin Sebor <msebor@redhat.com>
7113 PR tree-optimization/83603
7114 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
7115 arguments past the endof the argument list in functions declared
7116 without a prototype.
7117 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
7118 Avoid checking when arguments are null.
7120 2018-01-03 Martin Sebor <msebor@redhat.com>
7123 * doc/extend.texi (attribute const): Fix a typo.
7124 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
7125 issuing -Wsuggest-attribute for void functions.
7127 2018-01-03 Martin Sebor <msebor@redhat.com>
7129 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
7130 offset_int::from instead of wide_int::to_shwi.
7131 (maybe_diag_overlap): Remove assertion.
7132 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
7133 * gimple-ssa-sprintf.c (format_directive): Same.
7134 (parse_directive): Same.
7135 (sprintf_dom_walker::compute_format_length): Same.
7136 (try_substitute_return_value): Same.
7138 2018-01-03 Jeff Law <law@redhat.com>
7141 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
7142 non-constant residual for zero at runtime and avoid probing in
7143 that case. Reorganize code for trailing problem to mirror handling
7146 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
7148 PR tree-optimization/83501
7149 * tree-ssa-strlen.c (get_string_cst): New.
7150 (handle_char_store): Call get_string_cst.
7152 2018-01-03 Martin Liska <mliska@suse.cz>
7154 PR tree-optimization/83593
7155 * tree-ssa-strlen.c: Include tree-cfg.h.
7156 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
7157 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
7158 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
7160 (strlen_dom_walker::before_dom_children): Call
7161 gimple_purge_dead_eh_edges. Dump tranformation with details
7163 (strlen_dom_walker::before_dom_children): Update call by adding
7164 new argument cleanup_eh.
7165 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
7167 2018-01-03 Martin Liska <mliska@suse.cz>
7170 * cif-code.def (VARIADIC_THUNK): New enum value.
7171 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
7174 2018-01-03 Jan Beulich <jbeulich@suse.com>
7176 * sse.md (mov<mode>_internal): Tighten condition for when to use
7177 vmovdqu<ssescalarsize> for TI and OI modes.
7179 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7181 Update copyright years.
7183 2018-01-03 Martin Liska <mliska@suse.cz>
7186 * ipa-visibility.c (function_and_variable_visibility): Skip
7187 functions with noipa attribure.
7189 2018-01-03 Jakub Jelinek <jakub@redhat.com>
7191 * gcc.c (process_command): Update copyright notice dates.
7192 * gcov-dump.c (print_version): Ditto.
7193 * gcov.c (print_version): Ditto.
7194 * gcov-tool.c (print_version): Ditto.
7195 * gengtype.c (create_file): Ditto.
7196 * doc/cpp.texi: Bump @copying's copyright year.
7197 * doc/cppinternals.texi: Ditto.
7198 * doc/gcc.texi: Ditto.
7199 * doc/gccint.texi: Ditto.
7200 * doc/gcov.texi: Ditto.
7201 * doc/install.texi: Ditto.
7202 * doc/invoke.texi: Ditto.
7204 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7206 * vector-builder.h (vector_builder::m_full_nelts): Change from
7207 unsigned int to poly_uint64.
7208 (vector_builder::full_nelts): Update prototype accordingly.
7209 (vector_builder::new_vector): Likewise.
7210 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
7211 (vector_builder::operator ==): Likewise.
7212 (vector_builder::finalize): Likewise.
7213 * int-vector-builder.h (int_vector_builder::int_vector_builder):
7214 Take the number of elements as a poly_uint64 rather than an
7216 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
7217 from unsigned int to poly_uint64.
7218 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
7219 (vec_perm_indices::new_vector): Likewise.
7220 (vec_perm_indices::length): Likewise.
7221 (vec_perm_indices::nelts_per_input): Likewise.
7222 (vec_perm_indices::input_nelts): Likewise.
7223 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
7224 number of elements per input as a poly_uint64 rather than an
7225 unsigned int. Use the original encoding for variable-length
7226 vectors, rather than clamping each individual element.
7227 For the second and subsequent elements in each pattern,
7228 clamp the step and base before clamping their sum.
7229 (vec_perm_indices::series_p): Handle polynomial element counts.
7230 (vec_perm_indices::all_in_range_p): Likewise.
7231 (vec_perm_indices_to_tree): Likewise.
7232 (vec_perm_indices_to_rtx): Likewise.
7233 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
7234 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
7235 (tree_vector_builder::new_binary_operation): Handle polynomial
7236 element counts. Return false if we need to know the number
7237 of elements at compile time.
7238 * fold-const.c (fold_vec_perm): Punt if the number of elements
7239 isn't known at compile time.
7241 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7243 * vec-perm-indices.h (vec_perm_builder): Change element type
7244 from HOST_WIDE_INT to poly_int64.
7245 (vec_perm_indices::element_type): Update accordingly.
7246 (vec_perm_indices::clamp): Handle polynomial element_types.
7247 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7248 (vec_perm_indices::all_in_range_p): Likewise.
7249 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
7251 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
7252 polynomial vec_perm_indices element types.
7253 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
7254 * fold-const.c (fold_vec_perm): Likewise.
7255 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
7256 * tree-vect-generic.c (lower_vec_perm): Likewise.
7257 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7258 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
7259 element type to HOST_WIDE_INT.
7261 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7262 Alan Hayward <alan.hayward@arm.com>
7263 David Sherwood <david.sherwood@arm.com>
7265 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
7266 rather than an int. Use plus_constant.
7267 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
7268 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
7270 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7271 Alan Hayward <alan.hayward@arm.com>
7272 David Sherwood <david.sherwood@arm.com>
7274 * calls.c (emit_call_1, expand_call): Change struct_value_size from
7275 a HOST_WIDE_INT to a poly_int64.
7277 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7278 Alan Hayward <alan.hayward@arm.com>
7279 David Sherwood <david.sherwood@arm.com>
7281 * calls.c (load_register_parameters): Cope with polynomial
7282 mode sizes. Require a constant size for BLKmode parameters
7283 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
7284 forces a parameter to be padded at the lsb end in order to
7285 fill a complete number of words, require the parameter size
7286 to be ordered wrt UNITS_PER_WORD.
7288 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7289 Alan Hayward <alan.hayward@arm.com>
7290 David Sherwood <david.sherwood@arm.com>
7292 * reload1.c (spill_stack_slot_width): Change element type
7293 from unsigned int to poly_uint64_pod.
7294 (alter_reg): Treat mode sizes as polynomial.
7296 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7297 Alan Hayward <alan.hayward@arm.com>
7298 David Sherwood <david.sherwood@arm.com>
7300 * reload.c (complex_word_subreg_p): New function.
7301 (reload_inner_reg_of_subreg, push_reload): Use it.
7303 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7304 Alan Hayward <alan.hayward@arm.com>
7305 David Sherwood <david.sherwood@arm.com>
7307 * lra-constraints.c (process_alt_operands): Reject matched
7308 operands whose sizes aren't ordered.
7309 (match_reload): Refer to this check here.
7311 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7312 Alan Hayward <alan.hayward@arm.com>
7313 David Sherwood <david.sherwood@arm.com>
7315 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
7316 that the mode size is in the set {1, 2, 4, 8, 16}.
7318 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7319 Alan Hayward <alan.hayward@arm.com>
7320 David Sherwood <david.sherwood@arm.com>
7322 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
7323 Use plus_constant instead of gen_rtx_PLUS.
7325 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7326 Alan Hayward <alan.hayward@arm.com>
7327 David Sherwood <david.sherwood@arm.com>
7329 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
7330 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
7331 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
7332 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
7333 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
7334 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
7335 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
7336 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
7337 * config/i386/i386.c (ix86_push_rounding): ...this new function.
7338 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
7340 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
7341 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
7342 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
7343 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
7344 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
7345 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
7346 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
7347 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
7348 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
7349 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
7351 * expr.c (emit_move_resolve_push): Treat the input and result
7352 of PUSH_ROUNDING as a poly_int64.
7353 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
7354 (emit_push_insn): Likewise.
7355 * lra-eliminations.c (mark_not_eliminable): Likewise.
7356 * recog.c (push_operand): Likewise.
7357 * reload1.c (elimination_effects): Likewise.
7358 * rtlanal.c (nonzero_bits1): Likewise.
7359 * calls.c (store_one_arg): Likewise. Require the padding to be
7360 known at compile time.
7362 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7363 Alan Hayward <alan.hayward@arm.com>
7364 David Sherwood <david.sherwood@arm.com>
7366 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
7367 Use plus_constant instead of gen_rtx_PLUS.
7369 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7370 Alan Hayward <alan.hayward@arm.com>
7371 David Sherwood <david.sherwood@arm.com>
7373 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
7376 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7377 Alan Hayward <alan.hayward@arm.com>
7378 David Sherwood <david.sherwood@arm.com>
7380 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
7381 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
7382 via stack temporaries. Treat the mode size as polynomial too.
7384 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7385 Alan Hayward <alan.hayward@arm.com>
7386 David Sherwood <david.sherwood@arm.com>
7388 * expr.c (expand_expr_real_2): When handling conversions involving
7389 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
7390 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
7391 as a poly_uint64 too.
7393 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7394 Alan Hayward <alan.hayward@arm.com>
7395 David Sherwood <david.sherwood@arm.com>
7397 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
7399 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7400 Alan Hayward <alan.hayward@arm.com>
7401 David Sherwood <david.sherwood@arm.com>
7403 * combine.c (can_change_dest_mode): Handle polynomial
7404 REGMODE_NATURAL_SIZE.
7405 * expmed.c (store_bit_field_1): Likewise.
7406 * expr.c (store_constructor): Likewise.
7407 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
7408 and polynomial REGMODE_NATURAL_SIZE.
7409 (gen_lowpart_common): Likewise.
7410 * reginfo.c (record_subregs_of_mode): Likewise.
7411 * rtlanal.c (read_modify_subreg_p): Likewise.
7413 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7414 Alan Hayward <alan.hayward@arm.com>
7415 David Sherwood <david.sherwood@arm.com>
7417 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
7418 numbers of elements.
7420 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7421 Alan Hayward <alan.hayward@arm.com>
7422 David Sherwood <david.sherwood@arm.com>
7424 * match.pd: Cope with polynomial numbers of vector elements.
7426 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7427 Alan Hayward <alan.hayward@arm.com>
7428 David Sherwood <david.sherwood@arm.com>
7430 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
7431 in a POINTER_PLUS_EXPR.
7433 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7434 Alan Hayward <alan.hayward@arm.com>
7435 David Sherwood <david.sherwood@arm.com>
7437 * omp-simd-clone.c (simd_clone_subparts): New function.
7438 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
7439 (ipa_simd_modify_function_body): Likewise.
7441 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7442 Alan Hayward <alan.hayward@arm.com>
7443 David Sherwood <david.sherwood@arm.com>
7445 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
7446 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
7447 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
7448 (expand_vector_condition, vector_element): Likewise.
7449 (subparts_gt): New function.
7450 (get_compute_type): Use subparts_gt.
7451 (count_type_subparts): Delete.
7452 (expand_vector_operations_1): Use subparts_gt instead of
7453 count_type_subparts.
7455 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7456 Alan Hayward <alan.hayward@arm.com>
7457 David Sherwood <david.sherwood@arm.com>
7459 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
7460 (vect_compile_time_alias): ...this new function. Do the calculation
7461 on poly_ints rather than trees.
7462 (vect_prune_runtime_alias_test_list): Update call accordingly.
7464 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7465 Alan Hayward <alan.hayward@arm.com>
7466 David Sherwood <david.sherwood@arm.com>
7468 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
7470 (vect_schedule_slp_instance): Likewise.
7472 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7473 Alan Hayward <alan.hayward@arm.com>
7474 David Sherwood <david.sherwood@arm.com>
7476 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
7477 constant and extern definitions for variable-length vectors.
7478 (vect_get_constant_vectors): Note that the number of units
7479 is known to be constant.
7481 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7482 Alan Hayward <alan.hayward@arm.com>
7483 David Sherwood <david.sherwood@arm.com>
7485 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
7486 of units as polynomial. Choose between WIDE and NARROW based
7489 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7490 Alan Hayward <alan.hayward@arm.com>
7491 David Sherwood <david.sherwood@arm.com>
7493 * tree-vect-stmts.c (simd_clone_subparts): New function.
7494 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
7496 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7497 Alan Hayward <alan.hayward@arm.com>
7498 David Sherwood <david.sherwood@arm.com>
7500 * tree-vect-stmts.c (vectorizable_call): Treat the number of
7501 vectors as polynomial. Use build_index_vector for
7504 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7505 Alan Hayward <alan.hayward@arm.com>
7506 David Sherwood <david.sherwood@arm.com>
7508 * tree-vect-stmts.c (get_load_store_type): Treat the number of
7509 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
7510 for variable-length vectors.
7511 (vectorizable_mask_load_store): Treat the number of units as
7512 polynomial, asserting that it is constant if the condition has
7513 already been enforced.
7514 (vectorizable_store, vectorizable_load): Likewise.
7516 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7517 Alan Hayward <alan.hayward@arm.com>
7518 David Sherwood <david.sherwood@arm.com>
7520 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
7521 of units as polynomial. Punt if we can't tell at compile time
7522 which vector contains the final result.
7524 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7525 Alan Hayward <alan.hayward@arm.com>
7526 David Sherwood <david.sherwood@arm.com>
7528 * tree-vect-loop.c (vectorizable_induction): Treat the number
7529 of units as polynomial. Punt on SLP inductions. Use an integer
7530 VEC_SERIES_EXPR for variable-length integer reductions. Use a
7531 cast of such a series for variable-length floating-point
7534 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7535 Alan Hayward <alan.hayward@arm.com>
7536 David Sherwood <david.sherwood@arm.com>
7538 * tree.h (build_index_vector): Declare.
7539 * tree.c (build_index_vector): New function.
7540 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
7541 of units as polynomial, forcibly converting it to a constant if
7542 vectorizable_reduction has already enforced the condition.
7543 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
7544 to create a {1,2,3,...} vector.
7545 (vectorizable_reduction): Treat the number of units as polynomial.
7546 Choose vectype_in based on the largest scalar element size rather
7547 than the smallest number of units. Enforce the restrictions
7550 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7551 Alan Hayward <alan.hayward@arm.com>
7552 David Sherwood <david.sherwood@arm.com>
7554 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
7555 number of units as polynomial.
7557 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7558 Alan Hayward <alan.hayward@arm.com>
7559 David Sherwood <david.sherwood@arm.com>
7561 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
7562 * target.def (autovectorize_vector_sizes): Return the vector sizes
7563 by pointer, using vector_sizes rather than a bitmask.
7564 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
7565 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
7566 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
7568 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
7569 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
7570 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
7571 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
7572 * omp-general.c (omp_max_vf): Likewise.
7573 * omp-low.c (omp_clause_aligned_alignment): Likewise.
7574 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
7575 * tree-vect-loop.c (vect_analyze_loop): Likewise.
7576 * tree-vect-slp.c (vect_slp_bb): Likewise.
7577 * doc/tm.texi: Regenerate.
7578 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
7580 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
7581 the vector size as a poly_uint64 rather than an unsigned int.
7582 (current_vector_size): Change from an unsigned int to a poly_uint64.
7583 (get_vectype_for_scalar_type): Update accordingly.
7584 * tree.h (build_truth_vector_type): Take the size and number of
7585 units as a poly_uint64 rather than an unsigned int.
7586 (build_vector_type): Add a temporary overload that takes
7587 the number of units as a poly_uint64 rather than an unsigned int.
7588 * tree.c (make_vector_type): Likewise.
7589 (build_truth_vector_type): Take the number of units as a poly_uint64
7590 rather than an unsigned int.
7592 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7593 Alan Hayward <alan.hayward@arm.com>
7594 David Sherwood <david.sherwood@arm.com>
7596 * target.def (get_mask_mode): Take the number of units and length
7597 as poly_uint64s rather than unsigned ints.
7598 * targhooks.h (default_get_mask_mode): Update accordingly.
7599 * targhooks.c (default_get_mask_mode): Likewise.
7600 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
7601 * doc/tm.texi: Regenerate.
7603 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7604 Alan Hayward <alan.hayward@arm.com>
7605 David Sherwood <david.sherwood@arm.com>
7607 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
7608 * omp-general.c (omp_max_vf): Likewise.
7609 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
7610 (expand_omp_simd): Handle polynomial safelen.
7611 * omp-low.c (omplow_simd_context): Add a default constructor.
7612 (omplow_simd_context::max_vf): Change from int to poly_uint64.
7613 (lower_rec_simd_input_clauses): Update accordingly.
7614 (lower_rec_input_clauses): Likewise.
7616 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7617 Alan Hayward <alan.hayward@arm.com>
7618 David Sherwood <david.sherwood@arm.com>
7620 * tree-vectorizer.h (vect_nunits_for_cost): New function.
7621 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
7622 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
7623 (vect_analyze_slp_cost): Likewise.
7624 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
7625 (vect_model_load_cost): Likewise.
7627 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7628 Alan Hayward <alan.hayward@arm.com>
7629 David Sherwood <david.sherwood@arm.com>
7631 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
7632 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
7633 from an unsigned int * to a poly_uint64_pod *.
7634 (calculate_unrolling_factor): New function.
7635 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
7637 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7638 Alan Hayward <alan.hayward@arm.com>
7639 David Sherwood <david.sherwood@arm.com>
7641 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
7642 from an unsigned int to a poly_uint64.
7643 (_loop_vec_info::slp_unrolling_factor): Likewise.
7644 (_loop_vec_info::vectorization_factor): Change from an int
7646 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
7647 (vect_get_num_vectors): New function.
7648 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
7649 (vect_get_num_copies): Use vect_get_num_vectors.
7650 (vect_analyze_data_ref_dependences): Change max_vf from an int *
7651 to an unsigned int *.
7652 (vect_analyze_data_refs): Change min_vf from an int * to a
7654 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7655 than an unsigned HOST_WIDE_INT.
7656 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
7657 (vect_analyze_data_ref_dependence): Change max_vf from an int *
7658 to an unsigned int *.
7659 (vect_analyze_data_ref_dependences): Likewise.
7660 (vect_compute_data_ref_alignment): Handle polynomial vf.
7661 (vect_enhance_data_refs_alignment): Likewise.
7662 (vect_prune_runtime_alias_test_list): Likewise.
7663 (vect_shift_permute_load_chain): Likewise.
7664 (vect_supportable_dr_alignment): Likewise.
7665 (dependence_distance_ge_vf): Take the vectorization factor as a
7666 poly_uint64 rather than an unsigned HOST_WIDE_INT.
7667 (vect_analyze_data_refs): Change min_vf from an int * to a
7669 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
7670 vfm1 as a poly_uint64 rather than an int. Make the same change
7671 for the returned bound_scalar.
7672 (vect_gen_vector_loop_niters): Handle polynomial vf.
7673 (vect_do_peeling): Likewise. Update call to
7674 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
7675 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
7677 * tree-vect-loop.c (vect_determine_vectorization_factor)
7678 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
7679 (vect_get_known_peeling_cost): Likewise.
7680 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
7681 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
7682 (vect_transform_loop): Likewise. Use the lowest possible VF when
7683 updating the upper bounds of the loop.
7684 (vect_min_worthwhile_factor): Make static. Return an unsigned int
7686 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
7687 polynomial unroll factors.
7688 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
7689 (vect_make_slp_decision): Likewise.
7690 (vect_supported_load_permutation_p): Likewise, and polynomial
7692 (vect_analyze_slp_cost): Handle polynomial vf.
7693 (vect_slp_analyze_node_operations): Likewise.
7694 (vect_slp_analyze_bb_1): Likewise.
7695 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
7696 than an unsigned HOST_WIDE_INT.
7697 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
7698 (vectorizable_load): Handle polynomial vf.
7699 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
7701 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
7703 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7704 Alan Hayward <alan.hayward@arm.com>
7705 David Sherwood <david.sherwood@arm.com>
7707 * match.pd: Handle bit operations involving three constants
7708 and try to fold one pair.
7710 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
7712 * tree-vect-loop-manip.c: Include gimple-fold.h.
7713 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
7714 niters_maybe_zero parameters. Handle other cases besides a step of 1.
7715 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
7716 Add a path that uses a step of VF instead of 1, but disable it
7718 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
7719 and niters_no_overflow parameters. Update calls to
7720 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
7721 Create a new SSA name if the latter choses to use a ste other
7722 than zero, and return it via niters_vector_mult_vf_var.
7723 * tree-vect-loop.c (vect_transform_loop): Update calls to
7724 vect_do_peeling, vect_gen_vector_loop_niters and
7725 slpeel_make_loop_iterate_ntimes.
7726 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
7727 (vect_gen_vector_loop_niters): Update declarations after above changes.
7729 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
7731 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
7732 128-bit round to integer instructions.
7733 (ceil<mode>2): Likewise.
7734 (btrunc<mode>2): Likewise.
7735 (round<mode>2): Likewise.
7737 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7739 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
7740 unaligned VSX load/store on P8/P9.
7741 (expand_block_clear): Allow the use of unaligned VSX
7742 load/store on P8/P9.
7744 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7746 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
7748 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
7749 swap associated with both a load and a store.
7751 2018-01-02 Andrew Waterman <andrew@sifive.com>
7753 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
7754 * config/riscv/riscv.md (clear_cache): Use it.
7756 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
7758 * web.c: Remove out-of-date comment.
7760 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7762 * expr.c (fixup_args_size_notes): Check that any existing
7763 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
7764 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
7765 (emit_single_push_insn): ...here.
7767 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7769 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
7770 (const_vector_encoded_nelts): New function.
7771 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
7772 (const_vector_int_elt, const_vector_elt): Declare.
7773 * emit-rtl.c (const_vector_int_elt_1): New function.
7774 (const_vector_elt): Likewise.
7775 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
7776 of CONST_VECTOR_ELT.
7778 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7780 * expr.c: Include rtx-vector-builder.h.
7781 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
7782 directly on the tree encoding.
7783 (const_vector_from_tree): Likewise.
7784 * optabs.c: Include rtx-vector-builder.h.
7785 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
7786 sequence of "u" values.
7787 * vec-perm-indices.c: Include rtx-vector-builder.h.
7788 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
7789 directly on the vec_perm_indices encoding.
7791 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7793 * doc/rtl.texi (const_vector): Describe new encoding scheme.
7794 * Makefile.in (OBJS): Add rtx-vector-builder.o.
7795 * rtx-vector-builder.h: New file.
7796 * rtx-vector-builder.c: Likewise.
7797 * rtl.h (rtx_def::u2): Add a const_vector field.
7798 (CONST_VECTOR_NPATTERNS): New macro.
7799 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
7800 (CONST_VECTOR_DUPLICATE_P): Likewise.
7801 (CONST_VECTOR_STEPPED_P): Likewise.
7802 (CONST_VECTOR_ENCODED_ELT): Likewise.
7803 (const_vec_duplicate_p): Check for a duplicated vector encoding.
7804 (unwrap_const_vec_duplicate): Likewise.
7805 (const_vec_series_p): Check for a non-duplicated vector encoding.
7806 Say that the function only returns true for integer vectors.
7807 * emit-rtl.c: Include rtx-vector-builder.h.
7808 (gen_const_vec_duplicate_1): Delete.
7809 (gen_const_vector): Call gen_const_vec_duplicate instead of
7810 gen_const_vec_duplicate_1.
7811 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
7812 (gen_const_vec_duplicate): Use rtx_vector_builder.
7813 (gen_const_vec_series): Likewise.
7814 (gen_rtx_CONST_VECTOR): Likewise.
7815 * config/powerpcspe/powerpcspe.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.
7819 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
7820 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
7821 Build a new vector rather than modifying a CONST_VECTOR in-place.
7822 (handle_special_swappables): Update call accordingly.
7824 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7826 * simplify-rtx.c (simplify_const_binary_operation): Use
7827 CONST_VECTOR_ELT instead of XVECEXP.
7829 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7831 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
7832 the selector elements to be different from the data elements
7833 if the selector is a VECTOR_CST.
7834 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
7835 ssizetype for the selector.
7837 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7839 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
7840 before testing each element individually.
7841 * tree-vect-generic.c (lower_vec_perm): Likewise.
7843 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7845 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
7846 * selftest-run-tests.c (selftest::run_tests): Call it.
7847 * vector-builder.h (vector_builder::operator ==): New function.
7848 (vector_builder::operator !=): Likewise.
7849 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
7850 (vec_perm_indices::all_from_input_p): New function.
7851 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
7852 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
7853 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
7854 instead of reading the VECTOR_CST directly. Detect whether both
7855 vector inputs are the same before constructing the vec_perm_indices,
7856 and update the number of inputs argument accordingly. Use the
7857 utility functions added above. Only construct sel2 if we need to.
7859 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7861 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
7862 the broadcast of the low byte.
7863 (expand_mult_highpart): Use an explicit encoding for the permutes.
7864 * optabs-query.c (can_mult_highpart_p): Likewise.
7865 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
7866 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7867 (vectorizable_bswap): Likewise.
7868 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
7869 explicit encoding for the power-of-2 permutes.
7870 (vect_permute_store_chain): Likewise.
7871 (vect_grouped_load_supported): Likewise.
7872 (vect_permute_load_chain): Likewise.
7874 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7876 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
7877 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
7878 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
7879 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
7880 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
7881 (vect_gen_perm_mask_any): Likewise.
7883 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7885 * int-vector-builder.h: New file.
7886 * vec-perm-indices.h: Include int-vector-builder.h.
7887 (vec_perm_indices): Redefine as an int_vector_builder.
7888 (auto_vec_perm_indices): Delete.
7889 (vec_perm_builder): Redefine as a stand-alone class.
7890 (vec_perm_indices::vec_perm_indices): New function.
7891 (vec_perm_indices::clamp): Likewise.
7892 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
7893 (vec_perm_indices::new_vector): New function.
7894 (vec_perm_indices::new_expanded_vector): Update for new
7895 vec_perm_indices class.
7896 (vec_perm_indices::rotate_inputs): New function.
7897 (vec_perm_indices::all_in_range_p): Operate directly on the
7898 encoded form, without computing elided elements.
7899 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
7900 encoding. Update for new vec_perm_indices class.
7901 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
7902 the given vec_perm_builder.
7903 (expand_vec_perm_var): Update vec_perm_builder constructor.
7904 (expand_mult_highpart): Use vec_perm_builder instead of
7905 auto_vec_perm_indices.
7906 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
7907 vec_perm_indices instead of auto_vec_perm_indices. Use a single
7908 or double series encoding as appropriate.
7909 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
7910 vec_perm_indices instead of auto_vec_perm_indices.
7911 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
7912 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
7913 (vect_permute_store_chain): Likewise.
7914 (vect_grouped_load_supported): Likewise.
7915 (vect_permute_load_chain): Likewise.
7916 (vect_shift_permute_load_chain): Likewise.
7917 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
7918 (vect_transform_slp_perm_load): Likewise.
7919 (vect_schedule_slp_instance): Likewise.
7920 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
7921 (vectorizable_mask_load_store): Likewise.
7922 (vectorizable_bswap): Likewise.
7923 (vectorizable_store): Likewise.
7924 (vectorizable_load): Likewise.
7925 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
7926 vec_perm_indices instead of auto_vec_perm_indices. Use
7927 tree_to_vec_perm_builder to read the vector from a tree.
7928 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
7929 vec_perm_builder instead of a vec_perm_indices.
7930 (have_whole_vector_shift): Use vec_perm_builder and
7931 vec_perm_indices instead of auto_vec_perm_indices. Leave the
7932 truncation to calc_vec_perm_mask_for_shift.
7933 (vect_create_epilog_for_reduction): Likewise.
7934 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
7935 from auto_vec_perm_indices to vec_perm_indices.
7936 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
7937 instead of changing individual elements.
7938 (aarch64_vectorize_vec_perm_const): Use new_vector to install
7939 the vector in d.perm.
7940 * config/arm/arm.c (expand_vec_perm_d::perm): Change
7941 from auto_vec_perm_indices to vec_perm_indices.
7942 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
7943 instead of changing individual elements.
7944 (arm_vectorize_vec_perm_const): Use new_vector to install
7945 the vector in d.perm.
7946 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
7947 Update vec_perm_builder constructor.
7948 (rs6000_expand_interleave): Likewise.
7949 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
7950 (rs6000_expand_interleave): Likewise.
7952 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7954 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
7955 to qimode could truncate the indices.
7956 * optabs.c (expand_vec_perm_var): Likewise.
7958 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
7960 * Makefile.in (OBJS): Add vec-perm-indices.o.
7961 * vec-perm-indices.h: New file.
7962 * vec-perm-indices.c: Likewise.
7963 * target.h (vec_perm_indices): Replace with a forward class
7965 (auto_vec_perm_indices): Move to vec-perm-indices.h.
7966 * optabs.h: Include vec-perm-indices.h.
7967 (expand_vec_perm): Delete.
7968 (selector_fits_mode_p, expand_vec_perm_var): Declare.
7969 (expand_vec_perm_const): Declare.
7970 * target.def (vec_perm_const_ok): Replace with...
7971 (vec_perm_const): ...this new hook.
7972 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
7973 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
7974 * doc/tm.texi: Regenerate.
7975 * optabs.def (vec_perm_const): Delete.
7976 * doc/md.texi (vec_perm_const): Likewise.
7977 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
7978 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
7979 expand_vec_perm for constant permutation vectors. Assert that
7980 the mode of variable permutation vectors is the integer equivalent
7981 of the mode that is being permuted.
7982 * optabs-query.h (selector_fits_mode_p): Declare.
7983 * optabs-query.c: Include vec-perm-indices.h.
7984 (selector_fits_mode_p): New function.
7985 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
7986 is defined, instead of checking whether the vec_perm_const_optab
7987 exists. Use targetm.vectorize.vec_perm_const instead of
7988 targetm.vectorize.vec_perm_const_ok. Check whether the indices
7989 fit in the vector mode before using a variable permute.
7990 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
7991 vec_perm_indices instead of an rtx.
7992 (expand_vec_perm): Replace with...
7993 (expand_vec_perm_const): ...this new function. Take the selector
7994 as a vec_perm_indices rather than an rtx. Also take the mode of
7995 the selector. Update call to shift_amt_for_vec_perm_mask.
7996 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
7997 Use vec_perm_indices::new_expanded_vector to expand the original
7998 selector into bytes. Check whether the indices fit in the vector
7999 mode before using a variable permute.
8000 (expand_vec_perm_var): Make global.
8001 (expand_mult_highpart): Use expand_vec_perm_const.
8002 * fold-const.c: Includes vec-perm-indices.h.
8003 * tree-ssa-forwprop.c: Likewise.
8004 * tree-vect-data-refs.c: Likewise.
8005 * tree-vect-generic.c: Likewise.
8006 * tree-vect-loop.c: Likewise.
8007 * tree-vect-slp.c: Likewise.
8008 * tree-vect-stmts.c: Likewise.
8009 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
8011 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
8012 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
8013 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
8014 (aarch64_vectorize_vec_perm_const): ...this new function.
8015 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8016 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8017 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
8018 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
8019 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8020 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8021 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
8023 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
8024 check for NEON modes.
8025 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
8026 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
8027 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
8028 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
8030 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
8031 the old VEC_PERM_CONST conditions.
8032 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
8033 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
8034 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
8035 (ia64_vectorize_vec_perm_const_ok): Merge into...
8036 (ia64_vectorize_vec_perm_const): ...this new function.
8037 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
8038 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
8039 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
8040 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
8041 * config/mips/mips.c (mips_expand_vec_perm_const)
8042 (mips_vectorize_vec_perm_const_ok): Merge into...
8043 (mips_vectorize_vec_perm_const): ...this new function.
8044 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
8045 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
8046 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
8047 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
8048 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
8049 (rs6000_expand_vec_perm_const): Delete.
8050 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
8052 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8053 (altivec_expand_vec_perm_const_le): Take each operand individually.
8054 Operate on constant selectors rather than rtxes.
8055 (altivec_expand_vec_perm_const): Likewise. Update call to
8056 altivec_expand_vec_perm_const_le.
8057 (rs6000_expand_vec_perm_const): Delete.
8058 (rs6000_vectorize_vec_perm_const_ok): Delete.
8059 (rs6000_vectorize_vec_perm_const): New function.
8060 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8061 an element count and rtx array.
8062 (rs6000_expand_extract_even): Update call accordingly.
8063 (rs6000_expand_interleave): Likewise.
8064 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
8065 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
8066 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
8067 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
8068 (rs6000_expand_vec_perm_const): Delete.
8069 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
8070 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8071 (altivec_expand_vec_perm_const_le): Take each operand individually.
8072 Operate on constant selectors rather than rtxes.
8073 (altivec_expand_vec_perm_const): Likewise. Update call to
8074 altivec_expand_vec_perm_const_le.
8075 (rs6000_expand_vec_perm_const): Delete.
8076 (rs6000_vectorize_vec_perm_const_ok): Delete.
8077 (rs6000_vectorize_vec_perm_const): New function. Remove stray
8078 reference to the SPE evmerge intructions.
8079 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
8080 an element count and rtx array.
8081 (rs6000_expand_extract_even): Update call accordingly.
8082 (rs6000_expand_interleave): Likewise.
8083 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
8084 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
8086 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
8088 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8090 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
8091 vector mode and that that mode matches the mode of the data
8093 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
8094 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
8095 directly using expand_vec_perm_1 when forcing selectors into
8097 (expand_vec_perm_var): New function, split out from expand_vec_perm.
8099 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8101 * optabs-query.h (can_vec_perm_p): Delete.
8102 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
8103 * optabs-query.c (can_vec_perm_p): Split into...
8104 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
8105 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
8106 particular selector is valid.
8107 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
8108 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
8109 (vect_grouped_load_supported): Likewise.
8110 (vect_shift_permute_load_chain): Likewise.
8111 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
8112 (vect_transform_slp_perm_load): Likewise.
8113 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
8114 (vectorizable_bswap): Likewise.
8115 (vect_gen_perm_mask_checked): Likewise.
8116 * fold-const.c (fold_ternary_loc): Likewise. Don't take
8117 implementations of variable permutation vectors into account
8118 when deciding which selector to use.
8119 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
8120 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
8121 with a false third argument.
8122 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
8123 to test whether the constant selector is valid and can_vec_perm_var_p
8124 to test whether a variable selector is valid.
8126 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8128 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
8129 * optabs-query.c (can_vec_perm_p): Likewise.
8130 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
8131 instead of vec_perm_indices.
8132 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
8133 (vect_gen_perm_mask_checked): Likewise,
8134 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
8135 (vect_gen_perm_mask_checked): Likewise,
8137 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
8139 * optabs-query.h (qimode_for_vec_perm): Declare.
8140 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
8141 (qimode_for_vec_perm): ...this new function.
8142 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
8144 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8146 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
8147 does not have a conditional at the top.
8149 2018-01-02 Richard Biener <rguenther@suse.de>
8151 * ipa-inline.c (big_speedup_p): Fix expression.
8153 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8156 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
8159 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
8163 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
8164 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
8165 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
8166 cond_taken_branch_cost 3->4.
8168 2018-01-01 Jakub Jelinek <jakub@redhat.com>
8170 PR tree-optimization/83581
8171 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
8172 TODO_cleanup_cfg if any changes have been made.
8175 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
8176 convert_modes if target mode has the right side, but different mode
8180 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
8181 last argument when extracting from CONCAT. If either from_real or
8182 from_imag is NULL, use expansion through memory. If result is not
8183 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
8184 the parts directly to inner mode, if even that fails, use expansion
8188 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
8189 check for bswap in mode rather than HImode and use that in expand_unop
8192 Copyright (C) 2018 Free Software Foundation, Inc.
8194 Copying and distribution of this file, with or without modification,
8195 are permitted in any medium without royalty provided the copyright
8196 notice and this notice are preserved.