1 2018-06-15 Jakub Jelinek <jakub@redhat.com>
4 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
5 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
6 Only call store_expr for halves if the mode is the same.
9 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
10 Fix up comment formatting.
12 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
14 * typed-splay-tree.h (typed_splay_tree::remove): New function.
15 (typed_splay_tree::closure,
16 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
17 (typed_splay_tree::typed_splay_tree,
18 typed_splay_tree::operator =): Declared private.
19 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
20 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
21 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
22 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
23 typed_splay_tree::splay_tree_splay,
24 typed_splay_tree::splay_tree_foreach_helper,
25 typed_splay_tree::splay_tree_insert,
26 typed_splay_tree::splay_tree_remove,
27 typed_splay_tree::splay_tree_lookup,
28 typed_splay_tree::splay_tree_predecessor,
29 typed_splay_tree::splay_tree_successor,
30 typed_splay_tree::splay_tree_min,
31 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
32 (typed_splay_tree::root, typed_splay_tree::comp,
33 typed_splay_tree::delete_key,
34 typed_splay_tree::delete_value): New data members.
35 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
36 typed_splay_tree::remove.
38 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
40 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
41 -mginv and -mno-ginv to the assembler.
42 * config/mips/mips.opt (-mcrc): New option.
44 * doc/invoke.text (-mcrc): Document.
47 2018-06-15 Nick Clifton <nickc@redhat.com>
50 * tree.c (escaped_string): New class. Converts an unescaped
51 string into its escaped equivalent.
52 (warn_deprecated_use): Use the new class to convert the
53 deprecation message, if present.
54 (test_escaped_strings): New self test.
55 (test_c_tests): Add test_escaped_strings.
56 * doc/extend.texi (deprecated): Add a note that the
57 deprecation message is affected by the -fmessage-length
58 option, and that control characters will be escaped.
59 (#pragma GCC error): Document this pragma.
60 (#pragma GCC warning): Likewise.
61 * doc/invoke.texi (-fmessage-length): Document this option's
62 effect on the #warning and #error preprocessor directives and
63 the deprecated attribute.
65 2018-06-15 Richard Biener <rguenther@suse.de>
67 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
68 here, also noting vector size used.
69 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
70 size used in MSG_OPTIMIZED_LOCATIONS dump.
71 (pass_slp_vectorize::execute): Adjust.
73 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
76 * config/arc/arc.c (arc_return_address_register): Fix
79 2018-06-15 Richard Biener <rguenther@suse.de>
82 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
83 leave useless conversion stripping to force_gimple_operand_gsi.
84 (gimplify_build2): Likewise.
85 (gimplify_build1): Likewise.
87 2018-06-15 Richard Biener <rguenther@suse.de>
90 * tree-cfg.c (move_stmt_op): unshare invariant addresses
91 before adjusting their block.
93 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
95 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
96 multilibs for *-*-rtems*.
97 * config/riscv/t-rtems: New file.
99 2018-06-14 Jakub Jelinek <jakub@redhat.com>
102 * match.pd ((A +- CST1) +- CST2): Punt if last resort
103 unsigned_type_for returns NULL.
106 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
107 subregs of multi-word pseudos unless the float mode has word size.
109 2018-06-14 Richard Biener <rguenther@suse.de>
112 * tree-vect-generic.c (build_word_mode_vector_type): Remove
113 duplicate and harmful type_hash_canon.
114 * tree.c (type_hash_canon): Assert we didn't find ourselves.
116 2018-06-14 Richard Biener <rguenther@suse.de>
119 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
122 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
124 * config/rtems.h (STDINT_LONG32): Define.
126 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
127 Prachi Godbole <prachi.godbole@imgtec.com>
129 * config/mips/mips-cpus.def: Define P6600.
130 * config/mips/mips-tables.opt: Regenerate.
131 * config/mips/mips.c (mips_ucbranch_type): New enum.
132 (mips_rtx_cost_data): Add support for P6600.
133 (mips_issue_rate): Likewise.
134 (mips_multipass_dfa_lookahead): Likewise.
135 (mips_avoid_hazard): Likewise.
136 (mips_reorg_process_insns): Likewise.
137 (mips_classify_branch_p6600): New function.
138 * config/mips/mips.h (TUNE_P6600): New define.
139 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
140 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
141 * config/mips/mips.md: Include p6600.md.
142 (processor): Add p6600.
143 * config/mips/p6600.md: New file.
144 * doc/invoke.texi: Add p6600 to supported architectures.
146 2018-06-13 Martin Sebor <msebor@redhat.com>
148 PR tree-optimization/86114
149 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
151 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
153 2018-06-13 Richard Biener <rguenther@suse.de>
155 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
156 Properly set vector type of the intermediate stmt.
157 * tree-vect-stmts.c (vectorizable_operation): The destination
158 var always has vectype_out type.
160 2018-06-13 Jeff Law <law@redhat.com>
162 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
163 integer 0 for argument to print_rtl_with_bb.
164 (rl78_reorg): Likewise.
166 2018-06-13 David Malcolm <dmalcolm@redhat.com>
168 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
169 from rtx to rtx_insn *.
170 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
172 (add_sched_insns_for_speculation): Likewise for local "target",
173 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
174 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
175 from rtx_insn ** to rtx_code_label **.
176 (reorg_emit_nops): Likewise.
177 (c6x_reorg): Likewise for local "call_labels".
178 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
180 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
181 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
182 the loops over LABEL_REFS.
183 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
185 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
186 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
187 (split_branches): Strengthen local "olabel" from rtx to
188 rtx_insn *, adding a safe_as_a cast.
189 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
191 (add_insn_after): Likewise for first two params.
192 (add_insn_before): Likewise.
193 (remove_insn): Likewise for param.
194 (emit_pattern_before_noloc): Likewise for second and third params.
195 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
196 (emit_call_insn_before_noloc): Likewise.
197 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
199 (emit_barrier_before): Likewise.
200 (emit_label_before): Strengthen "label" param from "rtx" to
201 "rtx_code_label *". Strengthen "before" param from "rtx" to
203 (emit_insn_after_1): Strengthen "after" param from "rtx" to
205 (emit_pattern_after_noloc): Likewise.
206 (emit_insn_after_noloc): Likewise.
207 (emit_jump_insn_after_noloc): Likewise.
208 (emit_call_insn_after_noloc): Likewise.
209 (emit_debug_insn_after_noloc): Likewise.
210 (emit_barrier_after): Likewise.
211 (emit_label_after): Likewise for both params.
212 (emit_pattern_after_setloc): Likewise for "after" param. Convert
213 "loc" param from "int" to "location_t".
214 (emit_insn_after_setloc): Likewise.
215 (emit_jump_insn_after_setloc): Likewise.
216 (emit_call_insn_after_setloc): Likewise.
217 (emit_debug_insn_after_setloc): Likewise.
218 (emit_pattern_before_setloc): Likewise for "before" param. Convert
219 "loc" param from "int" to "location_t".
220 (emit_pattern_before): Convert NULL_RTX to NULL.
221 (emit_insn_before_setloc): Convert "loc" param from "int" to
223 (emit_jump_insn_before_setloc): Likewise.
224 (emit_call_insn_before_setloc): Likewise.
225 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
226 rtx_insn *. Convert "loc" param from "int" to "location_t".
227 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
228 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
229 Convert 3rd param from "int" to "location_t".
230 (emit_barrier_before, emit_barrier_after, next_real_insn):
231 Strengthen param from rtx to rtx_insn *.
232 (emit_label_before): Strengthen 1st param from "rtx" to
233 "rtx_code_label *". Strengthen 2nd param from "rtx" to
235 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
236 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
237 Strengthen 2nd param from "rtx" to "rtx_insn *".
238 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
239 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
240 Likewise. Convert 3rd param from "int" to "location_t".
241 (emit_label_after): Strengthen 1st param from "rtx" to
243 (next_real_insn, remove_insn): Strengthen param from "rtx" to
245 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
246 from "rtx" to "rtx_insn *".
248 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
250 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
251 bodies streamed in with -Q.
252 * dumpfile.c (dump_files): Add lto-stream-out dump file.
253 * dumpfile.h (tree_dump_index): Add lto_stream_out.
254 * gimple-streamer-out.c: Include gimple-pretty-print.h
255 (output_bb): Dump stmts streamed.
256 * lto-section-out.c: Include print-tree.h
257 (lto_begin_section): Dump sections created.
258 (lto_output_decl_index): Dump decl encoded.
259 * lto-streamer-out.c: Include print-tree.h
260 (create_output_block): Dump output block created.
261 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
262 (output_function): Dump function output.
263 (output_constructor): Dump constructor streamed.
264 (write_global_stream): Output indexes encoded.
265 (produce_asm_for_decls): Dump streams encoded.
266 * lto-streamer.c (streamer_dump_file): New global var.
267 * lto-streamer.h (streamer_dump_file): Declare.
268 * passes.c (ipa_write_summaries): Initialize streamer dump.
269 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
272 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
275 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
276 offsets for register save directives. Emit a second batch of save
277 directives, if need be, when the function accesses prior frames.
279 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
281 * config/arc/fpu.md (fmasf4): Force operand to register.
284 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
286 * config/arc/arc-protos.h (arc_pad_return): Remove.
287 * config/arc/arc.c (machine_function): Remove force_short_suffix
289 (arc_print_operand): Adjust printing of '&'.
290 (arc_verify_short): Remove conditional printing of short suffix.
291 (arc_final_prescan_insn): Remove reference to size_reason.
292 (pad_return): New function.
293 (arc_reorg): Call pad_return.
294 (arc_pad_return): Remove.
295 (arc_init_machine_status): Remove reference to force_short_suffix.
296 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
297 (attr length): When attribute iscompact is true force to 2
298 regardless; in the case of maybe check if we want to force the
299 instruction to have 4 bytes length.
300 (nopv): Change it to generate 4 byte long nop as well.
301 (blockage): New pattern.
302 (simple_return): Remove call to arc_pad_return.
303 (p_return_i): Likewise.
305 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
307 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
309 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
311 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
314 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
316 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
317 for ARC700 and ARCv2.
319 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
322 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
323 operands[2] instead of operands[1].
326 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
328 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
329 case, check whether the outer register overlaps an unallocatable
330 register, not just whether it fits the required class.
332 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
334 * poly-int.h (can_div_trunc_p): Add new overload in which all values
336 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
337 (memrefs_conflict_p): Likewise.
338 (init_alias_analysis): Likewise.
339 * cfgexpand.c (expand_debug_expr): Likewise.
340 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
341 * cse.c (fold_rtx): Likewise.
342 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
343 * expr.c (emit_block_move_hints): Likewise.
344 (clear_storage_hints, push_block, emit_push_insn): Likewise.
345 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
346 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
347 (emit_group_store): Likewise.
348 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
349 to read the PRE/POST_MODIFY increment.
350 * calls.c (store_one_arg): Use strip_offset.
351 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
353 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
355 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
356 (simplify_binary_operation_1): Extend CONST_INT handling to
358 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
359 than a HOST_WIDE_INT.
360 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
362 (adjust_mems, add_stores): Update accodingly.
363 (vt_canonicalize_addr): Track polynomial offsets.
364 (emit_note_insn_var_location): Likewise.
365 (vt_add_function_parameter): Likewise.
366 (vt_initialize): Likewise.
368 2018-06-12 Jeff Law <law@redhat.com>
370 * config.gcc (alpha*-*-freebsd*): Remove.
371 * config/alpha/freebsd.h: Remove.
373 2018-06-12 David Malcolm <dmalcolm@redhat.com>
376 * spellcheck-tree.c (levenshtein_distance): Rename to...
377 (get_edit_distance): ...this, and update for underlying renaming.
378 * spellcheck-tree.h (levenshtein_distance): Rename to...
379 (get_edit_distance): ...this.
380 * spellcheck.c (levenshtein_distance): Rename to...
381 (get_edit_distance): ...this. Convert from Levenshtein distance
382 to Damerau-Levenshtein distance by supporting transpositions of
383 adjacent characters. Rename "v1" to "v_next" and "v0" to
385 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
386 (selftest::test_edit_distance_unit_test_oneway): ...this, and
387 update for underlying renaming.
388 (selftest::levenshtein_distance_unit_test): Rename to...
389 (selftest::test_get_edit_distance_unit): ...this, and update for
391 (selftest::test_find_closest_string): Add example from PR 69968
392 where transposition helps
393 (selftest::test_metric_conditions): Update for renaming.
394 (selftest::test_metric_conditions): Likewise.
395 (selftest::spellcheck_c_tests): Likewise.
396 * spellcheck.h (levenshtein_distance): Rename both overloads to...
397 (get_edit_distance): ...this.
398 (best_match::consider): Update for renaming.
400 2018-06-12 Martin Sebor <msebor@redhat.com>
402 PR tree-optimization/85259
403 * builtins.c (compute_objsize): Handle constant offsets.
404 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
405 true iff a warning has been issued.
406 * gimple.h (gimple_nonartificial_location): New function.
407 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
408 gimple_nonartificial_location and handle -Wno-system-headers.
409 (handle_builtin_stxncpy): Same.
411 2018-06-12 Martin Sebor <msebor@redhat.com>
414 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
416 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
418 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
419 BUILTIN_VEC_XST entries for pointer to double and long long.
421 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
424 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
426 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
429 2018-06-12 Martin Liska <mliska@suse.cz>
431 * doc/options.texi: Document IntegerRange.
433 2018-06-12 Martin Liska <mliska@suse.cz>
435 * config/i386/i386.opt: Make MPX-related options as Deprecated.
436 * opt-functions.awk: Handle Deprecated flag.
437 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
439 (read_cmdline_option): Report warning for a deprecated option.
440 * opts.h (struct cl_option): Add new field cl_deprecated.
441 (CL_ERR_DEPRECATED): New.
443 2018-06-12 Martin Liska <mliska@suse.cz>
445 * doc/options.texi: Document Deprecated option flag.
447 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
449 * config/arc/arc-arch.h (arc_extras): New enum.
450 (arc_cpu_t):Add field extra.
451 (arc_cpu_types): Consider the extras.
452 * config/arc/arc-cpus.def: Add extras info.
453 * config/arc/arc-opts.h (processor_type): Consider extra field.
454 * config/arc/arc.c (arc_override_options): Handle extra field.
456 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
458 * config/arc/arc-arch.h: Update ARC_OPTX macro.
459 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
461 * config/arc/arc.c (arc_init): Update pic warning.
462 (irq_range): Update irq range parsing warnings.
463 (arc_override_options): Update various warning messages.
464 (arc_handle_aux_attribute): Likewise.
466 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
468 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
470 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
472 * doc/sourcebuild.texi: Document usage of line number 0 in verify
473 compiler messages directives.
475 2018-06-12 Matthew Fortune <mfortune@gmail.com>
477 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
478 * config/mips/mips-tables.opt: Regenerate.
479 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
481 * doc/invoke.texi: Document -march=i6500.
483 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
485 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
486 (i6400_gpmul): Add cpu_unit.
487 (i6400_gpdiv): Likewise.
488 (i6400_msa_add_d): Update reservations.
489 (i6400_msa_int_add) Likewise.
490 (i6400_msa_short_logic3) Likewise.
491 (i6400_msa_short_logic2) Likewise.
492 (i6400_msa_short_logic) Likewise.
493 (i6400_msa_move) Likewise.
494 (i6400_msa_cmp) Likewise.
495 (i6400_msa_short_float2) Likewise.
496 (i6400_msa_div_d) Likewise.
497 (i6400_msa_long_logic1) Likewise.
498 (i6400_msa_long_logic2) Likewise.
499 (i6400_msa_mult) Likewise.
500 (i6400_msa_long_float2) Likewise.
501 (i6400_msa_long_float4) Likewise.
502 (i6400_msa_long_float5) Likewise.
503 (i6400_msa_long_float8) Likewise.
504 (i6400_fpu_fadd): Include frint type.
505 (i6400_fpu_store): New define_insn_reservation.
506 (i6400_fpu_load): Likewise.
507 (i6400_fpu_move): Likewise.
508 (i6400_fpu_fcmp): Likewise.
509 (i6400_fpu_fmadd): Likewise.
510 (i6400_int_mult): Include imul3nc type and update reservation.
511 (i6400_int_div): Include idiv3 type and update reservation.
512 (i6400_int_load): Update to check type not move_type.
513 (i6400_int_store): Likewise.
514 (i6400_int_prefetch): Set zero latency.
516 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
518 * gcc.c: Document new %@{...} sequence.
519 (LINK_COMMAND_SPEC): Use it for the -L switches.
520 (cpp_unique_options): Use it for the -I switches.
521 (at_file_argbuf): New global variable.
522 (in_at_file): Likewise.
523 (alloc_args): Create at_file_argbuf.
524 (clear_args): Truncate at_file_argbuf.
525 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
526 (open_at_file): New function.
527 (close_at_file): Likewise.
528 (create_at_file): Delete.
529 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
532 (validate_switches_from_spec): Deal with %@{...} sequence.
533 (validate_switches): Likewise.
534 (driver::finalize): Call clear_args.
536 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
538 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
540 2018-06-11 Martin Sebor <msebor@redhat.com>
542 * doc/invoke.texi (-Wall): List -Wc++17-compat.
543 (Wno-class-memaccess): Add @opindex.
544 (Wno-templates, Wno-multiple-inheritance): Same.
545 (Wno-virtual-inheritance, Wno-namespaces): Same.
546 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
547 (Wno-format-overflow, Wno-format-truncation): Same.
548 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
549 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
550 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
551 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
552 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
553 (Wno-misspelled-isr): Same.
555 2018-06-11 Martin Sebor <msebor@redhat.com>
557 * PR tree-optimization/86083
558 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
560 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
562 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
564 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
567 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
568 on the correct operand.
569 (*movdi_internal64): Ditto.
571 2018-06-11 Martin Liska <mliska@suse.cz>
573 PR tree-optimization/86089
574 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
576 2018-06-11 Julia Koval <julia.koval@intel.com>
578 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
579 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
580 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
582 2018-06-11 Olivier Hainque <hainque@adacore.com>
584 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
585 for Ada with strict dwarf2.
587 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
590 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
593 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
595 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
597 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
599 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
602 2018-06-08 Martin Liska <mliska@suse.cz>
604 * tree-cfg.h (debug_function): Fix argument type to match
607 2018-06-08 Martin Liska <mliska@suse.cz>
609 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
610 Remove usage of MPX-related (and removed) fields.
611 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
613 2018-06-08 David Malcolm <dmalcolm@redhat.com>
615 * cfg.c (debug): Use TDF_NONE rather than 0.
616 * cfghooks.c (debug): Likewise.
617 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
618 (struct dump_option_value_info): Convert to...
619 (struct kv_pair): ...this template type.
620 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
622 (optinfo_verbosity_options): Likewise.
623 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
625 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
626 than int for "optgroup_flags" param.
627 (dump_generic_expr_loc): Use dump_flags_t rather than int for
629 (dump_dec): Likewise.
630 (dump_finish): Use TDF_NONE rather than 0.
631 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
632 rather than int for "optgroup_flags" param. Use TDF_NONE rather
633 than 0. Update for change to option_ptr.
634 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
635 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
636 0. Update for changes to optinfo_verbosity_options and
638 (opt_info_switch_p): Convert optgroup_flags from int to
640 (dump_basic_block): Use dump_flags_t rather than int
641 for "dump_kind" param.
642 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
643 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
644 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
645 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
646 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
647 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
648 TDF_NONE): Convert from macros to...
649 (enum dump_flag): ...this new enum.
650 (dump_flags_t): Update to use enum.
651 (operator|, operator&, operator~, operator|=, operator&=):
652 Implement for dump_flags_t.
653 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
654 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
655 Convert from macros to...
656 (enum optgroup_flag): ...this new enum.
657 (optgroup_flags_t): New typedef.
658 (operator|, operator|=): Implement for optgroup_flags_t.
659 (struct dump_file_info): Convert field "alt_flags" to
660 dump_flags_t. Convert field "optgroup_flags" to
662 (dump_basic_block): Use dump_flags_t rather than int for param.
663 (dump_generic_expr_loc): Likewise.
664 (dump_dec): Likewise.
665 (dump_register): Convert param "optgroup_flags" to
667 (opt_info_enable_passes): Likewise.
668 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
670 * gimple-pretty-print.c (debug): Likewise.
671 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
672 (merged_store_group::apply_stores): Likewise.
673 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
674 * gimple.c (verify_gimple_pp): Likewise.
675 * graphite-poly.c (print_pbb_body): Likewise.
676 * passes.c (pass_manager::register_one_dump_file): Convert
677 local "optgroup_flags" to optgroup_flags_t.
678 * print-tree.c (print_node): Use TDF_NONE rather than 0.
680 (debug_body): Likewise.
681 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
683 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
685 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
686 (convert_mult_to_fma): Likewise.
687 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
688 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
689 * tree-vect-data-refs.c (dump_lower_bound): Convert param
690 "dump_kind" to dump_flags_t.
692 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
694 * config/rs6000/rs6000.c (min, max): Delete.
696 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
698 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
699 -mabi=spe and -mabi=no-spe.
701 2018-06-08 Martin Liska <mliska@suse.cz>
703 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
704 where we expect an existing summary.
706 2018-06-08 Martin Liska <mliska@suse.cz>
708 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
709 * ipa-inline.h (estimate_edge_growth): Likewise.
711 2018-06-08 Martin Liska <mliska@suse.cz>
713 * cgraph.c (function_version_hasher::hash): Use
714 cgraph_node::get_uid ().
715 (function_version_hasher::equal):
716 * cgraph.h (cgraph_node::get_uid): New method.
717 * ipa-inline.c (update_caller_keys): Use
718 cgraph_node::get_uid ().
719 (update_callee_keys): Likewise.
720 * ipa-utils.c (searchc): Likewise.
721 (ipa_reduced_postorder): Likewise.
722 * lto-cgraph.c (input_node): Likewise.
723 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
724 * symbol-summary.h (symtab_insertion): Likewise.
725 (symtab_removal): Likewise.
726 (symtab_duplication): Likewise.
727 * tree-pretty-print.c (dump_function_header): Likewise.
728 * tree-sra.c (convert_callers_for_node): Likewise.
730 2018-06-08 Martin Liska <mliska@suse.cz>
732 * cgraph.c (symbol_table::create_edge): Always assign a new
734 (symbol_table::free_edge): Do not recycle numbers.
735 * cgraph.h (cgraph_edge::get): New method.
736 * symbol-summary.h (symtab_removal): Use it.
737 (symtab_duplication): Likewise.
738 (call_summary::hashable_uid): Remove.
740 2018-06-08 Martin Liska <mliska@suse.cz>
742 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
743 (initialize_growth_caches): Remove.
744 (free_growth_caches): Likewise.
745 (do_estimate_edge_time): Use edge_growth_cache.
746 (do_estimate_edge_size): Likewise.
747 (do_estimate_edge_hints): Likewise.
748 * ipa-inline.c (reset_edge_caches): Likewise.
749 (recursive_inlining): Likewise.
750 (inline_small_functions): Likewise.
751 * ipa-inline.h (initialize_growth_caches): Remove.
752 (estimate_edge_size): Likewise.
753 (estimate_edge_time): Likewise.
754 (estimate_edge_hints): Likewise.
755 (reset_edge_growth_cache): Likewise.
756 * symbol-summary.h (call_summary::remove): New method.
758 2018-06-08 Martin Liska <mliska@suse.cz>
760 * ipa-cp.c (class edge_clone_summary): New summary.
761 (grow_edge_clone_vectors): Remove.
762 (ipcp_edge_duplication_hook): Remove.
763 (class edge_clone_summary_t): New call_summary class.
764 (ipcp_edge_removal_hook): Remove.
765 (edge_clone_summary_t::duplicate): New function.
766 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
767 (create_specialized_node): Likewise.
768 (ipcp_driver): Initialize edge_clone_summaries and do not
771 2018-06-08 Martin Liska <mliska@suse.cz>
773 * symbol-summary.h (get): New function.
774 (call_summary::m_initialize_when_cloning): New class member.
776 2018-06-08 Martin Liska <mliska@suse.cz>
778 * cgraph.c (cgraph_node::remove): Do not recycle uid.
779 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
780 (symbol_table::allocate_cgraph_symbol): Do not set uid.
781 * passes.c (uid_hash_t): Record removed_nodes by their uids.
782 (remove_cgraph_node_from_order): Use the removed_nodes set.
783 (do_per_function_toporder): Likwise.
784 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
785 instead of summary_uid.
786 (symtab_removal): Likewise.
787 (symtab_duplication): Likewise.
789 2018-06-08 Martin Liska <mliska@suse.cz>
791 * ipa-cp.c (ipcp_store_bits_results): Use
792 ipcp_transformation_sum.
793 (ipcp_store_vr_results): Likewise.
794 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
796 (ipcp_transformation_initialize): ... this.
797 (ipa_set_node_agg_value_chain):
798 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
799 (write_ipcp_transformation_info): Likewise.
800 (read_ipcp_transformation_info): Likewise.
801 (ipcp_update_bits): Likewise.
802 (ipcp_update_vr): Likewise.
803 (ipcp_transform_function): Likewise.
804 * ipa-prop.h: Rename ipcp_transformation_summary to
806 (class ipcp_transformation_t): New function summary.
807 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
808 (ipa_get_agg_replacements_for_node): Likewise.
810 2018-06-08 Martin Liska <mliska@suse.cz>
812 * ipa-pure-const.c (struct funct_state_d): Do it class instead
814 (class funct_state_summary_t): New function_summary class.
815 (has_function_state): Remove.
816 (get_function_state): Likewise.
817 (set_function_state): Likewise.
818 (add_new_function): Likewise.
819 (funct_state_summary_t::insert): New function.
820 (duplicate_node_data): Remove.
821 (remove_node_data): Remove.
822 (funct_state_summary_t::duplicate): New function.
823 (register_hooks): Create new funct_state_summaries.
824 (pure_const_generate_summary): Use it.
825 (pure_const_write_summary): Likewise.
826 (pure_const_read_summary): Likewise.
827 (propagate_pure_const): Likewise.
828 (propagate_nothrow): Likewise.
829 (dump_malloc_lattice): Likewise.
830 (propagate_malloc): Likewise.
831 (execute): Do not register hooks, just remove summary
833 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
836 2018-06-08 Martin Liska <mliska@suse.cz>
838 * ipa-reference.c (remove_node_data): Remove.
839 (duplicate_node_data): Likewise.
840 (class ipa_ref_var_info_summary_t): New class.
841 (class ipa_ref_opt_summary_t): Likewise.
842 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
843 (get_reference_optimization_summary): Use
844 ipa_ref_opt_sum_summaries.
845 (set_reference_vars_info): Remove.
846 (set_reference_optimization_summary): Likewise.
847 (ipa_init): Create summaries.
848 (init_function_info): Use function summary.
849 (ipa_ref_opt_summary_t::duplicate): New function.
850 (ipa_ref_opt_summary_t::remove): New function.
851 (get_read_write_all_from_node): Fix GNU coding style.
852 (propagate): Use function summary.
853 (write_node_summary_p): Fix GNU coding style.
854 (stream_out_bitmap): Likewise.
855 (ipa_reference_read_optimization_summary): Use function summary.
856 (ipa_reference_c_finalize): Do not release hooks.
858 2018-06-08 Martin Liska <mliska@suse.cz>
860 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
861 (analyze_function_body): Extract multiple calls of get_create.
862 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
863 * ipa-inline.c (recursive_inlining): Use ::get method.
864 * ipa-inline.h (estimate_edge_growth): Likewise.
866 2018-06-08 Martin Liska <mliska@suse.cz>
868 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
870 (hsa_function_summary::hsa_function_summary): Use the new enum
872 (hsa_gpu_implementation_p): Use hsa_summaries::get.
873 * hsa-gen.c (hsa_get_host_function): Likewise.
874 (get_brig_function_name): Likewise.
875 * ipa-hsa.c (process_hsa_functions): Likewise.
876 (ipa_hsa_write_summary): Likewise.
877 * symbol-summary.h (symtab_duplication): Use ::get function/
880 2018-06-08 Martin Liska <mliska@suse.cz>
882 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
884 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
885 (hsa_register_kernel): Likewise.
886 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
887 * hsa-gen.c (hsa_get_host_function): Likewise.
888 (get_brig_function_name): Likewise.
889 (generate_hsa): Likewise.
890 (pass_gen_hsail::execute): Likewise.
891 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
892 (devirtualization_time_bonus): Likewise.
893 (ipcp_propagate_stage): Likewise.
894 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
895 (edge_set_predicate): Likewise.
896 (evaluate_conditions_for_known_args): Likewise.
897 (evaluate_properties_for_edge): Likewise.
898 (ipa_fn_summary::reset): Likewise.
899 (ipa_fn_summary_t::duplicate): Likewise.
900 (dump_ipa_call_summary): Likewise.
901 (ipa_dump_fn_summary): Likewise.
902 (analyze_function_body): Likewise.
903 (compute_fn_summary): Likewise.
904 (estimate_edge_devirt_benefit): Likewise.
905 (estimate_edge_size_and_time): Likewise.
906 (estimate_calls_size_and_time): Likewise.
907 (estimate_node_size_and_time): Likewise.
908 (inline_update_callee_summaries): Likewise.
909 (remap_edge_change_prob): Likewise.
910 (remap_edge_summaries): Likewise.
911 (ipa_merge_fn_summary_after_inlining): Likewise.
912 (ipa_update_overall_fn_summary): Likewise.
913 (read_ipa_call_summary): Likewise.
914 (inline_read_section): Likewise.
915 (write_ipa_call_summary): Likewise.
916 (ipa_fn_summary_write): Likewise.
917 (ipa_free_fn_summary): Likewise.
918 * ipa-hsa.c (process_hsa_functions): Likewise.
919 (ipa_hsa_write_summary): Likewise.
920 (ipa_hsa_read_section): Likewise.
921 * ipa-icf.c (sem_function::merge): Likewise.
922 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
923 (do_estimate_edge_time): Likewise.
924 (estimate_size_after_inlining): Likewise.
925 (estimate_growth): Likewise.
926 (growth_likely_positive): Likewise.
927 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
928 (inline_call): Likewise.
929 * ipa-inline.c (caller_growth_limits): Likewise.
930 (can_inline_edge_p): Likewise.
931 (can_inline_edge_by_limits_p): Likewise.
932 (compute_uninlined_call_time): Likewise.
933 (compute_inlined_call_time): Likewise.
934 (want_inline_small_function_p): Likewise.
935 (edge_badness): Likewise.
936 (update_caller_keys): Likewise.
937 (update_callee_keys): Likewise.
938 (recursive_inlining): Likewise.
939 (inline_small_functions): Likewise.
940 (inline_to_all_callers_1): Likewise.
941 (dump_overall_stats): Likewise.
942 (early_inline_small_functions): Likewise.
943 (early_inliner): Likewise.
944 * ipa-inline.h (estimate_edge_growth): Likewise.
945 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
946 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
947 * ipa-prop.h (IPA_NODE_REF): Likewise.
948 (IPA_EDGE_REF): Likewise.
949 * ipa-pure-const.c (malloc_candidate_p): Likewise.
950 (propagate_malloc): Likewise.
951 * ipa-split.c (execute_split_functions): Likewise.
952 * symbol-summary.h: Rename get to get_create.
954 (get_create): Likewise.
955 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
957 2018-06-08 Martin Liska <mliska@suse.cz>
959 * symbol-summary.h (release): Move definition out of class
961 (symtab_removal): Likewise.
962 (symtab_duplication): Likewise.
964 2018-06-08 Martin Liska <mliska@suse.cz>
966 * symbol-summary.h (function_summary): Move constructor
967 implementation out of class declaration.
969 (symtab_insertion): Likewise.
970 (symtab_removal): Likewise.
971 (symtab_duplication): Likewise.
974 2018-06-08 Martin Liska <mliska@suse.cz>
976 * Makefile.in: Remove support for MPX (macros, related functions,
977 fields in cgraph_node, ...).
978 * builtin-types.def (BT_BND): Likewise.
979 (BT_FN_BND_CONST_PTR): Likewise.
980 (BT_FN_CONST_PTR_BND): Likewise.
981 (BT_FN_VOID_PTR_BND): Likewise.
982 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
983 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
984 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
985 (expand_builtin_mempcpy_with_bounds): Likewise.
986 (expand_builtin_memset_with_bounds): Likewise.
987 (expand_builtin_memset_args): Likewise.
988 (std_expand_builtin_va_start): Likewise.
989 (expand_builtin): Likewise.
990 (expand_builtin_with_bounds): Likewise.
991 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
992 (DEF_LIB_BUILTIN_CHKP): Likewise.
993 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
994 (DEF_CHKP_BUILTIN): Likewise.
995 (BUILT_IN_MEMCPY): Likewise.
996 (BUILT_IN_MEMMOVE): Likewise.
997 (BUILT_IN_MEMPCPY): Likewise.
998 (BUILT_IN_MEMSET): Likewise.
999 (BUILT_IN_STPCPY): Likewise.
1000 (BUILT_IN_STRCAT): Likewise.
1001 (BUILT_IN_STRCHR): Likewise.
1002 (BUILT_IN_STRCPY): Likewise.
1003 (BUILT_IN_STRLEN): Likewise.
1004 (BUILT_IN_MEMCPY_CHK): Likewise.
1005 (BUILT_IN_MEMMOVE_CHK): Likewise.
1006 (BUILT_IN_MEMPCPY_CHK): Likewise.
1007 (BUILT_IN_MEMSET_CHK): Likewise.
1008 (BUILT_IN_STPCPY_CHK): Likewise.
1009 (BUILT_IN_STRCAT_CHK): Likewise.
1010 (BUILT_IN_STRCPY_CHK): Likewise.
1011 * calls.c (store_bounds): Likewise.
1012 (emit_call_1): Likewise.
1013 (special_function_p): Likewise.
1014 (maybe_warn_nonstring_arg): Likewise.
1015 (initialize_argument_information): Likewise.
1016 (finalize_must_preallocate): Likewise.
1017 (compute_argument_addresses): Likewise.
1018 (expand_call): Likewise.
1019 * cfgexpand.c (expand_call_stmt): Likewise.
1020 (expand_return): Likewise.
1021 (expand_gimple_stmt_1): Likewise.
1022 (pass_expand::execute): Likewise.
1023 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
1024 (cgraph_node::remove): Likewise.
1025 (cgraph_node::dump): Likewise.
1026 (cgraph_node::verify_node): Likewise.
1027 * cgraph.h (chkp_function_instrumented_p): Likewise.
1028 (symtab_node::get_alias_target): Likewise.
1029 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
1030 (cgraph_local_p): Likewise.
1031 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
1032 (cgraph_edge::rebuild_references): Likewise.
1033 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
1034 (walk_polymorphic_call_targets): Likewise.
1035 (cgraph_node::expand_thunk): Likewise.
1036 (symbol_table::output_weakrefs): Likewise.
1037 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
1038 (ix86_handle_option): Likewise.
1039 * config/i386/constraints.md: Likewise.
1040 * config/i386/i386-builtin-types.def (BND): Likewise.
1044 * config/i386/i386-builtin.def (BDESC_END): Likewise.
1045 (BDESC_FIRST): Likewise.
1047 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
1048 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
1049 * config/i386/i386.c (enum reg_class): Likewise.
1050 (ix86_target_string): Likewise.
1051 (ix86_option_override_internal): Likewise.
1052 (ix86_conditional_register_usage): Likewise.
1053 (ix86_valid_target_attribute_inner_p): Likewise.
1054 (ix86_set_indirect_branch_type): Likewise.
1055 (ix86_set_current_function): Likewise.
1056 (ix86_function_arg_regno_p): Likewise.
1057 (init_cumulative_args): Likewise.
1058 (ix86_function_arg_advance): Likewise.
1059 (ix86_function_arg): Likewise.
1060 (ix86_pass_by_reference): Likewise.
1061 (ix86_function_value_regno_p): Likewise.
1062 (ix86_function_value_1): Likewise.
1063 (ix86_function_value_bounds): Likewise.
1064 (ix86_return_in_memory): Likewise.
1065 (ix86_setup_incoming_vararg_bounds): Likewise.
1066 (ix86_va_start): Likewise.
1067 (indirect_thunk_need_prefix): Likewise.
1068 (print_reg): Likewise.
1069 (ix86_print_operand): Likewise.
1070 (ix86_expand_call): Likewise.
1071 (ix86_output_function_return): Likewise.
1072 (reg_encoded_number): Likewise.
1073 (BDESC_VERIFYS): Likewise.
1074 (ix86_init_mpx_builtins): Likewise.
1075 (ix86_init_builtins): Likewise.
1076 (ix86_emit_cmove): Likewise.
1077 (ix86_emit_move_max): Likewise.
1078 (ix86_expand_builtin): Likewise.
1079 (ix86_builtin_mpx_function): Likewise.
1080 (ix86_get_arg_address_for_bt): Likewise.
1081 (ix86_load_bounds): Likewise.
1082 (ix86_store_bounds): Likewise.
1083 (ix86_load_returned_bounds): Likewise.
1084 (ix86_store_returned_bounds): Likewise.
1085 (ix86_class_likely_spilled_p): Likewise.
1086 (ix86_hard_regno_mode_ok): Likewise.
1087 (x86_order_regs_for_local_alloc): Likewise.
1088 (ix86_mitigate_rop): Likewise.
1089 (ix86_bnd_prefixed_insn_p): Likewise.
1090 (ix86_mpx_bound_mode): Likewise.
1091 (ix86_make_bounds_constant): Likewise.
1092 (ix86_initialize_bounds): Likewise.
1093 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
1094 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
1095 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
1096 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
1097 (TARGET_CHKP_BOUND_MODE): Likewise.
1098 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
1099 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
1100 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
1101 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
1102 * config/i386/i386.h (TARGET_MPX): Likewise.
1103 (TARGET_MPX_P): Likewise.
1104 (VALID_BND_REG_MODE): Likewise.
1105 (FIRST_BND_REG): Likewise.
1106 (LAST_BND_REG): Likewise.
1107 (enum reg_class): Likewise.
1108 (BND_REG_P): Likewise.
1109 (BND_REGNO_P): Likewise.
1110 (BNDmode): Likewise.
1111 (ADJUST_INSN_LENGTH): Likewise.
1112 * config/i386/i386.md: Likewise.
1113 * config/i386/i386.opt: Likewise.
1114 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
1115 (defined): Likewise.
1116 (LINK_MPX): Likewise.
1117 (MPX_SPEC): Likewise.
1118 (LIBMPX_SPEC): Likewise.
1119 (LIBMPXWRAPPERS_SPEC): Likewise.
1120 (CHKP_SPEC): Likewise.
1121 * config/i386/predicates.md: Likewise.
1122 * dbxout.c (dbxout_type): Likewise.
1123 * doc/extend.texi: Likewise.
1124 * doc/invoke.texi: Likewise.
1125 * doc/md.texi: Likewise.
1126 * doc/tm.texi: Likewise.
1127 * doc/tm.texi.in: Likewise.
1128 * dwarf2out.c (is_base_type): Likewise.
1129 (gen_formal_types_die): Likewise.
1130 (gen_subprogram_die): Likewise.
1131 (gen_type_die_with_usage): Likewise.
1132 (gen_decl_die): Likewise.
1133 (dwarf2out_late_global_decl): Likewise.
1134 * expr.c (expand_assignment): Likewise.
1135 (emit_storent_insn): Likewise.
1136 (store_expr_with_bounds): Likewise.
1137 (store_expr): Likewise.
1138 (expand_expr_real_1): Likewise.
1139 * expr.h (store_expr_with_bounds): Likewise.
1140 * function.c (use_register_for_decl): Likewise.
1141 (struct bounds_parm_data): Likewise.
1142 (assign_parms_augmented_arg_list): Likewise.
1143 (assign_parm_find_entry_rtl): Likewise.
1144 (assign_parm_is_stack_parm): Likewise.
1145 (assign_parm_load_bounds): Likewise.
1146 (assign_bounds): Likewise.
1147 (assign_parms): Likewise.
1148 (expand_function_start): Likewise.
1149 * gcc.c (CHKP_SPEC): Likewise.
1150 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
1151 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
1152 (wrestrict_dom_walker::check_call): Likewise.
1153 * gimple.c (gimple_build_call_from_tree): Likewise.
1154 * gimple.h (enum gf_mask): Likewise.
1155 (gimple_call_with_bounds_p): Likewise.
1156 (gimple_call_set_with_bounds): Likewise.
1157 * gimplify.c (gimplify_init_constructor): Likewise.
1158 * ipa-cp.c (initialize_node_lattices): Likewise.
1159 (propagate_constants_across_call): Likewise.
1160 (find_more_scalar_values_for_callers_subset): Likewise.
1161 * ipa-hsa.c (process_hsa_functions): Likewise.
1162 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
1163 * ipa-icf.c (sem_function::merge): Likewise.
1164 * ipa-inline.c (early_inliner): Likewise.
1165 * ipa-pure-const.c (warn_function_noreturn): Likewise.
1166 (warn_function_cold): Likewise.
1167 (propagate_pure_const): Likewise.
1168 * ipa-ref.h (enum GTY): Likewise.
1169 * ipa-split.c (find_retbnd): Likewise.
1170 (consider_split): Likewise.
1171 (split_function): Likewise.
1172 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
1173 * ipa.c (walk_polymorphic_call_targets): Likewise.
1174 (symbol_table::remove_unreachable_nodes): Likewise.
1175 (process_references): Likewise.
1176 (cgraph_build_static_cdtor_1): Likewise.
1177 * lto-cgraph.c (lto_output_node): Likewise.
1178 (output_refs): Likewise.
1179 (compute_ltrans_boundary): Likewise.
1180 (input_overwrite_node): Likewise.
1181 (input_node): Likewise.
1182 (input_cgraph_1): Likewise.
1183 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
1184 * passes.c (pass_manager::execute_early_local_passes): Likewise.
1185 (class pass_chkp_instrumentation_passes): Likewise.
1186 (make_pass_chkp_instrumentation_passes): Likewise.
1187 * passes.def: Likewise.
1188 * rtl.h (struct GTY): Likewise.
1189 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
1190 * stor-layout.c (layout_type): Likewise.
1191 * symtab.c: Likewise.
1192 * target.def: Likewise.
1193 * targhooks.c (default_chkp_bound_type): Likewise.
1194 (default_chkp_bound_mode): Likewise.
1195 (default_builtin_chkp_function): Likewise.
1196 (default_chkp_function_value_bounds): Likewise.
1197 (default_chkp_make_bounds_constant): Likewise.
1198 (default_chkp_initialize_bounds): Likewise.
1199 * targhooks.h (default_chkp_bound_type): Likewise.
1200 (default_chkp_bound_mode): Likewise.
1201 (default_builtin_chkp_function): Likewise.
1202 (default_chkp_function_value_bounds): Likewise.
1203 (default_chkp_make_bounds_constant): Likewise.
1204 (default_chkp_initialize_bounds): Likewise.
1205 * toplev.c (compile_file): Likewise.
1206 (process_options): Likewise.
1207 * tree-core.h (DEF_BUILTIN): Likewise.
1208 (DEF_BUILTIN_CHKP): Likewise.
1209 * tree-inline.c (declare_return_variable): Likewise.
1210 (remap_gimple_stmt): Likewise.
1211 (copy_bb): Likewise.
1212 (initialize_inlined_parameters): Likewise.
1213 (expand_call_inline): Likewise.
1214 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
1215 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
1216 (make_pass_ipa_chkp_produce_thunks): Likewise.
1217 (make_pass_chkp): Likewise.
1218 (make_pass_chkp_opt): Likewise.
1219 (make_pass_chkp_instrumentation_passes): Likewise.
1220 * tree-pretty-print.c (dump_generic_node): Likewise.
1221 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
1222 * tree-ssa-dce.c (propagate_necessity): Likewise.
1223 (eliminate_unnecessary_stmts): Likewise.
1224 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
1225 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
1226 * tree-ssa-sccvn.h: Likewise.
1227 * tree-ssa-strlen.c (get_string_length): Likewise.
1228 (valid_builtin_call): Likewise.
1229 (adjust_last_stmt): Likewise.
1230 (handle_builtin_strchr): Likewise.
1231 (handle_builtin_strcpy): Likewise.
1232 (handle_builtin_stxncpy): Likewise.
1233 (handle_builtin_memcpy): Likewise.
1234 (handle_builtin_strcat): Likewise.
1235 (strlen_check_and_optimize_stmt): Likewise.
1236 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
1237 * tree-streamer-in.c: Likewise.
1238 * tree-streamer.c (record_common_node): Likewise.
1239 * tree.c (tree_code_size): Likewise.
1240 (wide_int_to_tree_1): Likewise.
1241 (type_contains_placeholder_1): Likewise.
1242 (build_common_tree_nodes): Likewise.
1243 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
1244 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
1245 (POINTER_BOUNDS_P): Likewise.
1246 (BOUNDED_TYPE_P): Likewise.
1247 (BOUNDED_P): Likewise.
1248 (CALL_WITH_BOUNDS_P): Likewise.
1249 (pointer_bounds_type_node): Likewise.
1250 * value-prof.c (gimple_ic): Likewise.
1251 * var-tracking.c (vt_add_function_parameters): Likewise.
1252 * varasm.c (make_decl_rtl): Likewise.
1253 (assemble_start_function): Likewise.
1254 (output_constant): Likewise.
1255 (maybe_assemble_visibility): Likewise.
1256 * varpool.c (ctor_for_folding): Likewise.
1257 * chkp-builtins.def: Remove.
1258 * ipa-chkp.c: Remove.
1259 * ipa-chkp.h: Remove.
1260 * rtl-chkp.c: Remove.
1261 * rtl-chkp.h: Remove.
1262 * tree-chkp-opt.c: Remove.
1263 * tree-chkp.c: Remove.
1264 * tree-chkp.h: Remove.
1266 2018-06-07 Carl Love <cel@us.ibm.com>
1268 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
1269 vextract_fp_from_shortl): Add BE support.
1271 2018-06-07 Paul Koning <ni1d@arrl.net>
1273 * compare-elim.c (try_merge_compare): Don't merge compare if
1274 address contains a side effect.
1275 (try_eliminate_compare): Likewise.
1277 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
1279 * config.gcc: Support "tremont".
1280 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
1281 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1283 * config/i386/i386.c (m_TREMONT): Define.
1284 (processor_target_table): Add "tremont".
1285 (PTA_TREMONT): Define.
1286 (ix86_lea_outperforms): Add TARGET_TREMONT.
1287 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
1288 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
1289 and M_INTEL_GOLDMONT_PLUS.
1290 (fold_builtin_cpu): Add "tremont".
1291 (ix86_add_stmt_cost): Add TARGET_TREMONT.
1292 (ix86_option_override_internal): Add "tremont".
1293 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
1294 (processor_type): Add PROCESSOR_TREMONT.
1295 * config/i386/x86-tune.def: Add m_TREMONT.
1296 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
1298 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1300 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
1301 symbol defined for msp430i* devices to be lower case.
1303 gcc/testsuite/gcc.target/msp430/
1304 * msp430i-device-symbol.c: New test.
1305 * msp430f-device-symbol.c: New test.
1306 * msp430.h: New test header file.
1308 2018-06-07 Richard Biener <rguenther@suse.de>
1310 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
1311 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
1312 Properly wrap signed arithmetic if overflow wraps.
1314 2018-06-07 Jakub Jelinek <jakub@redhat.com>
1316 PR tree-optimization/69615
1317 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
1318 of a cast from a same precision integral SSA_NAME in a bb dominated
1319 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
1320 cast to utype if rhs2 has already a compatible type.
1322 2018-06-07 Richard Biener <rguenther@suse.de>
1324 PR tree-optimization/85935
1325 * graphite-scop-detection.c (find_params_in_bb): Analyze
1326 condition operands with respect to the correct loop. Assert
1327 the analysis doesn't fail.
1329 2018-06-04 Carl Love <cel@us.ibm.com>
1331 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
1332 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
1333 as it is slightly cheaper.
1334 (first_match_or_eos_index_<mode>):
1335 Calculate index using natural element order.
1336 (first_match_index_<mode>):
1337 Calculate index using natural element order.
1338 (first_match_or_eos_index_<mode>):
1339 Calculate index using natural order.
1340 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
1341 for BE and LE modes.
1342 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
1343 P9V_BUILTIN_VCLZLSBB_V16QI.
1344 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
1347 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
1349 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
1350 indentation and line wrap for many prototypes. Add missing
1351 @smallexample directives around block of prototypes for vec_xl and
1354 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
1356 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
1357 track if we pass or return IEEE 128-bit floating point.
1358 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
1359 C++ mangling that is compatible with GCC 8.1.
1360 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
1361 (init_cumulative_args): Note if we pass or return IEEE 128-bit
1362 floating point types.
1363 (rs6000_function_arg_advance_1): Likewise.
1364 (rs6000_mangle_type): Optionally generate mangled names that match
1365 what GCC 8.1 generated for IEEE 128-bit floating point types.
1366 (rs6000_globalize_decl_name): If we have an external function that
1367 passes or returns IEEE 128-bit types, generate a weak reference
1368 from the mangled name used in GCC 8.1 to the current mangled
1370 (rs6000_init_builtins): Make __ibm128 use the long double type if
1371 long double is IBM extended double. Make __float128 use the long
1372 double type if long double is IEEE 128-bit.
1375 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
1376 macro for __ibm128 built-in functions.
1377 (PACK_IF): Add __ibm128 pack/unpack functions.
1378 (UNPACK_IF): Likewise.
1379 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
1380 enable long double built-in functions if long double is IEEE
1381 128-bit floating point.
1382 (rs6000_invalid_builtin): Update long double built-in function
1384 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
1385 functions, adjust the built-in function to use the long double
1386 built-in function if __ibm128 and long double are the same type.
1387 * doc/extend.texi (PowerPC builtins): Update documention for
1388 __builtin_{,un}pack_longdouble. Add documentation for
1389 __builtin_{,un}pack_ibm128.
1391 2018-06-06 Jim Wilson <jimw@sifive.com>
1393 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
1394 (struct machine_function): New field interrupt_mode.
1395 (riscv_handle_type_attribute): New function. Add forward declaration.
1396 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
1397 (riscv_expand_epilogue): Check interrupt_mode field.
1398 (riscv_set_current_function): Check interrupt attribute args and
1399 set interrupt_mode field.
1400 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
1401 (riscv_sret, riscv_uret): New.
1402 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
1403 new arguments to interrupt attribute.
1405 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
1408 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
1409 Don't handle -mcpu=power8 if -mpower9-vector is also used.
1411 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
1413 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
1414 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
1415 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1416 several redundant entries.
1418 2018-06-06 David Malcolm <dmalcolm@redhat.com>
1420 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
1421 type from "rtx" to "rtx_insn *".
1422 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
1423 for local "call_insn", removing cast.
1424 (ix86_expand_call): Likewise, introducing a "call_insn" local.
1426 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
1428 PR tree-optimization/86066
1429 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
1430 for BIT_INSERT_EXPR stores.
1432 2018-06-06 Richard Biener <rguenther@suse.de>
1434 PR tree-optimization/86062
1435 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1436 component refs ontop
1437 of to be offsetted base.
1439 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1441 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
1442 to be static and remove check on interrupt attribute name.
1444 gcc/testsuite/gcc.target/msp430/
1445 * function-attributes-4.c: New test.
1446 * static-interrupts.c: New test.
1448 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
1450 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
1451 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
1453 2018-06-05 Steve Ellcey <sellcey@cavium.com>
1456 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
1458 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
1459 Remove second argument, change how error is called.
1460 (aarch64_layout_arg): Remove second argument from
1461 aarch64_err_no_fpadvsimd call.
1462 (aarch64_init_cumulative_args): Ditto.
1463 (aarch64_gimplify_va_arg_expr): Ditto.
1464 * config/aarch64/aarch64.md (mov<mode>): Ditto.
1466 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
1468 * config/i386/i386.md (simple_return_indirect_internal): New expander.
1469 (*simple_return_indirect_internal<mode>): Rename from
1470 simple_return_indirect_internal. Use W mode iterator.
1471 (rstorssp): New expander.
1472 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
1473 (clrssbsy): New expander.
1474 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
1476 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
1478 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
1480 (cmse_check_pointed_object): Likewise.
1482 2018-06-05 Martin Liska <mliska@suse.cz>
1484 PR gcov-profile/47618
1485 * doc/invoke.texi: Document how -fprofile-dir format
1488 2018-06-05 Richard Biener <rguenther@suse.de>
1490 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
1491 removal pretend DOM info isn't available so we do not update
1492 it and only remove edges, not dominated blocks. Actually free
1493 DOM info in case we removed something. Remove unreachable blocks.
1494 (mfb_keep_latches): Work with either DOM info or marked backedges.
1495 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
1496 first. Mark backedges if DOM info isn't available.
1497 (Re-)compute DOM info after cleanup_control_flow_pre.
1499 2018-06-05 Richard Biener <rguenther@suse.de>
1501 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
1502 (locus_discrim_hasher::hash): Adjust.
1503 (locus_discrim_hasher::equal): Likewise.
1504 (next_discriminator_for_locus): Work on line directly.
1505 (same_line_p): Pass in expanded locus1 as well.
1506 (assign_discriminators): Avoid redundant location expansions.
1508 2018-06-05 Richard Biener <rguenther@suse.de>
1510 PR tree-optimization/86046
1511 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
1512 if required after clearing TREE_ADDRESSABLE.
1514 2018-06-05 Richard Biener <rguenther@suse.de>
1516 PR tree-optimization/86047
1517 * tree-ssa-loop.c (for_each_index): Glob handling of all
1518 decls and constants and really handle all of them.
1520 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1523 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
1524 qualifier_void_pointer and qualifier_const_void_pointer.
1525 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
1526 (arm_init_builtins): Handle the above.
1527 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
1528 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
1529 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
1532 2018-06-05 Martin Liska <mliska@suse.cz>
1534 * auto-profile.c (read_autofdo_file): Do not use
1535 gcov_ctr_summary struct.
1536 (afdo_callsite_hot_enough_for_early_inline): Likewise.
1537 * coverage.c (struct counts_entry): Likewise.
1538 (read_counts_file): Read just single summary entry.
1539 (get_coverage_counts): Use gcov_summary struct.
1540 * coverage.h (get_coverage_counts): Likewise.
1541 * gcov-dump.c (dump_working_sets): Likewise.
1542 (tag_summary): Dump just single summary.
1543 * gcov-io.c (gcov_write_summary): Write just histogram
1545 (gcov_read_summary): Read just single summary.
1546 (compute_working_sets): Use gcov_summary struct.
1547 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
1548 of GCOV_COUNTERS_SUMMABLE.
1549 (GCOV_COUNTERS_SUMMABLE): Remove.
1550 (GCOV_FIRST_VALUE_COUNTER): Replace with
1551 GCOV_COUNTER_V_INTERVAL.
1552 (struct gcov_ctr_summary): Remove.
1553 (struct gcov_summary): Directly use fields of former
1555 (compute_working_sets): Use gcov_summary struct.
1556 * gcov.c (read_count_file): Do not use ctrs fields.
1557 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
1559 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
1561 * profile.c: Likewise.
1562 * profile.h: Likewise.
1564 2018-06-05 Martin Liska <mliska@suse.cz>
1566 PR gcov-profile/84846
1567 * gcov.c (output_lines): Print working directory only
1568 in intermediate format.
1570 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
1572 * config/s390/s390-builtin-types.def: Add void function type.
1573 * config/s390/s390-builtins.def: Use the function type for the
1576 2018-06-04 Jim Wilson <jimw@sifive.com>
1578 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
1580 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
1581 and maybe_eh_return. Change regno to unsigned int. Use new args to
1582 handle EH_RETURN_DATA_REGNO registers properly.
1583 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
1584 (riscv_expand_epilogue): Update comment. Change argument name and
1585 type. Update code to use new name and type. Pass new args to
1586 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
1588 * config/riscv/riscv.md (NORMAL_RETURN): New.
1589 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
1590 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
1591 (eh_return): Call gen_eh_return_internal and emit barrier.
1592 (eh_return_internal): Call riscv_expand_epilogue.
1594 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
1596 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
1597 bit_insertion field and declare can_be_merged_into method.
1598 (merged_store_group::can_be_merged_into): New method.
1599 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
1600 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
1601 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
1603 2018-06-04 Richard Biener <rguenther@suse.de>
1605 PR tree-optimization/85955
1606 * builtins.c (fold_builtin_sincos): Convert pointers to
1607 destination to appropriate type before dereferencing.
1609 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
1611 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
1613 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
1615 * expr.c (expand_expr_real_1): Force the operand into memory if
1616 its TYPE_MODE is BLKmode and if there is no integer mode for
1617 the number of bits being extracted.
1619 2018-06-04 Jakub Jelinek <jakub@redhat.com>
1623 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
1624 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
1626 2018-06-04 Richard Biener <rguenther@suse.de>
1628 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
1629 (cleanup_tree_cfg_noloop): ... single caller. Do
1630 start_recording_case_labels later.
1632 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
1634 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
1635 to _IMMINTRIN_H_INCLUDED.
1636 * config/i386/pconfigintrin.h: Ditto.
1637 * config/i386/waitpkgintrin.h: Ditto.
1638 * config/i386/immintrin.h: Add includes for sgxintrin.h,
1639 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
1640 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
1641 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
1642 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
1643 waitpkgintrin.h and cldemoteintrin.h.
1645 2018-06-04 Richard Biener <rguenther@suse.de>
1647 PR tree-optimization/86038
1648 * tracer.c (find_best_successor): Check probability for
1649 being initialized, bail out if not.
1651 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
1654 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
1655 of bits to ignore when comparing architectures.
1657 2018-06-04 Jakub Jelinek <jakub@redhat.com>
1659 PR tree-optimization/69615
1660 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
1661 maximum or minimum of the type, try to merge it also as if
1662 range1 is + [-, x - 1] or + [x + 1, -].
1665 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
1667 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
1669 PR tree-optimization/86034
1670 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
1671 the unsigned bitfield type in a bit insertion sequence if it does not
1672 have a larger precision than the bitfield size.
1673 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
1675 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
1677 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
1679 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
1681 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
1682 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
1683 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
1684 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
1686 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
1688 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
1689 Disable -fdelete-null-pointer-checks for ELF toolchain.
1691 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
1692 Kito Cheng <kito.cheng@gmail.com>
1694 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
1695 (nds32le-*-*, nds32be-*-*): Integrate checking process.
1696 (nds32*-*-*): Add glibc and uclibc conditions.
1697 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
1698 (TARGET_EXCEPT_UNWIND_INFO): Define.
1699 * config/nds32/elf.h: New file.
1700 * config/nds32/linux.h: New file.
1701 * config/nds32/nds32-elf.opt: New file.
1702 * config/nds32/nds32-linux.opt: New file.
1703 * config/nds32/nds32-fp-as-gp.c
1704 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
1705 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
1707 (nds32_asm_file_end): Ditto.
1708 (nds32_print_operand): Ditto.
1709 (nds32_insert_attributes): Ditto.
1710 (nds32_init_libfuncs): New function.
1711 (TARGET_HAVE_TLS): Define.
1712 (TARGET_INIT_LIBFUNCS): Define.
1713 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
1715 (TARGET_ELF): Apply different mcmodel setting.
1716 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
1717 been migrated into elf.h and linux.h files.
1718 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
1719 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
1720 (mcmodel): The content has been migrated into nds32-elf.opt and
1721 nds32-linux.opt files.
1722 * config/nds32/t-elf: New file.
1723 * config/nds32/t-linux: New file.
1725 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
1726 Shiva Chen <shiva0217@gmail.com>
1728 * config/nds32/constants.md (unspec_volatile_element): Add
1729 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
1730 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
1732 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
1733 (make_pass_nds32_fp_as_gp): Declare.
1734 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
1736 (nds32_asm_function_end_prologue): Remove unused asm output.
1737 (nds32_asm_function_begin_epilogue): Remove unused asm output.
1738 (nds32_asm_file_start): Output necessary fp_as_gp information.
1739 (nds32_option_override): Adjust register usage.
1740 (nds32_expand_prologue): Consider fp_as_gp situation.
1741 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
1742 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
1745 (simple_return): Ditto.
1746 (omit_fp_begin): Output special directive for fp_as_gp.
1747 (omit_fp_end): Output special directive for fp_as_gp.
1748 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
1749 mforbid-fp-as-gp): New options.
1751 2018-06-01 Mark Wielaard <mark@klomp.org>
1753 * dwarf2out.c (dwarf2out_finish): Remove generation of
1754 DW_AT_loclists_base.
1756 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
1758 * gimple-ssa-store-merging.c: Include gimple-fold.h.
1759 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
1760 (struct merged_store_group): Add bit_insertion field.
1761 (dump_char_array): Use standard hexadecimal format.
1762 (merged_store_group::merged_store_group): Set bit_insertion to false.
1763 (merged_store_group::apply_stores): Use optimal buffer size. Deal
1764 with BIT_INSERT_EXPR stores. Move up code updating the mask and
1765 also print the mask in the dump file.
1766 (pass_store_merging::gate): Minor tweak.
1767 (imm_store_chain_info::coalesce_immediate): Fix wrong association
1768 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
1769 stores with INTEGER_CST stores.
1770 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
1771 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
1772 and use it throughout. Generate bit insertion sequences if need be.
1773 (pass_store_merging::process_store): Remove redundant condition.
1774 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
1776 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
1778 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
1779 the 128-bit floating point types. Fix function comment.
1781 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1783 * config/aarch64/aarch64-simd.md
1784 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
1786 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
1789 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
1791 PR tree-optimization/85989
1792 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
1794 (backprop::intersect_uses): Check it when deciding whether this
1795 is a backedge reference.
1796 (backprop::process_block): Add each phi to m_visited_phis
1797 after visiting it, then clear it at the end.
1799 2018-06-01 Richard Biener <rguenther@suse.de>
1801 * tree-vectorizer.h (vect_dr_stmt): New function.
1802 (vect_get_load_cost): Adjust.
1803 (vect_get_store_cost): Likewise.
1804 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
1805 Use vect_dr_stmt instead of DR_SMTT.
1806 (vect_record_base_alignments): Likewise.
1807 (vect_calculate_target_alignment): Likewise.
1808 (vect_compute_data_ref_alignment): Likewise and make static.
1809 (vect_update_misalignment_for_peel): Likewise.
1810 (vect_verify_datarefs_alignment): Likewise.
1811 (vector_alignment_reachable_p): Likewise.
1812 (vect_get_data_access_cost): Likewise. Pass down
1813 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
1814 (vect_get_peeling_costs_all_drs): Likewise.
1815 (vect_peeling_hash_get_lowest_cost): Likewise.
1816 (vect_enhance_data_refs_alignment): Likewise.
1817 (vect_find_same_alignment_drs): Likewise.
1818 (vect_analyze_data_refs_alignment): Likewise.
1819 (vect_analyze_group_access_1): Likewise.
1820 (vect_analyze_group_access): Likewise.
1821 (vect_analyze_data_ref_access): Likewise.
1822 (vect_analyze_data_ref_accesses): Likewise.
1823 (vect_vfa_segment_size): Likewise.
1824 (vect_small_gap_p): Likewise.
1825 (vectorizable_with_step_bound_p): Likewise.
1826 (vect_prune_runtime_alias_test_list): Likewise.
1827 (vect_analyze_data_refs): Likewise.
1828 (vect_supportable_dr_alignment): Likewise.
1829 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
1830 (vect_gen_prolog_loop_niters): Likewise.
1831 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
1832 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
1834 (vect_recog_mask_conversion_pattern): Likewise.
1835 (vect_try_gather_scatter_pattern): Likewise.
1836 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
1837 to vect_get_store_cost.
1838 (vect_get_store_cost): Get stmt_info instead of DR.
1839 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
1840 (vect_get_load_cost): Get stmt_info instead of DR.
1842 2018-06-01 Richard Biener <rguenther@suse.de>
1845 * gimple-fold.c (var_decl_component_p): Also allow offsetted
1846 vars wrapped in MEM_REFs.
1848 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
1850 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
1851 Fix subreg tests so that we only return a choice between
1852 GENERAL_REGS and FP_REGS if the original classes included both.
1854 2018-06-01 Richard Biener <rguenther@suse.de>
1857 * tree-ssa-structalias.c (get_function_part_constraint):
1858 Handle NULL fi->decl.
1859 (find_func_aliases_for_call): Properly handle indirect
1860 fi from direct call.
1861 (find_func_clobbers): Likewise.
1862 (ipa_pta_execute): Likewise.
1863 (create_variable_info_for): For functions that are ifunc_resolver
1864 resolve to a varinfo that contains the result of the resolver call.
1865 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
1868 2018-05-31 Michael Collison <michael.collison@arm.com>
1870 * config/aarch64/aarch64.md:
1871 (*fix_to_zero_extenddfdi2): New pattern.
1872 * gcc.target/aarch64/fix_extend1.c: New testcase.
1874 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
1878 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
1879 and BUILT_IN_STRNCMP_EQ.
1880 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
1881 BUILT_IN_STRNCMP_EQ.
1882 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
1883 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
1884 (gimple_fold_builtin): Likewise.
1885 * tree-ssa-strlen.c (compute_string_length): New function.
1886 (determine_min_obsize): New function.
1887 (handle_builtin_string_cmp): New function to handle calls to
1888 string compare functions.
1889 (strlen_optimize_stmt): Add handling to builtin string compare
1891 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
1892 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
1893 * tree.c (build_common_builtin_nodes): Add new defines of
1894 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
1896 2018-05-31 Jakub Jelinek <jakub@redhat.com>
1899 * bb-reorder.c (pass_partition_blocks::gate): Return false for
1900 functions with naked attribute.
1902 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
1904 * config/i386/sse.md (avx_vec_concat<mode>):
1905 Substitute concat_tg_mode mode attribute with xtg_mode.
1906 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
1907 (concat_tg_mode): Remove mode attribute.
1909 2018-05-31 Martin Sebor <msebor@redhat.com>
1912 * calls.c (alloc_max_size): Correct a logic error/typo.
1913 Treat excessive arguments as infinite. Warn for invalid arguments.
1914 * doc/invoke.texi (-Walloc-size-larger-than): Update.
1916 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
1919 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
1920 and movx for Haswell.
1922 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
1923 Cesar Philippidis <cesar@codesourcery.com>
1926 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
1927 when emitting error on private/firstprivate reductions.
1928 * omp-low.c (lower_omp_target): Avoid reference-type processing
1929 on pointers for firstprivate clause.
1931 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
1933 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
1936 * config/aarch64/aarch64-simd.md
1937 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
1938 (aarch64_ld1_x3_<mode>): Likewise
1939 (aarch64_st1x2<VALLDIF:mode>): Likewise
1940 (aarch64_st1_x2_<mode>): Likewise
1941 (aarch64_st1x3<VALLDIF:mode>): Likewise
1942 (aarch64_st1_x3_<mode>): Likewise
1943 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
1944 (vld1_s8_x3): Likewise.
1945 (vld1_u16_x3): Likewise.
1946 (vld1_s16_x3): Likewise.
1947 (vld1_u32_x3): Likewise.
1948 (vld1_s32_x3): Likewise.
1949 (vld1_u64_x3): Likewise.
1950 (vld1_s64_x3): Likewise.
1951 (vld1_f16_x3): Likewise.
1952 (vld1_f32_x3): Likewise.
1953 (vld1_f64_x3): Likewise.
1954 (vld1_p8_x3): Likewise.
1955 (vld1_p16_x3): Likewise.
1956 (vld1_p64_x3): Likewise.
1957 (vld1q_u8_x3): Likewise.
1958 (vld1q_s8_x3): Likewise.
1959 (vld1q_u16_x3): Likewise.
1960 (vld1q_s16_x3): Likewise.
1961 (vld1q_u32_x3): Likewise.
1962 (vld1q_s32_x3): Likewise.
1963 (vld1q_u64_x3): Likewise.
1964 (vld1q_s64_x3): Likewise.
1965 (vld1q_f16_x3): Likewise.
1966 (vld1q_f32_x3): Likewise.
1967 (vld1q_f64_x3): Likewise.
1968 (vld1q_p8_x3): Likewise.
1969 (vld1q_p16_x3): Likewise.
1970 (vld1q_p64_x3): Likewise.
1971 (vst1_s64_x2): Likewise.
1972 (vst1_u64_x2): Likewise.
1973 (vst1_f64_x2): Likewise.
1974 (vst1_s8_x2): Likewise.
1975 (vst1_p8_x2): Likewise.
1976 (vst1_s16_x2): Likewise.
1977 (vst1_p16_x2): Likewise.
1978 (vst1_s32_x2): Likewise.
1979 (vst1_u8_x2): Likewise.
1980 (vst1_u16_x2): Likewise.
1981 (vst1_u32_x2): Likewise.
1982 (vst1_f16_x2): Likewise.
1983 (vst1_f32_x2): Likewise.
1984 (vst1_p64_x2): Likewise.
1985 (vst1q_s8_x2): Likewise.
1986 (vst1q_p8_x2): Likewise.
1987 (vst1q_s16_x2): Likewise.
1988 (vst1q_p16_x2): Likewise.
1989 (vst1q_s32_x2): Likewise.
1990 (vst1q_s64_x2): Likewise.
1991 (vst1q_u8_x2): Likewise.
1992 (vst1q_u16_x2): Likewise.
1993 (vst1q_u32_x2): Likewise.
1994 (vst1q_u64_x2): Likewise.
1995 (vst1q_f16_x2): Likewise.
1996 (vst1q_f32_x2): Likewise.
1997 (vst1q_f64_x2): Likewise.
1998 (vst1q_p64_x2): Likewise.
1999 (vst1_s64_x3): Likewise.
2000 (vst1_u64_x3): Likewise.
2001 (vst1_f64_x3): Likewise.
2002 (vst1_s8_x3): Likewise.
2003 (vst1_p8_x3): Likewise.
2004 (vst1_s16_x3): Likewise.
2005 (vst1_p16_x3): Likewise.
2006 (vst1_s32_x3): Likewise.
2007 (vst1_u8_x3): Likewise.
2008 (vst1_u16_x3): Likewise.
2009 (vst1_u32_x3): Likewise.
2010 (vst1_f16_x3): Likewise.
2011 (vst1_f32_x3): Likewise.
2012 (vst1_p64_x3): Likewise.
2013 (vst1q_s8_x3): Likewise.
2014 (vst1q_p8_x3): Likewise.
2015 (vst1q_s16_x3): Likewise.
2016 (vst1q_p16_x3): Likewise.
2017 (vst1q_s32_x3): Likewise.
2018 (vst1q_s64_x3): Likewise.
2019 (vst1q_u8_x3): Likewise.
2020 (vst1q_u16_x3): Likewise.
2021 (vst1q_u32_x3): Likewise.
2022 (vst1q_u64_x3): Likewise.
2023 (vst1q_f16_x3): Likewise.
2024 (vst1q_f32_x3): Likewise.
2025 (vst1q_f64_x3): Likewise.
2026 (vst1q_p64_x3): Likewise.
2028 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2030 * config/msp430/msp430.c (msp430_output_labelref): Prepend
2031 user_label_prefix to name.
2033 * tree-core.h: Update comment about the format of NAME string
2034 passed to handler in attribute_spec.
2036 * config/msp430/msp430.md: Remove erroneous subreg expression from
2037 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
2038 zero_extend{q,h}isi2.
2040 2018-05-30 Borislav Petkov <bp@suse.de>
2042 * doc/extend.texi: Document some architecture specific
2043 constraints and sort entries.
2045 2018-05-30 Martin Sebor <msebor@redhat.com>
2048 * builtins.c (expand_builtin_stpcpy_1): New function.
2049 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
2050 only if the former succeeds.
2052 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2054 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
2057 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2059 * doc/invoke.texi (-flinker-output): Document
2061 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2063 * passes.c (ipa_write_summaries): Only modify statements if body
2065 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
2066 incrementally linking.
2067 (ipa_passes): Likewise.
2068 * lto-cgraph.c (lto_output_node): When incrementally linking do not
2069 pass down resolution info.
2070 * common.opt (flag_incremental_link): Update info.
2071 * gcc.c (plugin specs): Turn flinker-output=* to
2072 -plugin-opt=-linker-output-known
2073 * toplev.c (compile_file): Also cut compilation when doing incremental
2075 * flag-types. (enum lto_partition_model): Add
2076 LTO_LINKER_OUTPUT_NOLTOREL.
2077 (invoke.texi): Add -flinker-output docs.
2078 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
2079 link same way as WPA; do not stream in dead initializers.
2081 * dwarf2out.c (dwarf2out_die_ref_for_decl,
2082 darf2out_register_external_decl): Support incremental link.
2084 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2086 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
2088 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2090 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
2091 it down to simple_object_copy_lto_debug_sections.
2092 (run_gcc): Determine incremental LTO link time and configure
2093 lto1 into non-wpa mode, disable renaming of debug sections.
2095 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
2097 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
2098 descriptions of various incorrectly documented functions.
2100 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
2103 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
2104 address check not strict.
2106 2018-05-30 Richard Biener <rguenther@suse.de>
2108 PR tree-optimization/85964
2109 * tracer.c (better_p): Drop initialized count check, we only
2110 call the function with initialized counts now.
2111 (find_best_successor): Do find a best edge if one
2112 has uninitialized count.
2113 (find_best_predecessor): Likewise. Do BB frequency check only
2114 if count is initialized.
2116 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
2118 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
2119 (aarch64_ldrstr_offset_compare): New.
2120 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
2121 load/store orderings.
2122 (aarch64_gen_adjusted_ldpstp): Likewise.
2124 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
2126 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
2127 Check for subset of GENERAL_REGS and FP_REGS.
2128 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
2131 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
2133 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
2134 and wi::to_poly_offset. Add the current offset and then check
2135 whether the sum fits, rather than using an unchecked addition of
2136 a checked term. Check for a shwi rather than a uhwi.
2137 * expr.c (get_bit_range): Use tree_to_poly_uint64.
2138 (store_constructor): Use poly_int_tree_p.
2139 (expand_expr_real_1): Likewise.
2140 * function.c (assign_temp): Likewise.
2141 * fold-const.c (const_binop): Use poly_int_tree_p and
2143 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
2145 * ipa-icf-gimple.c (func_checker::compare_operand): Use
2146 to_poly_offset for MEM offsets.
2147 * ipa-icf.c (sem_variable::equals): Likewise.
2148 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
2149 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
2150 wi::to_poly_offset for BIT_FIELD_REF offsets.
2151 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
2153 * var-tracking.c (emit_note_insn_var_location): Use
2154 tree_to_poly_uint64.
2156 2018-05-29 Jim Wilson <jimw@sifive.com>
2158 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
2160 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
2163 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
2164 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
2166 (sse4_1_round<mode>2): Use nonimmediate_operand
2167 for operand 1 predicate.
2169 2018-05-29 Martin Sebor <msebor@redhat.com>
2170 Richard Biener <rguenther@suse.de>
2173 * calls.c (get_size_range): Call determine_value_range instead
2174 of get_value_range..
2175 * tree-vrp.h (determine_value_range): Declared new function.
2176 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
2178 2018-05-29 Richard Biener <rguenther@suse.de>
2180 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
2181 sure to use non-pattern stmts for get_earlier_stmt arguments.
2182 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
2183 called on pattern stmts.
2184 (get_later_stmt): Likewise.
2186 2018-05-29 Martin Liska <mliska@suse.cz>
2188 PR gcov-profile/85759
2189 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
2192 2018-05-29 Jakub Jelinek <jakub@redhat.com>
2194 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
2196 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
2200 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
2201 VEC_PACK_FLOAT_EXPR): New tree codes.
2202 * tree-pretty-print.c (op_code_prio): Handle
2203 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
2204 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
2205 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
2206 * tree-inline.c (estimate_operator_cost): Likewise.
2207 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
2208 * fold-const.c (const_binop): Likewise.
2209 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
2210 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
2211 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
2212 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
2213 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
2214 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
2215 * expr.c (expand_expr_real_2): Likewise.
2216 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
2217 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
2218 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
2220 * optabs.c (expand_widen_pattern_expr): For
2221 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
2222 sign from result type rather than operand's type.
2223 (expand_binop_directly): For vec_packu_float_optab and
2224 vec_packs_float_optab allow result type to be different from operand's
2226 * optabs-tree.c (optab_for_tree_code): Handle
2227 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
2228 VEC_PACK_FLOAT_EXPR. Formatting fixes.
2229 * tree-vect-generic.c (expand_vector_operations_1): Handle
2230 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
2231 VEC_PACK_FLOAT_EXPR.
2232 * tree-vect-stmts.c (supportable_widening_operation): Handle
2234 (supportable_narrowing_operation): Handle FLOAT_EXPR.
2235 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
2236 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
2237 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
2238 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
2240 (vec_pack<floatprefix>_float_<mode>): New expander.
2241 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
2243 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
2244 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
2245 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
2246 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
2247 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
2249 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
2250 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
2251 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
2252 VEC_PACK_FLOAT_EXPR): Document.
2254 2018-05-29 Richard Biener <rguenther@suse.de>
2256 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
2258 (stmt_vec_info_vec): Make pointer.
2259 (init_stmt_vec_info_vec): Remove.
2260 (free_stmt_vec_info_vec): Likewise.
2261 (set_stmt_vec_info_vec): New function.
2262 (free_stmt_vec_infos): Likewise.
2263 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
2264 (set_vinfo_for_stmt): Likewise.
2265 (get_earlier_stmt): Likewise.
2266 (get_later_stmt): Likewise.
2267 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
2268 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
2269 (vec_info::~vec_info): Free stmt_vec_infos.
2270 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
2271 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
2272 (pass_slp_vectorize::execute): Likewise.
2273 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
2274 (free_stmt_vec_info_vec): Likewise.
2275 (set_stmt_vec_info_vec): New function.
2276 (free_stmt_vec_infos): Likewise.
2277 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
2278 the global stmt_vec_info_vec.
2279 * tree-parloops.c (gather_scalar_reductions): Use
2280 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
2283 2018-05-29 Richard Biener <rguenther@suse.de>
2285 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
2287 2018-05-29 Martin Liska <mliska@suse.cz>
2288 David Malcolm <dmalcolm@redhat.com>
2290 * vec.c (test_reverse): New.
2291 (vec_c_tests): Add new test.
2292 * vec.h (vl_ptr>::reverse): New function.
2294 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
2296 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
2298 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
2301 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
2303 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
2305 2018-05-28 Richard Biener <rguenther@suse.de>
2307 PR tree-optimization/85933
2308 * tree-vect-data-refs.c (vect_record_base_alignments): Only
2309 look at stmts marked as vectorizable.
2311 2018-05-28 Richard Biener <rguenther@suse.de>
2313 PR tree-optimization/85934
2314 * tree-vect-generic.c (expand_vector_operations_1): Hoist
2315 vector boolean check before scalar optimization.
2317 2018-05-28 Jakub Jelinek <jakub@redhat.com>
2319 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
2322 2018-05-28 Mark Wielaard <mark@klomp.org>
2324 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
2325 if it is an expression containing a minus sign.
2327 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
2329 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
2331 2018-05-27 Paul Koning <ni1d@arrl.net>
2333 * config/pdp11/pdp11.md (truncsihi2): Remove.
2335 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
2336 Chung-Ju Wu <jasonwucj@gmail.com>
2338 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
2340 (unaligned_store_dw): Ditto.
2341 * config/nds32/nds32-memory-manipulation.c
2342 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
2343 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
2344 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
2345 (emit_setmem_word_loop): Rename to ...
2346 (emit_setmem_doubleword_loop): ... this.
2347 (nds32_gen_dup_4_byte_to_word_value): New function.
2348 (nds32_gen_dup_8_byte_to_double_word_value): New function.
2349 (nds32_expand_setmem_loop): Refine implementation.
2350 (nds32_expand_setmem_loop_v3m): Ditto.
2351 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
2354 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
2356 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
2358 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
2360 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
2361 (nds32_init_machine_status): Initialize machine->attr_naked_p and
2362 machine->attr_no_prologue_p.
2363 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
2364 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
2365 (nds32_expand_epilogue): Consider attr_naked_p.
2366 (nds32_expand_epilogue_v3pop): Likewise.
2367 (nds32_can_use_return_insn): Likewise.
2368 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
2369 attr_no_prologue_p fields.
2370 * config/nds32/nds32.opt (mret-in-naked-func): New option.
2372 2018-05-27 Jakub Jelinek <jakub@redhat.com>
2375 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
2377 * config/i386/sse.md
2378 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
2380 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
2382 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
2384 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
2386 (*<floatsuffix>floatv2div2sf2): Rename to ...
2387 (*float<floatunssuffix>v2div2sf2): ... this.
2388 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
2389 (float<floatunssuffix>v2div2sf2_mask): ... this.
2390 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
2391 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
2392 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
2394 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
2396 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
2398 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
2400 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
2402 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
2404 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
2405 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
2406 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
2407 gen_ufix_truncv8dfv8si2.
2408 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
2409 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
2410 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
2411 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
2412 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
2413 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
2414 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
2415 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
2417 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
2421 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
2423 2018-05-25 Jim Wilson <jimw@sifive.com>
2425 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
2426 * config/riscv/riscv.c (struct machine_function): Add
2427 interrupt_handler_p and attribute_checked_p fields.
2428 (riscv_attribute_table): Add interrupt.
2429 (riscv_interrupt_type_p): New.
2430 (riscv_save_reg_p): Save extra regs for interrupt handler.
2431 (riscv_use_save_libcall): Return false for interrupt handler.
2432 (riscv_first_stack_step): Add forward declaration.
2433 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
2434 for interrupt handler with large frame. Use it for saved reg list.
2435 (riscv_expand_prologue): Move flag_stack_usage_info support to
2436 eliminate duplication.
2437 (riscv_expand_epilogue): Generate mret for interrupt handler.
2438 (riscv_epilogue_uses): New.
2439 (riscv_can_use_return_insn): Return false for interrupt handler.
2440 (riscv_function_ok_for_sibcall): Likewise.
2441 (riscv_set_current_function): Add interrupt handler support.
2442 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
2443 * config/riscv/riscv.md (UNSPECV_MRET): New.
2445 (riscv_frflags, riscv_fsflags): Use tab after opcode.
2447 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
2449 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
2451 PR tree-optimization/85712
2452 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
2453 this candidate has already been replaced in-situ by a copy.
2455 2018-05-25 Jason Merrill <jason@redhat.com>
2457 PR c++/80485 - inline function non-zero address.
2458 * symtab.c (nonzero_address): Check DECL_COMDAT.
2460 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
2463 * config/alpha/alpha.md (ashlsi3): New insn pattern.
2464 (*ashlsi_se): Rename from *ashldi_se. Define as sign
2465 extension of SImode operation. Use const123_operand predicate.
2466 (*saddsi_1): Remove.
2467 (*saddl_se_1): Ditto.
2469 (*ssubl_se_1): Ditto.
2470 * config/alpha/predicates.md (const123_operand): New predicate.
2471 * config/alpha/constraints.md (P): Use IN_RANGE.
2473 2018-05-25 Richard Biener <rguenther@suse.de>
2475 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
2477 (ref_maybe_used_by_stmt_p): Likewise.
2478 (stmt_may_clobber_ref_p): Likewise.
2479 (stmt_may_clobber_ref_p_1): Likewise.
2480 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
2482 (ref_maybe_used_by_stmt_p): Likewise.
2483 (stmt_may_clobber_ref_p): Likewise.
2484 (stmt_may_clobber_ref_p_1): Likewise.
2485 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
2486 the alias oracle to disambiguate DRs with stmts DR analysis
2488 (vect_analyze_data_refs): Do not give up on not analyzable
2489 DRs for BB vectorization. Remove code truncating the dataref
2492 2018-05-25 Jakub Jelinek <jakub@redhat.com>
2495 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
2496 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
2497 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
2499 2018-05-25 Richard Biener <rguenther@suse.de>
2501 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
2502 function, combining stmt data ref gathering and fatal analysis
2504 (vect_analyze_data_refs): Remove now redudnant code and simplify.
2505 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
2506 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
2507 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
2508 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
2510 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2512 PR tree-optimization/85720
2513 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
2514 SCC if all partitions are builtins.
2515 (version_loop_by_alias_check): New parameter. Generate cancelable
2516 runtime alias check if all partitions are builtins.
2517 (distribute_loop): Update call to above function.
2519 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2521 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
2522 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
2523 (parm_default_def_partition_arg): Ditto.
2524 (set_parm_default_def_partition): Ditto.
2525 (get_parm_default_def_partitions): Ditto and make it static.
2526 (get_undefined_value_partitions): Ditto and make it static.
2527 (remove_ssa_form): Refactor call to init_var_map here.
2528 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
2529 computation for loop region.
2530 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
2531 (register_default_def): Delete.
2532 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
2533 (parm_default_def_partition_arg): Ditto.
2534 (set_parm_default_def_partition): Ditto.
2535 (get_parm_default_def_partitions): Ditto and make it static.
2536 (get_undefined_value_partitions): Ditto and make it static.
2537 (coalesce_with_default, coalesce_with_default): Update comment.
2538 (create_coalesce_list_for_region): New func factored out from
2539 create_outofssa_var_map.
2540 (populate_coalesce_list_for_outofssa): New func factored out from
2541 create_outofssa_var_map and coalesce_ssa_name.
2542 (create_outofssa_var_map): Delete.
2543 (coalesce_ssa_name): Refactor to support live range computation.
2544 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
2545 (get_parm_default_def_partitions): Delete.
2546 (get_undefined_value_partitions): Ditto.
2547 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
2548 computation for loop region.
2549 (new_tree_live_info, loe_visit_block): Ditto.
2550 (live_worklist, set_var_live_on_entry): Ditto.
2551 (calculate_live_on_exit, verify_live_on_entry): Ditto.
2552 * tree-ssa-live.h (struct _var_map): New fields.
2553 (init_var_map): Change decl.
2554 (region_contains_p): New.
2556 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2558 * tree-ssa-live.h (live_merge_and_clear): Delete.
2560 2018-05-25 Richard Biener <rguenther@suse.de>
2563 * tree-dump.c (dequeue_and_dump): Remove access to removed
2564 operand 2 of a SWITCH_EXPR.
2566 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2568 * doc/sourcebuild.texi (vect_double_cond_arith): Include
2569 multiplication and division.
2570 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
2571 (cond_udiv@var{m}, cond_umod@var{m}): Document.
2572 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
2573 (cond_udiv_optab, cond_umod_optab): New optabs.
2574 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
2575 (IFN_COND_RDIV): New internal functions.
2576 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
2577 TRUNC_MOD_EXPR and RDIV_EXPR.
2578 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
2579 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
2581 (SVE_INT_BINARY): Include mult.
2582 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
2583 (optab, sve_int_op): Handle mult.
2584 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
2586 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
2587 for SVE_INT_BINARY_SD.
2589 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2591 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
2592 (optab, sve_int_op): Handle div and udiv.
2593 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
2594 for SVE_INT_BINARY_SD.
2595 (*<optab><mode>3): New insn for the same.
2597 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2599 * tree-vect-patterns.c: Include predict.h.
2600 (vect_recog_divmod_pattern): Restrict check for division support
2601 to when optimizing for size.
2603 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2605 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
2606 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
2607 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
2608 (gimple_match_op::set_op): Likewise.
2609 (gimple_resimplify4): Declare.
2610 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
2611 (expr::gen_transform): Likewise.
2612 (decision_tree::gen): Generate a simplification routine for 4 operands.
2613 * gimple-match-head.c (gimple_simplify): Add an overload for
2614 4 operands. In the top-level function, handle up to 4 call
2615 arguments and call gimple_resimplify4.
2616 (gimple_resimplify4): New function.
2617 (build_call_internal): Pass a fourth operand.
2618 (maybe_push_to_seq): Likewise.
2619 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
2620 Fold VEC_COND_EXPRs of an operation and a default value into
2621 an IFN_COND_* function if possible.
2622 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
2624 (SVE_COND_FP_BINARY): Include them.
2625 (optab, sve_fp_op): Handle them.
2626 (SVE_INT_BINARY_REV): New code iterator.
2627 (SVE_COND_FP_BINARY_REV): New int iterator.
2628 (commutative): New int attribute.
2629 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
2631 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
2633 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
2634 (*cond_<optab><mode>): New patterns for reversed operands.
2636 2018-05-25 Richard Biener <rguenther@suse.de>
2638 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
2639 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
2640 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
2641 (STMT_VINFO_GROUPED_ACCESS): Adjust.
2642 * tree-vect-data-refs.c (everywhere): Adjust users.
2643 * tree-vect-loop.c (everywhere): Likewise.
2644 * tree-vect-slp.c (everywhere): Likewise.
2645 * tree-vect-stmts.c (everywhere): Likewise.
2646 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
2648 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2650 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
2652 (gcc_cv_as_section_exclude): ... this.
2653 Try Solaris as #exclude syntax.
2654 * configure: Regenerate.
2655 * config.in: Regenerate.
2656 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
2658 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
2659 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
2661 * varasm.c (default_elf_asm_named_section): Don't check if
2662 HAVE_GAS_SECTION_EXCLUDE is defined.
2664 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2666 * doc/md.texi: Update the documentation of the cond_* optabs
2667 to mention the new final operand. Fix GET_MODE_NUNITS call.
2668 Describe the scalar case too.
2669 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
2670 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
2672 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
2673 (get_conditional_internal_fn): Update comment.
2674 * tree-vect-loop.c (vectorizable_reduction): Pass the original
2675 accumulator value as a final argument to conditional functions.
2676 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
2677 a define_expand and add an "else" operand. Assert for now that
2678 the else operand is equal to operand 2. Use SVE_INT_BINARY and
2679 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
2680 (*cond_<optab><mode>): New patterns.
2681 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
2682 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
2683 (UNSPEC_COND_EOR): Delete.
2684 (optab): Remove associated mappings.
2685 (SVE_INT_BINARY): New code iterator.
2686 (sve_int_op): Remove int attribute and add "minus" to the code
2688 (SVE_COND_INT_OP): Delete.
2689 (SVE_COND_FP_OP): Rename to...
2690 (SVE_COND_FP_BINARY): ...this.
2692 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2694 * optabs.c (can_reuse_operands_p): New function.
2695 (maybe_legitimize_operands): Try to reuse the results for
2698 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
2700 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
2701 Add {q} suffix to insn mnemonic.
2703 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2705 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
2706 (msp430_warn_func_return): New.
2708 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
2710 * fold-const.c (tree_nonzero_bits): New function.
2711 * fold-const.h (tree_nonzero_bits): Likewise.
2712 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
2713 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
2715 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
2719 * varasm.c (assemble_alias): Check ifunc_resolver only on
2722 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
2725 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
2726 when memory input operand is handled.
2728 2018-05-24 Luis Machado <luis.machado@linaro.org>
2730 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
2732 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
2734 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
2736 * match.pd: Delay FMA folds until after vectorization.
2738 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
2741 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
2742 address check not strict.
2744 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
2746 * gimple-match.h (gimple_match_op): New class.
2747 (mprts_hook): Replace parameters with a gimple_match_op *.
2748 (maybe_build_generic_op): Likewise.
2749 (gimple_simplified_result_is_gimple_val): Replace parameters with
2750 a const gimple_match_op *.
2751 (gimple_simplify): Replace code_helper * and tree * parameters with
2752 a gimple_match_op * parameter.
2753 (gimple_resimplify1): Replace code_helper *, tree and tree *
2754 parameters with a gimple_match_op * parameter.
2755 (gimple_resimplify2): Likewise.
2756 (gimple_resimplify3): Likewise.
2757 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
2758 parameters with a gimple_match_op * parameter.
2759 * gimple-match-head.c (gimple_simplify): Change prototypes of
2760 auto-generated functions to take a gimple_match_op * instead of
2761 separate code_helper * and tree * parameters. Make the same
2762 change in the top-level overload and update calls to the
2763 gimple_resimplify routines. Update calls to the auto-generated
2764 functions and to maybe_push_res_to_seq in the publicly-facing
2765 operation-specific gimple_simplify overloads.
2766 (gimple_match_op::MAX_NUM_OPS): Define.
2767 (gimple_resimplify1): Replace rcode and ops with a single res_op
2768 parameter. Update call to gimple_simplify.
2769 (gimple_resimplify2): Likewise.
2770 (gimple_resimplify3): Likewise.
2771 (mprts_hook): Replace parameters with a gimple_match_op *.
2772 (maybe_build_generic_op): Likewise.
2773 (build_call_internal): Replace type, nargs and ops with
2774 a gimple_match_op *.
2775 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
2776 with a single gimple_match_op *. Update calls to mprts_hook,
2777 build_call_internal and gimple_simplified_result_is_gimple_val.
2778 Factor out code that is common to the tree_code and combined_fn cases.
2779 * genmatch.c (expr::gen_transform): Replace tem_code and
2780 tem_ops with a gimple_match_op called tem_op. Update calls
2781 to the gimple_resimplify functions and maybe_push_res_to_seq.
2782 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
2783 res_ops. Update call to the gimple_resimplify functions.
2784 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
2785 (decision_tree::gen): Make the functions take a gimple_match_op *
2786 called res_op instead of separate res_code and res_ops parameters.
2787 Update call accordingly.
2788 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
2789 and ops with a single res_op parameter. Update calls to
2790 maybe_build_generic_op and maybe_push_res_to_seq.
2791 (fold_stmt_1): Update calls to gimple_simplify and
2792 replace_stmt_with_simplification.
2793 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
2794 and gimple_simplified_result_is_gimple_val.
2795 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
2797 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
2798 with a gimple_match_op *.
2799 (vn_nary_build_or_lookup): Likewise. Update call to
2800 vn_nary_build_or_lookup_1.
2801 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
2802 gimple_match_op *. Update calls to the gimple_resimplify routines
2803 and to gimple_simplified_result_is_gimple_val.
2804 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
2805 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
2806 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
2807 (visit_nary_op): Likewise.
2808 (visit_reference_op_load): Likewise.
2810 2018-05-23 Luis Machado <luis.machado@linaro.org>
2812 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
2813 modifier for printing the step amount.
2815 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
2818 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
2821 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
2823 * doc/sourcebuild.texi (Endianness): New subsubsection.
2825 2018-05-23 Luis Machado <luis.machado@linaro.org>
2827 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2828 <prefetch_dynamic_strides>: New const bool field.
2829 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2830 prefetch_dynamic_strides.
2831 (exynosm1_prefetch_tune): Likewise.
2832 (thunderxt88_prefetch_tune): Likewise.
2833 (thunderx_prefetch_tune): Likewise.
2834 (thunderx2t99_prefetch_tune): Likewise.
2835 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
2837 (aarch64_override_options_internal): Update to set
2838 PARAM_PREFETCH_DYNAMIC_STRIDES.
2839 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2840 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2841 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2842 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2843 prefetch-dynamic-strides setting.
2845 2018-05-23 Luis Machado <luis.machado@linaro.org>
2847 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2848 <minimum_stride>: New const int field.
2849 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2850 minimum_stride field defaulting to -1.
2851 (exynosm1_prefetch_tune): Likewise.
2852 (thunderxt88_prefetch_tune): Likewise.
2853 (thunderx_prefetch_tune): Likewise.
2854 (thunderx2t99_prefetch_tune): Likewise.
2855 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
2856 <default_opt_level>: Set to 3.
2857 (aarch64_override_options_internal): Update to set
2858 PARAM_PREFETCH_MINIMUM_STRIDE.
2859 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2860 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2861 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2862 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2863 stride is constant and is below the minimum stride threshold.
2865 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2867 * config/arm/arm-cpus.in (mode26): Delete.
2868 (armv4): Delete mode26 reference.
2869 * config/arm/arm.c (arm_configure_build_target): Delete use of
2872 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
2874 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
2876 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
2877 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
2879 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
2880 (floatunsdidf2): Ditto.
2882 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
2884 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
2885 (fixuns_trunc<mode>si2_avx512f): Ditto.
2886 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
2887 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
2888 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
2890 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
2892 PR rtl-optimization/79985
2893 * df-scan.c (df_insn_refs_collect): Remove special case for
2894 global registers and asm statements.
2896 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
2898 * extend.texi (Global Register Variables): Rewrite the bullet list.
2899 Note that the register is available for allocation. Note that access
2900 via inline asm must use constraints. Add note about async-signal
2901 handlers. Remove paragraph about automagic register selection.
2903 2018-05-23 Richard Biener <rguenther@suse.de>
2905 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
2906 of fixed offset from memset VN.
2908 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
2910 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
2912 (alloc_cand_and_find_basis): Initialize first_interp field.
2913 (slsr_process_mul): Modify first_interp field.
2914 (slsr_process_add): Likewise.
2915 (slsr_process_cast): Modify first_interp field for each new
2917 (slsr_process_copy): Likewise.
2918 (dump_candidate): Dump first_interp field.
2919 (replace_mult_candidate): Process all interpretations, not just
2921 (replace_rhs_if_not_dup): Likewise.
2922 (replace_one_candidate): Likewise.
2924 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
2926 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
2928 (aarch64_needs_frame_chain): New function.
2929 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
2931 2018-05-23 Sudakshina Das <sudi.das@arm.com>
2934 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
2935 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
2936 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
2937 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
2938 as true for strict-align.
2939 (aarch64_can_inline_p): Perform checks even when callee has no
2940 attributes to check for strict alignment.
2941 * doc/extend.texi (AArch64 Function Attributes): Document
2943 * doc/invoke.texi: (AArch64 Options): Likewise.
2945 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
2947 PR tree-optimization/85853
2948 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
2949 the handling of the root of the node to...
2950 (vect_slp_analyze_node_operations_1): ...this new function,
2951 and run the whole thing with the child nodes' def types
2952 set according to their SLP node's def type.
2954 2018-05-23 Richard Biener <rguenther@suse.de>
2957 * tree-data-ref.c (create_runtime_alias_checks): Defer
2958 and ignore overflow warnings.
2960 2018-05-23 Yury Gribov <tetra2005@gmail.com>
2962 PR tree-optimization/85822
2963 * tree-vrp.c (is_masked_range_test): Fix handling of negative
2966 2018-05-23 Richard Biener <rguenther@suse.de>
2968 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
2969 memset constants via native_interpret_expr.
2971 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
2974 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
2976 (cgraph_node::create_alias): Likewise.
2977 (cgraph_node::get_availability): Check ifunc_resolver instead
2978 of looking up ifunc attribute.
2979 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
2980 * varasm.c (do_assemble_alias): Likewise.
2981 (assemble_alias): Likewise.
2982 (default_binds_local_p_3): Likewise.
2983 * cgraph.h (cgraph_node): Add ifunc_resolver.
2984 (cgraph_node::only_called_directly_or_aliased_p): Return false
2986 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
2988 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
2989 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
2990 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
2991 instead of looking up ifunc attribute.
2993 2018-05-22 Luis Machado <luis.machado@linaro.org>
2995 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
2997 2018-05-22 Martin Sebor <msebor@redhat.com>
3000 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
3001 only when expasion succeeds.
3002 (expand_builtin_strcmp): Same.
3003 (expand_builtin_strncmp): Same.
3005 2018-05-22 Martin Sebor <msebor@redhat.com>
3007 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
3009 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3010 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3012 * config/aarch64/aarch64-ldpstp.md: Replace uses of
3013 aarch64_mem_pair_operand with memory_operand and delete operand swapping
3015 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3016 Add check for legitimate_address.
3017 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
3018 (aarch64_swap_ldrstr_operands): New.
3019 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
3022 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3023 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3025 * config/aarch64/aarch64.md: New patterns to generate stp
3027 (store_pair_sw, store_pair_dw): New patterns to generate stp for
3028 single words and double words.
3029 (load_pair_sw, load_pair_dw): Likewise.
3030 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
3032 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
3034 * config/aarch64/aarch64-ldpstp.md: Modify peephole
3035 for different mode ldpstp and add peephole for merged zero stores.
3037 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3039 (aarch64_gen_store_pair): Rename calls to match new patterns.
3040 (aarch64_gen_load_pair): Rename calls to match new patterns.
3041 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
3042 (load_pair<DREG:mode><DREG2:mode>): ... This.
3043 (store_pair<mode>): Rename to...
3044 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
3045 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
3047 (V_INT_EQUIV): Handle SImode.
3048 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
3051 2018-05-22 Martin Sebor <msebor@redhat.com>
3054 * calls.c (maybe_warn_nonstring_arg): Use string length to set
3055 or ajust the presumed bound on an operation to avoid unnecessary
3058 2018-05-22 Martin Sebor <msebor@redhat.com>
3060 PR tree-optimization/85826
3061 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
3062 assuming that a DECL necesarily has a constant size.
3064 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
3067 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
3069 2018-05-22 Richard Biener <rguenther@suse.de>
3071 PR tree-optimization/85834
3072 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
3073 non-constant and non-zero memset arguments.
3075 2018-05-22 Martin Liska <mliska@suse.cz>
3078 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
3080 2018-05-22 Richard Biener <rguenther@suse.de>
3082 PR tree-optimization/85863
3083 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
3084 comparisons when vectype is specified.
3085 (vectorizable_condition): Do not specify vectype for
3086 vect_is_simple_cond when SLP vectorizing.
3088 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
3091 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
3092 define __ibm128 as long double.
3093 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
3094 as a distinct type when IEEE 128-bit support is enabled.
3095 (init_float128_ieee): Fix up conversions between IFmode and IEEE
3096 128-bit types to use the correct functions.
3097 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
3098 convert between 128-bit floating point types that have different
3099 modes but the same representation, instead of using gen_lowpart to
3101 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
3103 (IFKF_reg): New attributes to give the register constraints for
3105 (extend<mode>tf2_internal): New insns to mark an explicit
3106 conversion between 128-bit floating point types that have a
3107 different mode but share the same representation.
3109 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
3111 PR tree-optimization/85814
3112 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
3113 a null return from get_strinfo when unsharing the next
3114 strinfo in the chain.
3116 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
3119 * varasm.c (weak_finish): Clean up weak_decls.
3121 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3123 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
3124 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
3125 UNSPEC_UADALP values.
3126 * config/aarch64/iterators.md (ABAL): New int iterator.
3129 (sur): Add mappings for the above.
3130 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
3132 (aarch64_<sur>abal<mode>_4): Likewise.
3133 (aarch64_<sur>adalp<mode>_3): Likewise.
3134 (<sur>sadv16qi): New define_expand.
3136 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
3138 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
3139 (*movdf_internal): Ditto.
3140 (*rcpsf2_sse): Ditto.
3141 (*rsqrtsf2_sse): Ditto.
3142 (*sqrt<mode>2_sse): Ditto.
3144 2018-05-21 Tamar Christina <tamar.christina@arm.com>
3146 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
3148 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
3149 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
3150 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
3151 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
3153 * config/aarch64/arm_neon.h: Likewise.
3154 * config/aarch64/iterators.md (VQ_I): New.
3156 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
3158 * config.gcc: Add arc/t-multilib-linux to tmake_file for
3160 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
3163 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
3165 * config/nds32/constraints.md (S): New constraint.
3166 * config/nds32/nds32.md (call_internal): Use constraint S.
3167 (call_value_internal): Likewise.
3168 (sibcall_internal): Likewise.
3169 (sibcall_value_internal): Likewise.
3171 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
3172 Chung-Ju Wu <jasonwucj@gmail.com>
3174 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
3177 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
3178 Chung-Ju Wu <jasonwucj@gmail.com>
3180 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
3181 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
3182 (nds32_rtx_costs_impl): Simplify.
3183 (nds32_address_cost_impl): Simplify.
3184 (nds32_init_rtx_costs): New function.
3185 (nds32_rtx_costs_speed_prefer): Likewise.
3186 (nds32_rtx_costs_size_prefer): Likewise.
3187 (nds32_address_cost_speed_prefer): Likewise.
3188 (nds32_address_cost_speed_fwprop): Likewise.
3189 (nds32_address_cost_size_prefer): Likewise.
3190 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
3191 * config/nds32/nds32.c (nds32_option_override): Use
3192 nds32_init_rtx_costs function.
3194 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
3196 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
3197 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
3198 (TARGET_PIPELINE_N8): Likewise.
3199 (TARGET_PIPELINE_N10): Likewise.
3200 (TARGET_PIPELINE_N13): Likewise.
3201 (TARGET_PIPELINE_GRAYWOLF): Likewise.
3203 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
3205 * config/nds32/nds32-fpu.md: Update copyright year.
3207 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3209 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
3211 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3213 * config/nds32/nds32.c
3214 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
3215 * config/nds32/nds32.opt (minline-asm-r15): New option.
3217 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3219 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
3221 * config/nds32/nds32.md (abssi2): New pattern.
3223 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
3225 * config/i386/i386.md (rex64namesuffix): New mode attribute.
3226 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
3227 Merge insn pattern from sse_cvtsi2ss<round_name> and
3228 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
3229 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
3230 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
3231 using SWI48 mode iterator.
3232 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
3233 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
3234 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
3235 pattern from sse_cvttss2si<round_saeonly_name>
3236 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
3237 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
3238 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
3239 using SWI48 mode iterator.
3240 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
3241 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
3242 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
3243 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
3244 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
3245 using SWI48 mode iterator.
3246 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
3247 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
3248 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
3249 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
3250 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
3251 SWI48 mode iterator.
3252 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
3253 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
3254 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
3255 pattern from sse_cvttsd2si<round_saeonly_name>
3256 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
3258 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3260 * config/nds32/nds32-md-auxiliary.c
3261 (nds32_valid_smw_lwm_base_p): Refine.
3262 (nds32_output_smw_single_word): Refine.
3263 (nds32_output_smw_double_word): New.
3264 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
3266 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3268 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
3269 (nds32_output_stack_pop): Refine.
3270 (nds32_expand_unaligned_load): Refine.
3271 (nds32_expand_unaligned_store): Refine.
3273 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3274 Chung-Ju Wu <jasonwucj@gmail.com>
3276 * config/nds32/constants.md: Add TP_REGNUM constant.
3277 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
3278 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
3280 * config/nds32/nds32-doubleword.md: Consider flag_pic.
3281 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
3282 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
3283 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
3284 and PIC code generation.
3285 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
3287 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
3289 * config/nds32/nds32.md: Support TLS and PIC.
3290 * config/nds32/nds32.c: Support TLS and PIC.
3291 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
3292 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
3295 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3297 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
3298 mode with E_ prefix.
3300 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3301 Chung-Ju Wu <jasonwucj@gmail.com>
3303 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
3304 * config/nds32/nds32-md-auxiliary.c
3305 (symbolic_reference_mentioned_p): New.
3306 (nds32_legitimize_ict_address): New.
3307 (nds32_expand_ict_move): New.
3308 (nds32_indirect_call_referenced_p): New.
3309 (nds32_symbol_binds_local_p): Delete.
3310 (nds32_long_call_p): Modify.
3311 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
3312 * config/nds32/nds32-protos.h
3313 (symbolic_reference_mentioned_p): Declare.
3314 (nds32_legitimize_ict_address): Declare.
3315 (nds32_expand_ict_move): Declare.
3316 (nds32_indirect_call_referenced_p): Declare.
3317 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
3318 (nds32_relax_group): Use nds32_ict_const_p as condition.
3319 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
3320 (nds32_asm_file_start): Output ict_model directive in asm code.
3321 (nds32_legitimate_address_p): Consider indirect call.
3322 (nds32_print_operand): Consider indirect call.
3323 (nds32_print_operand_address): Consider indirect call.
3324 (nds32_insert_attributes): Handle "indirect_call" attribute.
3325 (TARGET_LEGITIMATE_ADDRESS_P): Define.
3326 (TARGET_LEGITIMATE_CONSTANT_P): Define.
3327 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
3328 (TARGET_DELEGITIMIZE_ADDRESS): Define.
3329 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
3330 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
3331 (TARGET_ICT_MODEL_SMALL): Define.
3332 (TARGET_ICT_MODEL_LARGE): Define.
3333 * config/nds32/nds32.md (movsi): Consider ict model.
3334 (call, call_value): Consider ict model.
3335 (sibcall, sibcall_value): Consider ict model.
3336 * config/nds32/nds32.opt (mict-model): New option.
3337 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
3340 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
3341 Monk Chiang <sh.chiang04@gmail.com>
3342 Jim Wilson <jimw@sifive.com>
3344 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
3345 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
3346 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
3347 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
3348 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
3349 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
3350 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
3351 compute save_libcall_adjustment properly.
3352 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
3353 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
3354 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
3355 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
3356 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
3357 (ABI_SPEC): Handle mabi=ilp32e.
3358 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
3359 (RVE): Add RVE mask.
3360 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
3361 <-march>: Add rv32e as an example.
3363 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
3366 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
3367 (intra_create_variable_infos): Handle C++ constructors.
3369 2018-05-18 Martin Liska <mliska@suse.cz>
3371 * passes.def: Remove a redundant pass.
3373 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
3376 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
3378 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3380 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
3382 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
3384 (armv2, armv2a, armv3, armv3m): Delete architectures.
3385 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
3386 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
3387 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
3389 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
3390 (*mulsidi3adddi): Likewise.
3391 (mulsidi3): Likewise.
3392 (*mulsidi3_nov6): Likewise.
3393 (umulsidi3): Likewise.
3394 (umulsidi3_nov6): Likewise.
3395 (umaddsidi4): Likewise.
3396 (*umulsidi3adddi): Likewise.
3397 (smulsi3_highpart): Likewise.
3398 (*smulsi3_highpart_nov6): Likewise.
3399 (umulsi3_highpart): Likewise.
3400 (*umulsi3_highpart_nov6): Likewise.
3401 * config/arm/arm.h (arm_arch3m): Delete.
3402 * config/arm/arm.c (arm_arch3m): Delete.
3403 (arm_option_override_internal): Update armv3-related comment.
3404 (arm_configure_build_target): Delete use of isa_bit_mode32.
3405 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
3406 (arm_rtx_costs_internal): Delete check of arm_arch3m.
3407 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
3409 (mulusa3): Likewise.
3410 * config/arm/arm-protos.h (arm_arch3m): Delete.
3411 * config/arm/arm-tables.opt: Regenerate.
3412 * config/arm/arm-tune.md: Likewise.
3413 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
3414 deleted architectures.
3416 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3418 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
3419 (armv5t, armv5te): New features.
3420 (ARMv5, ARMv5e): Delete fgroups.
3421 (ARMv5t, ARMv5te): Adjust for above changes.
3423 (armv5, armv5e): Delete arches.
3424 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
3426 (*call_reg_arm): Likewise.
3427 (*call_value_reg_armv5): Likewise.
3428 (*call_value_reg_arm): Likewise.
3429 (*call_symbol): Likewise.
3430 (*call_value_symbol): Likewise.
3431 (*sibcall_insn): Likewise.
3432 (*sibcall_value_insn): Likewise.
3434 (prefetch): Likewise.
3435 (define_split and define_peephole2 dependent on arm_arch5):
3437 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
3439 (TARGET_ARM_QBIT): Likewise.
3440 (TARGET_DSP_MULTIPLY): Likewise.
3441 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
3442 (arm_arch5, arm_arch5e): Delete.
3443 (arm_arch5t, arm_arch5te): Declare.
3444 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
3445 (arm_arch5t): Declare.
3446 (arm_option_reconfigure_globals): Update for the above.
3447 (arm_options_perform_arch_sanity_checks): Update comment, replace
3448 use of arm_arch5 with arm_arch5t.
3449 (use_return_insn): Likewise.
3450 (arm_emit_call_insn): Likewise.
3451 (output_return_instruction): Likewise.
3452 (arm_final_prescan_insn): Likewise.
3453 (arm_coproc_builtin_available): Likewise.
3454 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
3455 arm_arch5e with arm_arch5t and arm_arch5te.
3456 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
3457 (arm_arch5t, arm_arch5te): Declare.
3458 * config/arm/arm-tables.opt: Regenerate.
3459 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
3460 * config/arm/t-multilib: Likewise.
3461 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
3462 instead of arm_arch5.
3463 (*call_reg_thumb1): Likewise.
3464 (*call_value_reg_thumb1_v5): Likewise.
3465 (*call_value_reg_thumb1): Likewise.
3466 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
3468 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
3470 2018-05-18 Martin Liska <mliska@suse.cz>
3472 PR gcov-profile/84846
3473 * doc/gcov.texi: Document -t option of gcov tool.
3475 2018-05-18 Martin Liska <mliska@suse.cz>
3477 PR gcov-profile/84846
3478 * gcov.c (print_usage): Add new -t option.
3479 (process_args): Handle the option.
3480 (generate_results): Use stdout as output when requested by
3483 2018-05-18 Martin Liska <mliska@suse.cz>
3485 PR gcov-profile/84846
3486 * coverage.c (coverage_init): Write PWD to .gcno file.
3487 * doc/gcov.texi: Document how working directory is printed.
3488 * gcov-dump.c (dump_gcov_file): Print PWD.
3489 * gcov.c (output_intermediate_file): Likewise.
3490 (read_graph_file): Read PWD string.
3491 (output_lines): Print PWD.
3493 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3496 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
3497 for retval and return false if all args to phi are zero.
3499 2018-05-18 Richard Biener <rguenther@suse.de>
3501 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
3503 (evrp_dom_walker::before_dom_children): Call it.
3505 2018-05-18 Richard Biener <rguenther@suse.de>
3507 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
3508 results when processing array refs with variable index.
3510 2018-05-18 Toon Moene <toon@moene.org>
3512 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
3513 directly after that of -floop-interchange. Indicate that both
3514 options are enabled by default when specifying -O3.
3516 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3518 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
3519 iterator. Delete separate integer-mode vec_set<mode> expander.
3520 (aarch64_simd_vec_setv2di): Delete.
3521 (vec_setv2di): Delete.
3522 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
3523 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
3524 the "w, r" alternative.
3526 2018-05-18 Martin Liska <mliska@suse.cz>
3528 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
3529 * tree-pass.h (make_pass_lower_switch_O0): New function.
3530 * tree-switch-conversion.c (node_has_low_bound): Remove.
3531 (node_has_high_bound): Likewise.
3532 (node_is_bounded): Likewise.
3533 (class pass_lower_switch): Make it a template type and create
3535 (pass_lower_switch::execute): Add template argument.
3536 (make_pass_lower_switch): New function.
3537 (make_pass_lower_switch_O0): New function.
3538 (do_jump_if_equal): Remove.
3539 (emit_case_nodes): Simplify to just handle all 3 cases and leave
3540 all the hard work to tree optimization passes.
3542 2018-05-18 Martin Liska <mliska@suse.cz>
3544 * dbgcnt.c (limit_low): Renamed from limit.
3545 (limit_high): New variable.
3546 (dbg_cnt_is_enabled): Check for upper limit.
3547 (dbg_cnt): Adjust dumping.
3548 (dbg_cnt_set_limit_by_index): Add new argument for high
3550 (dbg_cnt_set_limit_by_name): Likewise.
3551 (dbg_cnt_process_single_pair): Parse new format.
3552 (dbg_cnt_process_opt): Use strtok.
3553 (dbg_cnt_list_all_counters): Remove 'value' and add
3555 * doc/invoke.texi: Document changes.
3557 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
3559 * doc/sourcebuild.texi (scalar_all_fma): Document.
3560 * tree.def (FMA_EXPR): Delete.
3561 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
3562 * internal-fn.c (ternary_direct): New macro.
3563 (expand_ternary_optab_fn): Likewise.
3564 (direct_ternary_optab_supported_p): Likewise.
3565 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
3566 * builtins.c (fold_builtin_fma): Delete.
3567 (fold_builtin_3): Don't call it.
3568 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
3569 * expr.c (expand_expr_real_2): Likewise.
3570 * fold-const.c (operand_equal_p): Likewise.
3571 (fold_ternary_loc): Likewise.
3572 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
3573 * gimple.c (DEFTREECODE): Likewise.
3574 * gimplify.c (gimplify_expr): Likewise.
3575 * optabs-tree.c (optab_for_tree_code): Likewise.
3576 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
3577 * tree-eh.c (operation_could_trap_p): Likewise.
3578 (stmt_could_throw_1_p): Likewise.
3579 * tree-inline.c (estimate_operator_cost): Likewise.
3580 * tree-pretty-print.c (dump_generic_node): Likewise.
3581 (op_code_prio): Likewise.
3582 * tree-ssa-loop-im.c (stmt_cost): Likewise.
3583 * tree-ssa-operands.c (get_expr_operands): Likewise.
3584 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
3585 * fold-const-call.h (fold_fma): Delete.
3586 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
3587 CFN_FNMA and CFN_FNMS.
3589 * genmatch.c (combined_fn): New enum.
3590 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
3591 (commutative_op): New function.
3592 (commutate): Use it. Handle more than 2 operands.
3593 (dt_operand::gen_gimple_expr): Use commutative_op.
3594 (parser::parse_expr): Allow :c to be used with non-binary
3595 operators if the commutative operand is known.
3596 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
3597 CFN_FMS, CFN_FNMA and CFN_FNMS.
3598 (backprop::process_assign_use): Remove FMA_EXPR handling.
3599 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
3600 (gen_hsa_fma): New function.
3601 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
3602 IFN_FNMA and IFN_FNMS.
3603 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
3604 * gimple-fold.h (follow_all_ssa_edges): Declare.
3605 * gimple-fold.c (follow_all_ssa_edges): New function.
3606 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
3607 gimple_build interface and use follow_all_ssa_edges to fold the result.
3608 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
3609 instead of checking for optabs directly.
3610 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
3611 rather than FMA_EXPRs.
3612 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
3613 call to IFN_FMA instead of an FMA_EXPR.
3615 2018-05-17 Jim Wilson <jimw@sifive.com>
3617 * expr.c (do_tablejump): When converting index to Pmode, if we have a
3618 sign extended promoted subreg, and the range does not have the sign bit
3619 set, then do a sign extend.
3621 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
3622 test, check for sign extended subreg and/or constant operands, and
3623 do a sign extend in that case.
3625 2018-05-17 Steve Ellcey <sellcey@cavium.com>
3627 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
3628 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
3630 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
3631 Change logics_shift_reg to logics_shift_imm.
3632 (thunderx2t99_fp_loadpair_basic): Delete.
3633 (thunderx2t99_fp_storepair_basic): Delete.
3634 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
3635 (thunderx2t99_asimd_polynomial): Delete.
3636 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
3637 and neon_fp_mul_d_scalar_q.
3638 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
3639 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
3640 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
3641 (thunderx2t99_asimd_lut): Add missing tbl types.
3642 (thunderx2t99_asimd_ext): Delete.
3643 (thunderx2t99_asimd_load1_1_mult): Delete.
3644 (thunderx2t99_asimd_load1_2_mult): Delete.
3645 (thunderx2t99_asimd_load1_ldp): New.
3646 (thunderx2t99_asimd_load1): New.
3647 (thunderx2t99_asimd_load2): Add missing *load2* types.
3648 (thunderx2t99_asimd_load3): New.
3649 (thunderx2t99_asimd_load4): New.
3650 (thunderx2t99_asimd_store1_1_mult): Delete.
3651 (thunderx2t99_asimd_store1_2_mult): Delete.
3652 (thunderx2t99_asimd_store2_mult): Delete.
3653 (thunderx2t99_asimd_store2_onelane): Delete.
3654 (thunderx2t99_asimd_store_stp): New.
3655 (thunderx2t99_asimd_store1): New.
3656 (thunderx2t99_asimd_store2): New.
3657 (thunderx2t99_asimd_store3): New.
3658 (thunderx2t99_asimd_store4): New.
3660 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
3662 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
3663 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
3665 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
3666 Segher Boessenkool <segher@kernel.crashing.org>
3669 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
3672 2018-05-17 Richard Biener <rguenther@suse.de>
3674 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
3675 for pruning loop and prune defs feeding only already visited PHIs.
3677 2018-05-17 Richard Biener <rguenther@suse.de>
3679 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
3681 2018-05-17 Bin Cheng <bin.cheng@arm.com>
3682 Richard Biener <rguenther@suse.de>
3684 PR tree-optimization/85793
3685 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
3686 for VMAT_ELEMENTWISE.
3688 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
3690 * internal-fn.h (lookup_internal_fn): Declare
3691 * internal-fn.c (lookup_internal_fn): New function.
3692 * gimple.c (gimple_build_call_from_tree): Handle calls to
3694 * gimple-pretty-print.c (dump_gimple_call): Print "." before
3695 internal function names.
3696 * tree-pretty-print.c (dump_generic_node): Likewise.
3697 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
3699 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
3701 * gimple-fold.h (gimple_build): Make the function forms take
3702 combined_fn rather than built_in_function.
3703 (gimple_simplify): Likewise.
3704 * gimple-match-head.c (gimple_simplify): Likewise.
3705 * gimple-fold.c (gimple_build): Likewise.
3706 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
3707 rather than gimple_build_call_internal.
3708 (get_initial_defs_for_reduction): Likewise.
3709 (vect_create_epilog_for_reduction): Likewise.
3710 (vectorizable_live_operation): Likewise.
3712 2018-05-17 Martin Liska <mliska@suse.cz>
3714 * gimple-ssa-sprintf.c (format_directive): Do not use
3715 space in between 'G_' and '('.
3717 2018-05-17 Jakub Jelinek <jakub@redhat.com>
3720 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
3721 even if the mask is not all ones.
3724 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
3726 (ix86_gimple_fold_builtin): Likewise.
3729 * config/i386/i386.c: Include tree-vector-builder.h.
3730 (ix86_vector_shift_count): New function.
3731 (ix86_fold_builtin): Fold shift builtins by scalar count.
3732 (ix86_gimple_fold_builtin): Likewise.
3734 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
3735 _mm512_setzero): New intrinsics.
3737 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
3738 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3740 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
3741 code generation for cases where splatting a value is not useful.
3742 * simplify-rtx.c (simplify_ternary_operation): Simplify
3743 vec_merge across a vec_duplicate and a paradoxical subreg forming
3744 a vector mode to a vec_concat.
3746 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
3748 * config.gcc: Support "goldmont-plus".
3749 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
3751 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3752 PROCESSOR_GOLDMONT_PLUS.
3753 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
3754 (processor_target_table): Add "goldmont-plus".
3755 (PTA_GOLDMONT_PLUS): Define.
3756 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
3757 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
3758 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
3759 (fold_builtin_cpu): Add "goldmont-plus".
3760 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
3761 (ix86_option_override_internal): Add "goldmont-plus".
3762 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
3763 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
3764 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
3765 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
3767 2018-05-17 Richard Biener <rguenther@suse.de>
3769 PR tree-optimization/85757
3770 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
3771 remove defs that only feed that PHI from further processing.
3773 2018-05-16 Jim Wilson <jimw@sifive.com>
3775 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
3777 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
3779 2018-05-16 Mark Wielaard <mark@klomp.org>
3781 * dwarf2out.c (count_index_strings): New function.
3782 (output_indirect_strings): Call count_index_strings and generate
3783 header for dwarf_version >= 5.
3785 2018-05-16 Mark Wielaard <mark@klomp.org>
3787 * dwarf2out.c (dwarf_FORM): New function.
3788 (set_indirect_string): Use dwarf_FORM.
3789 (reset_indirect_string): Likewise.
3790 (size_of_die): Likewise.
3791 (value_format): Likewise.
3792 (output_die): Likewise.
3793 (add_skeleton_AT_string): Likewise.
3794 (output_macinfo_op): Likewise.
3795 (index_string): Likewise.
3796 (output_index_string_offset): Likewise.
3797 (output_index_string): Likewise.
3798 (count_index_strings): Likewise.
3800 2018-05-16 Carl Love <cel@us.ibm.com>
3802 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
3803 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
3805 2018-05-16 Martin Jambor <mjambor@suse.cz>
3807 * ipa-prop.c (ipa_free_all_edge_args): Remove.
3808 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
3810 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
3812 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
3813 (fnma<mode>4): Likewise.
3814 (fms<mode>4): Likewise.
3815 (fnms<mode>4): Likewise.
3816 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
3817 (aarch64_fnma<mode>4): Likewise.
3818 (aarch64_fms<mode>4): Likewise.
3819 (aarch64_fnms<mode>4): Likewise.
3820 (aarch64_fnmadd<mode>4): Likewise.
3822 2018-05-16 Jason Merrill <jason@redhat.com>
3824 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
3826 2018-05-16 Richard Biener <rguenther@suse.de>
3828 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
3829 (dump_stmt_cost): Declare.
3830 (add_stmt_cost): Dump cost we add.
3831 (add_stmt_costs): New function.
3832 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
3834 (vect_analyze_stmt): Adjust prototype.
3835 (vectorizable_condition): Likewise.
3836 (vectorizable_live_operation): Likewise.
3837 (vectorizable_reduction): Likewise.
3838 (vectorizable_induction): Likewise.
3839 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
3840 cost vector to pass to vectorizable_ and record afterwards.
3841 (vect_model_reduction_cost): Take cost vector argument and adjust.
3842 (vect_model_induction_cost): Likewise.
3843 (vectorizable_reduction): Likewise.
3844 (vectorizable_induction): Likewise.
3845 (vectorizable_live_operation): Likewise.
3846 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
3847 SLP_TREE_NUMBER_OF_VEC_STMTS.
3848 (vect_analyze_slp_cost_1): Remove.
3849 (vect_analyze_slp_cost): Likewise.
3850 (vect_slp_analyze_node_operations): Take visited args and
3851 a target cost vector. Avoid processing already visited stmt sets.
3852 (vect_slp_analyze_operations): Use a local cost vector to gather
3853 costs and register those of non-discarded instances.
3854 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
3855 (vect_schedule_slp_instance): Remove copying of
3856 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
3858 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
3859 adding cost. Record cost entry location.
3860 (vect_prologue_cost_for_slp_op): Function to compute cost of
3861 a constant or invariant generated for SLP vect in the prologue,
3862 split out from vect_analyze_slp_cost_1.
3863 (vect_model_simple_cost): Make static. Adjust for SLP costing.
3864 (vect_model_promotion_demotion_cost): Likewise.
3865 (vect_model_store_cost): Likewise, make static.
3866 (vect_model_load_cost): Likewise.
3867 (vectorizable_bswap): Add cost vector arg and adjust.
3868 (vectorizable_call): Likewise.
3869 (vectorizable_simd_clone_call): Likewise.
3870 (vectorizable_conversion): Likewise.
3871 (vectorizable_assignment): Likewise.
3872 (vectorizable_shift): Likewise.
3873 (vectorizable_operation): Likewise.
3874 (vectorizable_store): Likewise.
3875 (vectorizable_load): Likewise.
3876 (vectorizable_condition): Likewise.
3877 (vectorizable_comparison): Likewise.
3878 (can_vectorize_live_stmts): Likewise.
3879 (vect_analyze_stmt): Likewise.
3880 (vect_transform_stmt): Adjust calls to vectorizable_*.
3881 * tree-vectorizer.c: Include gimple-pretty-print.h.
3882 (dump_stmt_cost): New function.
3884 2018-05-16 Richard Biener <rguenther@suse.de>
3886 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
3887 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
3888 * tree-ssa-dse.c: Include tree-ssa-loop.h.
3889 (check_name): New callback.
3890 (dse_classify_store): Track cycles via a visited bitmap of PHI
3891 defs and simplify handling of in-loop and across loop dead stores
3892 and properly fail for loop-variant refs. Handle byte-tracking with
3893 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
3896 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
3898 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
3899 (vect_get_mask_type_for_stmt): Likewise.
3900 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
3902 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
3903 to determine the statement's vector type and the vector type that
3904 should be used for calculating nunits. Deal with cases in which
3905 the type has to be deferred.
3906 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
3907 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
3908 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
3909 (vect_determine_vf_for_stmt): New functions, split out from...
3910 (vect_determine_vectorization_factor): ...here.
3911 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
3912 (vect_get_mask_type_for_stmt): New functions, split out from
3913 vect_determine_vectorization_factor.
3915 2018-05-16 Richard Biener <rguenther@suse.de>
3917 * tree-cfg.c (verify_gimple_assign_ternary): Properly
3918 verify the [VEC_]COND_EXPR embedded comparison.
3920 2018-05-15 Martin Sebor <msebor@redhat.com>
3922 PR tree-optimization/85753
3923 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
3924 RECORD_TYPE in addition to ARRAY_TYPE.
3926 2018-05-15 Martin Sebor <msebor@redhat.com>
3929 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
3931 2018-05-15 Richard Biener <rguenther@suse.de>
3933 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
3934 add by_clobber_p one. Change algorithm to collect all defs
3935 representing uses we need to walk and try reducing them to
3936 a single one before failing.
3937 (dse_dom_walker::dse_optimize_stmt): Adjust.
3939 2018-05-13 Mark Wielaard <mark@klomp.org>
3941 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
3942 (size_of_loc_descr): Likewise.
3943 (output_loc_operands): Likewise.
3944 (output_loc_operands_raw): Likewise.
3945 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
3946 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
3947 (hash_loc_operands): Likewise.
3948 (compare_loc_operands): Likewise.
3950 2018-05-14 Mark Wielaard <mark@klomp.org>
3952 * dwarf2out.c (count_index_addrs): New function.
3953 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
3955 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3957 PR tree-optimization/83648
3958 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
3959 return value as malloc candidate.
3961 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3964 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
3965 param as true in call to suggest_attribute.
3967 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
3969 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
3970 -mreadonly-in-sdata.
3972 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3974 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
3976 (aarch64_crypto_aesd_fused): Likewise.
3978 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
3980 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
3981 (movsi_aarch64): Likewise.
3982 (load_pairsi): Likewise.
3983 (load_pairdi): Likewise.
3984 (store_pairsi): Likewise.
3985 (store_pairdi): Likewise.
3986 (load_pairsf): Likewise.
3987 (load_pairdf): Likewise.
3988 (store_pairsf): Likewise.
3989 (store_pairdf): Likewise.
3990 (zero_extend): Likewise.
3991 (trunc): Swap alternatives.
3992 (fcvt_target): Add '?' to prefer w over r.
3994 2018-05-14 Jakub Jelinek <jakub@redhat.com>
3997 * config/i386/i386.md: Disallow non-commutative arithmetics in
3998 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
3999 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
4000 in the peephole2 before it.
4002 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
4004 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
4005 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
4006 (ix86_handle_option): Handle -mcldemote.
4007 * config.gcc: New header.
4008 * config/i386/cldemoteintrin.h: New file.
4009 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
4010 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
4012 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4013 OPTION_MASK_ISA_CLDEMOTE.
4014 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
4015 (ix86_valid_target_attribute_inner_p): Ditto.
4016 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
4017 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
4018 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
4019 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
4020 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
4022 * config/i386/i386.opt: Add -mcldemote.
4023 * config/i386/x86intrin.h: New header.
4024 * doc/invoke.texi: Add -mcldemote.
4026 2018-05-14 Richard Biener <rguenther@suse.de>
4028 * doc/match-and-simplify.texi: Adjust :s documentation.
4030 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
4032 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
4033 intended memcpy size.
4034 (REORDER_45): Likewise.
4036 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
4038 * sort.cc: New file.
4039 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
4040 * vec.c (qsort_chk): Use gcc_qsort.
4041 * Makefile.in (OBJS-libcommon): Add sort.o.
4042 (build/sort.o): New target. Use it...
4043 (BUILD_RTL): ... here, and...
4044 (build/gencfn-macros): ... here, and...
4045 (build/genmatch): ... here.
4047 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4048 Chung-Ju Wu <jasonwucj@gmail.com>
4050 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
4051 * config/nds32/nds32-graywolf.md: New file.
4052 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
4053 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
4055 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
4056 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
4057 * config/nds32/nds32.md (pipeline_model): Add graywolf.
4058 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
4059 * config/nds32/pipelines.md: Include n15 settings.
4061 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4062 Chung-Ju Wu <jasonwucj@gmail.com>
4064 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
4065 * config/nds32/nds32-n13.md: New file.
4066 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
4067 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
4069 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
4070 * config/nds32/nds32.md (pipeline_model): Add n13.
4071 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
4072 * config/nds32/pipelines.md: Include n13 settings.
4074 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4075 Chung-Ju Wu <jasonwucj@gmail.com>
4077 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
4078 * config/nds32/nds32-n10.md: New file.
4079 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
4080 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
4082 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
4083 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
4084 * config/nds32/nds32.md (pipeline_model): Add n10.
4085 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
4086 * config/nds32/pipelines.md: Include n10 settings.
4088 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
4089 Kito Cheng <kito.cheng@gmail.com>
4090 Chung-Ju Wu <jasonwucj@gmail.com>
4092 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
4093 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4094 Add enum values for DSP extension instructions.
4095 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
4097 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
4098 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
4100 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
4101 * config/nds32/nds32-dspext.md: New file for DSP implementation.
4102 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
4103 * config/nds32/nds32-intrinsic.md: Likewise.
4104 * config/nds32/nds32_intrinsic.h: Likewise.
4105 * config/nds32/nds32-md-auxiliary.c: Likewise.
4106 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
4107 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
4108 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
4109 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
4110 * config/nds32/nds32-protos.h: New declarations for DSP extension.
4111 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
4112 TYPE_DMAC in switch statement.
4113 * config/nds32/nds32.c: New checking and implementation for DSP
4114 extension instructions.
4115 * config/nds32/nds32.h: Likewise.
4116 * config/nds32/nds32.md: Likewise.
4117 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
4118 * config/nds32/predicates.md: Implement new predicates for DSP
4121 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
4123 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
4124 Reformat alternatives and attributes so it is easier to identify
4125 which constraints/attributes go with which instruction.
4126 (mov<mode>_hardfloat32, FMOVE64): Likewise.
4127 (mov<mode>_softfloat32, FMOVE64): Likewise.
4128 (mov<mode>_hardfloat64, FMOVE64): Likewise.
4129 (mov<mode>_softfloat64, FMOVE64): Likewise.
4131 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
4133 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4135 (Basic PowerPC Built-in Functions): The new name of the
4136 subsection previously known as "PowerPC Built-in Functions".
4137 (Basic PowerPC Built-in Functions Available on all Configurations):
4139 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
4140 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
4141 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
4142 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
4144 2018-05-11 Martin Jambor <mjambor@suse.cz>
4147 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
4150 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4153 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
4155 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
4157 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
4158 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
4159 (ix86_handle_option): Handle -mwaitpkg.
4160 * config.gcc: New header.
4161 * config/i386/cpuid.h (bit_WAITPKG): New bit.
4162 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
4163 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
4165 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4166 OPTION_MASK_ISA_WAITPKG.
4167 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
4168 (ix86_option_override_internal): Add PTA_WAITPKG.
4169 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
4170 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
4171 IX86_BUILTIN_TPAUSE.
4172 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
4173 __builtin_ia32_umwait and __builtin_ia32_tpause.
4174 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
4175 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
4176 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
4177 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
4178 UNSPECV_TPAUSE): New.
4179 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
4180 * config/i386/i386.opt: Add -mwaitpkg.
4181 * config/i386/waitpkgintrin.h: New file.
4182 * config/i386/x86intrin.h: New header.
4183 * doc/invoke.texi: Add -mwaitpkg.
4185 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4188 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
4190 (cortex-m0): Use armv6s-m isa.
4191 (cortex-m0plus): Likewise.
4192 (cortex-m1): Likewise.
4193 (cortex-m0.small-multiply): Likewise.
4194 (cortex-m0plus.small-multiply): Likewise.
4195 (cortex-m1.small-multiply): Likewise.
4197 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
4198 Jakub Jelinek <jakub@redhat.com>
4200 PR tree-optimization/85692
4201 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
4202 source permute as well.
4204 2018-05-11 Martin Liska <mliska@suse.cz>
4207 * doc/extend.texi: Document LLVM style format for no_sanitize
4210 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
4212 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
4213 mode_supports_vsx_dform_quad to mode_supports_dq_form.
4214 (mode_supports_vsx_dform_quad): Likewise.
4215 (mode_supports_vmx_dform): Move these functions to be next to the
4216 other mode_supports functions.
4217 (mode_supports_dq_form): Likewise.
4218 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
4219 mode_supports_dq_form.
4220 (reg_offset_addressing_ok_p): Likewise.
4221 (offsettable_ok_by_alignment): Likewise.
4222 (rs6000_legitimate_offset_address_p): Likewise.
4223 (legitimate_lo_sum_address_p): Likewise.
4224 (rs6000_legitimize_address): Likewise.
4225 (rs6000_legitimize_reload_address): Likewise.
4226 (rs6000_secondary_reload_inner): Likewise.
4227 (rs6000_preferred_reload_class): Likewise.
4228 (rs6000_output_move_128bit): Likewise.
4230 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
4232 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
4233 Generate SImode target register for null target.
4234 <case IX86_BUILTIN_XGETBV>: Ditto.
4235 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
4236 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
4238 2018-05-10 Carl Love <cel@us.ibm.com>
4240 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
4241 dcbtt and dcbtstt if operands[2] is 0.
4243 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
4246 * config/i386/sse.md (usadv64qi): New expander.
4248 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
4250 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
4251 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
4252 -maltivec=be support.
4253 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
4254 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
4255 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
4256 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
4257 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
4258 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
4259 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
4260 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
4261 altivec_vsumsws): Adjust.
4262 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
4263 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
4264 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
4266 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
4267 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
4268 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
4269 (altivec_lve<VI_char>x): Delete expand.
4270 (*altivec_lve<VI_char>x_internal): Rename to...
4271 (altivec_lve<VI_char>x): ... this.
4272 (altivec_lvxl_<mode>): Delete expand.
4273 (*altivec_lvxl_<mode>_internal): Rename to ...
4274 (altivec_lvxl_<mode>): ... this.
4275 (altivec_stvxl_<mode>): Delete expand.
4276 (*altivec_stvxl_<mode>_internal): Rename to ...
4277 (altivec_stvxl_<mode>): ... this.
4278 (altivec_stve<VI_char>x): Delete expand.
4279 (*altivec_stve<VI_char>x_internal): Rename to ...
4280 (altivec_stve<VI_char>x): ... this.
4281 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
4282 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
4283 reduc_plus_scal_<mode>): Adjust.
4284 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
4286 (rs6000_cpu_cpp_builtins): Adjust.
4287 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
4288 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
4289 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
4290 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
4291 -maltivec=be support.
4292 (rs6000_split_vec_extract_var): Adjust.
4293 (rs6000_split_v4si_init): Adjust.
4294 (swap_selector_for_mode): Delete.
4295 (altivec_expand_lvx_be, altivec_expand_stvx_be,
4296 altivec_expand_stvex_be): Delete.
4297 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
4298 -maltivec=be support.
4299 (rs6000_gimple_fold_builtin): Ditto.
4300 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
4302 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
4303 (TARGET_DIRECT_MOVE_64BIT): Adjust.
4304 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
4305 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
4306 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
4307 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
4308 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
4309 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
4310 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
4311 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
4312 anonymous split): Adjust.
4313 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
4314 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
4316 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
4318 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
4319 when --with-gxx-include-dir is also specified.
4320 * configure: Regenerate.
4322 2018-05-09 Jim Wilson <jimw@sifive.com>
4325 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
4326 * config/riscv/t-withmultilib: New.
4327 * config/riscv/withmultilib.h: New.
4328 * doc/install.texi: Document RISC-V --with-multilib-list support.
4330 2018-05-09 Richard Biener <rguenther@suse.de>
4332 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
4334 (vect_bb_vectorization_profitable_p): Adjust. Compute
4335 actual scalar cost using the cost vector and the add_stmt_cost
4338 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4340 PR rtl-optimization/85645
4341 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
4342 in the REG_CFA_REGISTER note for LR, don't leave it empty.
4344 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4346 PR rtl-optimization/85645
4347 * shrink-wrap.c (spread_components): Return a boolean saying if
4348 anything was changed.
4349 (try_shrink_wrapping_separate): Iterate spread_components until
4350 nothing changes anymore.
4352 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4354 PR rtl-optimization/85645
4355 * regrename.c (build_def_use): Also kill the chains that include the
4356 destination of a REG_CFA_REGISTER note.
4358 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4360 PR rtl-optimization/85645
4361 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
4362 insn that has a REG_CFA_REGISTER note.
4364 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
4366 * cfgexpand.c (expand_clobber): New function.
4367 (expand_gimple_stmt_1): Use it.
4368 * tree-vect-stmts.c (vect_clobber_variable): New function,
4370 (vectorizable_simd_clone_call): ...here.
4371 (vectorizable_store): Emit a clobber either side of an
4372 IFN_STORE_LANES sequence.
4373 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
4375 2018-05-09 Tom de Vries <tom@codesourcery.com>
4378 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
4379 (define_insn "trap_if_false"): Add exit after trap.
4381 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
4383 PR rtl-optimization/85638
4384 * bb-reorder.c: Include common/common-target.h.
4385 (create_forwarder_block): New function extracted from...
4386 (fix_up_crossing_landing_pad): ...here. Rename into...
4387 (dw2_fix_up_crossing_landing_pad): ...this.
4388 (sjlj_fix_up_crossing_landing_pad): New function.
4389 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
4390 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
4391 from both partitions and exit the loop after one iteration.
4393 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
4396 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4398 (Basic PowerPC Built-in Functions): The new name of the
4399 subsection previously known as "PowerPC Built-in Functions".
4400 (Basic PowerPC Built-in Functions Available on all Configurations):
4402 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
4404 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
4406 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
4408 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
4411 2018-05-08 Jim Wilson <jimw@sifive.com>
4413 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
4414 (LD_EMUL_SUFFIX): New.
4415 (LINK_SPEC): Use it.
4417 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
4419 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4421 (Basic PowerPC Built-in Functions): The new name of the
4422 subsection previously known as "PowerPC Built-in Functions".
4423 (Basic PowerPC Built-in Functions Available on all Configurations):
4425 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
4427 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
4429 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
4431 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
4434 2018-05-08 Jakub Jelinek <jakub@redhat.com>
4437 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
4438 after cmpelim optimization.
4440 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
4442 * config.gcc: Support "goldmont".
4443 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
4444 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4446 * config/i386/i386.c (m_GOLDMONT): Define.
4447 (processor_target_table): Add "goldmont".
4448 (PTA_GOLDMONT): Define.
4449 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
4450 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
4451 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
4452 (fold_builtin_cpu): Add "goldmont".
4453 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
4454 (ix86_option_override_internal): Add "goldmont".
4455 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
4456 (processor_type): Add PROCESSOR_GOLDMONT.
4457 * config/i386/i386.md: Add CPU "glm".
4458 * config/i386/glm.md: New file.
4459 * config/i386/x86-tune.def: Add m_GOLDMONT.
4460 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
4462 2018-05-08 Jakub Jelinek <jakub@redhat.com>
4465 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
4467 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
4468 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
4469 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
4472 * config/i386/i386.c (ix86_fold_builtin): Handle
4473 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
4476 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
4477 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
4479 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
4482 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
4483 (check_arch): Likewise.
4484 (check_fpu): Return the result rather than printing it.
4485 (end arch): Fix operator precedence.
4486 (end cpu): Likewise.
4487 (END): Print the result from check_fpu.
4489 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
4490 Alan Hayward <alan.hayward@arm.com>
4491 David Sherwood <david.sherwood@arm.com>
4493 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
4494 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
4495 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
4496 (*fcmuo<mode>_and): New patterns.
4498 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
4500 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
4501 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
4502 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
4503 (cmp_op, sve_imm_con): New code attributes.
4504 (SVE_COND_INT_CMP, imm_con): Delete.
4505 (cmp_op): Remove above unspecs from int attribute.
4506 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
4508 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
4509 comparison-specific unspecs.
4510 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
4511 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
4512 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
4513 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
4514 (*vec_fcm<cmp_op><mode>): Rename to...
4515 (*fcm<cmp_op><mode>): ...this and adjust likewise.
4516 (*vec_fcmuo<mode>): Rename to...
4517 (*fcmuo<mode>): ...this and adjust likewise.
4518 (*pred_fcm<cmp_op><mode>): New pattern.
4519 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
4520 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
4522 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
4524 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
4525 (aarch64_emit_sve_predicated_cond): New function.
4526 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
4527 (aarch64_emit_unspec_cond_or): Replace with...
4528 (aarch64_emit_sve_or_conds): ...this new function. Use
4529 aarch64_emit_sve_ptrue_op for the individual comparisons and
4530 aarch64_emit_binop to OR them together.
4531 (aarch64_emit_inverted_unspec_cond): Replace with...
4532 (aarch64_emit_sve_inverted_cond): ...this new function. Use
4533 aarch64_emit_sve_ptrue_op for the comparison and
4534 aarch64_emit_unop to invert the result.
4535 (aarch64_expand_sve_vec_cmp_float): Update after the above
4536 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
4538 2018-05-07 Nathan Sidwell <nathan@acm.org>
4540 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
4541 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
4542 (Backwards Compatibility): Likewise.
4544 2018-05-07 Luis Machado <luis.machado@linaro.org>
4548 2018-05-07 Luis Machado <luis.machado@linaro.org>
4550 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4551 <prefetch_dynamic_strides>: New const bool field.
4552 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4553 prefetch_dynamic_strides.
4554 (exynosm1_prefetch_tune): Likewise.
4555 (thunderxt88_prefetch_tune): Likewise.
4556 (thunderx_prefetch_tune): Likewise.
4557 (thunderx2t99_prefetch_tune): Likewise.
4558 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
4560 (aarch64_override_options_internal): Update to set
4561 PARAM_PREFETCH_DYNAMIC_STRIDES.
4562 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4563 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4564 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4565 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4566 prefetch-dynamic-strides setting.
4568 2018-05-07 Luis Machado <luis.machado@linaro.org>
4570 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4571 <minimum_stride>: New const int field.
4572 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4573 minimum_stride field.
4574 (exynosm1_prefetch_tune): Likewise.
4575 (thunderxt88_prefetch_tune): Likewise.
4576 (thunderx_prefetch_tune): Likewise.
4577 (thunderx2t99_prefetch_tune): Likewise.
4578 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
4579 (aarch64_override_options_internal): Update to set
4580 PARAM_PREFETCH_MINIMUM_STRIDE.
4581 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4582 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4583 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4584 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4585 stride is constant and is below the minimum stride threshold.
4587 2018-05-07 Luis Machado <luis.machado@linaro.org>
4589 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
4592 2018-05-07 Luis Machado <luis.machado@linaro.org>
4594 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4595 <prefetch_dynamic_strides>: New const bool field.
4596 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4597 prefetch_dynamic_strides.
4598 (exynosm1_prefetch_tune): Likewise.
4599 (thunderxt88_prefetch_tune): Likewise.
4600 (thunderx_prefetch_tune): Likewise.
4601 (thunderx2t99_prefetch_tune): Likewise.
4602 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
4604 (aarch64_override_options_internal): Update to set
4605 PARAM_PREFETCH_DYNAMIC_STRIDES.
4606 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4607 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4608 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4609 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4610 prefetch-dynamic-strides setting.
4612 2018-05-07 Luis Machado <luis.machado@linaro.org>
4614 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4615 <minimum_stride>: New const int field.
4616 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4617 minimum_stride field.
4618 (exynosm1_prefetch_tune): Likewise.
4619 (thunderxt88_prefetch_tune): Likewise.
4620 (thunderx_prefetch_tune): Likewise.
4621 (thunderx2t99_prefetch_tune): Likewise.
4622 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
4623 (aarch64_override_options_internal): Update to set
4624 PARAM_PREFETCH_MINIMUM_STRIDE.
4625 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4626 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4627 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4628 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4629 stride is constant and is below the minimum stride threshold.
4631 2018-05-06 Jakub Jelinek <jakub@redhat.com>
4634 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
4635 the type is addressable. Don't force op into register if it has
4638 2018-05-05 Roland McGrath <mcgrathr@google.com>
4641 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
4642 any section for which we don't know a specific type it should have,
4643 regardless of name. Previously this was done only for the exact
4644 names ".init_array", ".fini_array", and ".preinit_array".
4645 (default_elf_asm_named_section): Add comment about
4646 relationship with default_section_type_flags and SECTION_NOTYPE.
4647 (get_section): Don't consider it a type conflict if one side has
4648 SECTION_NOTYPE and the other doesn't, as long as neither has the
4649 SECTION_BSS et al used in the default_section_type_flags logic.
4651 2018-05-05 Tom de Vries <tom@codesourcery.com>
4654 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
4655 (workaround_barsyncs): New function.
4656 (nvptx_reorg): Use workaround_barsyncs.
4657 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4658 (define_expand "nvptx_membar_cta"): New define_expand.
4659 (define_insn "*nvptx_membar_cta"): New insn.
4661 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
4663 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
4664 To improve optimization opportunities.
4665 * builtin-types.def: The new needed builtin types for the above.
4667 2018-05-04 Richard Biener <rguenther@suse.de>
4669 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
4670 * gimple-ssa-store-merging.c
4671 (imm_store_chain_info::output_merged_store): Remove redundant create,
4672 release split_store vector contents on failure.
4673 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
4674 scalar stmt vector on cache hit.
4676 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
4678 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
4680 * config.gcc (powerpc-xilinx-eabi*): Remove.
4681 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
4683 (fusion_addis_mem_combo_load): Ditto.
4684 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
4686 (rs6000_cpu_cpp_builtins): Ditto.
4687 * config/rs6000/rs6000-linux.c
4688 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
4689 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
4690 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
4692 (rs6000_setup_reg_addr_masks): Ditto.
4693 (rs6000_init_hard_regno_mode_ok): Ditto.
4694 (rs6000_option_override_internal): Ditto.
4695 (legitimate_lo_sum_address_p): Ditto.
4696 (rs6000_legitimize_address): Ditto.
4697 (rs6000_legitimize_reload_address): Ditto.
4698 (rs6000_legitimate_address_p): Ditto.
4699 (abi_v4_pass_in_fpr): Ditto.
4700 (setup_incoming_varargs): Ditto.
4701 (rs6000_gimplify_va_arg): Ditto.
4702 (rs6000_split_multireg_move): Ditto.
4703 (rs6000_savres_strategy): Ditto.
4704 (rs6000_emit_prologue_components): Ditto.
4705 (rs6000_emit_epilogue_components): Ditto.
4706 (rs6000_emit_prologue): Ditto.
4707 (rs6000_emit_epilogue): Ditto.
4708 (rs6000_elf_file_end): Ditto.
4709 (rs6000_function_value): Ditto.
4710 (rs6000_libcall_value): Ditto.
4711 * config/rs6000/rs6000.h: Ditto.
4712 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
4713 (TARGET_MINMAX): ... this. New.
4714 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
4715 * config/rs6000/rs6000.md: Remove Xilinx FP support.
4716 (*movsi_internal1_single): Delete.
4717 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
4718 mfpu=, mxilinx-fpu): Delete.
4719 * config/rs6000/singlefp.h: Delete.
4720 * config/rs6000/sysv4.h: Remove Xilinx FP support.
4721 * config/rs6000/t-rs6000: Ditto.
4722 * config/rs6000/t-xilinx: Delete.
4723 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
4724 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
4725 (VStype_simple): Delete.
4726 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
4727 * config/rs6000/xfpu.h: Delete.
4728 * config/rs6000/xfpu.md: Delete.
4729 * config/rs6000/xilinx.h: Delete.
4730 * config/rs6000/xilinx.opt: Delete.
4731 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
4732 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
4734 2018-05-04 Tom de Vries <tom@codesourcery.com>
4737 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
4740 2018-05-04 Richard Biener <rguenther@suse.de>
4743 * tree-complex.c (update_complex_assignment): We are always in SSA form.
4744 (expand_complex_div_wide): Likewise.
4745 (expand_complex_operations_1): Likewise.
4746 (expand_complex_libcall): Preserve EH info of the original stmt.
4747 (tree_lower_complex): Handle removed blocks.
4748 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
4749 on complex multiplication and division libcall builtins.
4751 2018-05-04 Richard Biener <rguenther@suse.de>
4754 * fold-const.c (negate_expr_p): Restrict negation of operand
4755 zero of a division to when we know that can happen without
4757 (fold_negate_expr_1): Likewise.
4759 2018-05-04 Jakub Jelinek <jakub@redhat.com>
4762 * real.h (real_nextafter): Declare.
4763 * real.c (real_nextafter): New function.
4764 * fold-const-call.c (fold_const_nextafter): New function.
4765 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
4766 CASE_CFN_NEXTTOWARD.
4767 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
4768 even when arg1_mode is different from arg0_mode.
4770 2018-05-03 Nathan Sidwell <nathan@acm.org>
4772 * doc/extend.texi (Deprecated Features): Remove
4774 (Backwards Compatibility): Likewise.
4775 * doc/invoke.texi (C++ Language Options): Likewise.
4776 (C++ Dialect Options): Likewise.
4778 2018-05-03 Jakub Jelinek <jakub@redhat.com>
4781 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
4782 _mm512_mask_mullox_epi64): New intrinsics.
4784 2018-05-03 Tom de Vries <tom@codesourcery.com>
4787 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
4788 dump files): Add offload-tree.
4790 2018-05-03 Richard Biener <rguenther@suse.de>
4792 PR tree-optimization/85615
4793 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
4794 to loops not nested in BBs loop father to avoid creating multi-entry
4797 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4799 PR tree-optimization/70291
4800 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
4801 arguments. Change return type to tree. Emit libcall as a new
4802 statement rather than replacing existing one when inplace_p is true.
4803 (expand_complex_multiplication_components): New function.
4804 (expand_complex_multiplication): Expand floating-point complex
4805 multiplication using the above.
4806 (expand_complex_division): Rename inner_type parameter to type.
4807 Update expand_complex_libcall call-site.
4808 (expand_complex_operations_1): Update expand_complex_multiplication
4809 and expand_complex_division call-sites.
4811 2018-05-02 Jakub Jelinek <jakub@redhat.com>
4814 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
4815 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
4816 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
4817 the highest significant bit of the shift count mask is clear. In
4818 check whether and[sq]i3 is needed verify that all significant bits
4819 of the shift count other than the highest are set.
4821 2018-05-02 Tom de Vries <tom@codesourcery.com>
4824 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
4825 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
4826 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
4827 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
4828 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
4829 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
4830 __builtin_goacc_parlevel_size.
4832 2018-05-02 Richard Biener <rguenther@suse.de>
4834 PR tree-optimization/85597
4835 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
4836 do not use split vect_get_vec_defs call but call vect_get_slp_defs
4839 2018-05-02 Tom de Vries <tom@codesourcery.com>
4842 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
4843 dump files): Add ltrans-tree.
4845 2018-05-02 Tom de Vries <tom@codesourcery.com>
4848 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
4849 dump files): Add wpa-ipa.
4851 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
4853 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
4854 powerpc*-*-linux*paired* target.
4855 * config/rs6000/750cl.h: Delete.
4856 * config/rs6000/paired.h: Delete.
4857 * config/rs6000/paired.md: Delete.
4858 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
4860 * config/rs6000/rs6000-builtin.def: Remove paired float support.
4861 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
4862 comment. Remove paired float support.
4863 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
4864 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
4866 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
4867 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
4869 * config/rs6000/rs6000.c: Remove paired float support.
4870 (paired_expand_vector_init, paired_expand_vector_move,
4871 paired_emit_vector_compare, paired_emit_vector_cond_expr,
4872 (paired_expand_lv_builtin, paired_expand_stv_builtin,
4873 paired_expand_builtin, paired_expand_predicate_builtin,
4874 paired_init_builtins): Delete.
4875 * config/rs6000/rs6000.h: Remove paired float support.
4876 * config/rs6000/rs6000.md: Remove paired float support.
4877 (move_from_CR_ov_bit): Delete.
4878 * config/rs6000/rs6000.opt (mpaired): Delete.
4879 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
4880 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
4882 2018-05-02 Richard Biener <rguenther@suse.de>
4885 * gimplify.c (gimplify_save_expr): When in SSA form allow
4886 SAVE_EXPRs to compute to SSA vars.
4888 2018-05-02 Jakub Jelinek <jakub@redhat.com>
4891 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
4892 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
4893 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
4894 clobber operands[2], instead use a new pseudo. Formatting fixes.
4896 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
4898 PR tree-optimization/85586
4899 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
4900 exit early for statements in the same group if the accesses are
4903 2018-05-02 Tom de Vries <tom@codesourcery.com>
4906 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
4909 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
4911 PR tree-optimization/85143
4912 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
4914 2018-05-01 Tom de Vries <tom@codesourcery.com>
4917 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
4918 not found" error message.
4920 2018-05-01 Tom de Vries <tom@codesourcery.com>
4923 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
4924 * vec.c (test_ordered_remove_if): New function.
4925 (vec_c_tests): Call test_ordered_remove_if.
4926 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
4927 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
4928 * tree-vect-patterns.c (vect_pattern_recog_1): Use
4929 VEC_ORDERED_REMOVE_IF.
4931 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4933 PR tree-optimization/82665
4934 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
4935 pointer subtraction where arguments come from a memchr call.
4937 2018-05-01 Jakub Jelinek <jakub@redhat.com>
4939 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
4940 --push-state --as-needed and --pop-state instead of --as-needed and
4941 --no-as-needed if ld supports it.
4942 * configure: Regenerated.
4945 * doc/install.texi2html: Replace _002d with - and _002a with * in
4946 generated html files using sed.
4948 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4951 * gcc-rich-location.c (blank_line_before_p): New function.
4952 (use_new_line): New function.
4953 (gcc_rich_location::add_fixit_insert_formatted): New function.
4954 * gcc-rich-location.h
4955 (gcc_rich_location::add_fixit_insert_formatted): New function.
4957 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4959 * selftest.c (assert_streq): Rename "expected" and "actual" to
4960 "val1" and "val2". Extend NULL-handling to cover both inputs
4961 symmetrically, while still requiring both to be non-NULL for a pass.
4962 * selftest.h (assert_streq): Rename "expected" and "actual" to
4964 (ASSERT_EQ): Likewise.
4965 (ASSERT_EQ_AT): Likewise.
4966 (ASSERT_KNOWN_EQ): Likewise.
4967 (ASSERT_KNOWN_EQ_AT): Likewise.
4968 (ASSERT_NE): Likewise.
4969 (ASSERT_MAYBE_NE): Likewise.
4970 (ASSERT_MAYBE_NE_AT): Likewise.
4971 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
4972 the assertion to pass.
4973 (ASSERT_STREQ_AT): Likewise.
4975 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
4977 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
4978 interaction with -pie.
4980 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4982 * selftest.h: Fix alphabetization of per-source-file selftest
4985 2018-04-30 Jason Merrill <jason@redhat.com>
4987 PR c++/61982 - dead stores to destroyed objects.
4988 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
4991 2018-04-30 Jason Merrill <jason@redhat.com>
4993 * tree.c (build_clobber): New.
4994 * tree.h: Declare it.
4995 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
4997 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4999 * diagnostic-show-locus.c (layout::layout): Update for
5000 location_get_source_line returning a char_span.
5001 (struct char_span): Move to input.h.
5002 (struct correction): Update for fields in char_span becoming
5004 (struct source_line): Update for location_get_source_line
5005 returning a char_span.
5006 (layout::print_line): Likewise.
5007 * edit-context.c (edited_file::print_content): Likewise.
5008 (edited_file::print_diff_hunk): Likewise.
5009 (edited_file::print_run_of_changed_lines): Likewise.
5010 (edited_file::get_num_lines): Likewise.
5011 (edited_line::edited_line): Likewise.
5012 * final.c (asm_show_source): Likewise.
5013 * input.c (location_get_source_line): Convert return type
5014 from const char * to char_span, losing the final "line_len"
5016 (dump_location_info): Update for the above.
5017 (get_substring_ranges_for_loc): Likewise. Use a char_span
5018 when handling the literal within the line.
5019 (test_reading_source_line): Update for location_get_source_line
5020 returning a char_span.
5021 * input.h (class char_span): Move here from
5022 diagnostic-show-locus.c, converting from a struct to a class.
5023 Make data members private.
5024 (char_span::operator bool): New.
5025 (char_span::length): New.
5026 (char_span::get_buffer): New.
5027 (char_span::operator[]): New.
5028 (char_span::subspan): Make const.
5029 (char_span::xstrdup): New.
5030 (location_get_source_line): Convert return type from const char *
5031 to char_span, losing the final "line_size" param.
5033 2018-04-30 Jan Hubicka <jh@suse.cz>
5035 * lto-wrapper.c (ltrans_priorities): New static var.
5036 (cmp_priority): New.
5037 (run_gcc): Read priorities and if doing parallel build order
5038 the Makefile by them.
5040 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5042 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
5044 2018-04-30 Richard Biener <rguenther@suse.de>
5046 * tree-cfg.c (verify_address): Remove base argument, add
5047 flag whether to check TREE_ADDRESSABLE and do that.
5048 (verify_expr): Remove.
5049 (verify_types_in_gimple_reference): Add pieces from verify_expr.
5050 (verify_gimple_assign_single): Likewise.
5051 (verify_gimple_switch): Likewise.
5052 (verify_expr_location_1): Dereference tp once. Add (disabled)
5053 piece from verify_expr.
5054 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
5056 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5058 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
5060 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5062 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
5063 (small_data_pattern): Likewise.
5064 (arc_rewrite_small_data): Likewise.
5065 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
5066 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
5067 (get_symbol_alignment): New function.
5068 (legitimate_small_data_address_p): Likewise.
5069 (legitimate_scaled_address): Update, call
5070 legitimate_small_data_address_p.
5071 (output_sdata): New static variable.
5072 (arc_print_operand): Update how we handle small data operands.
5073 (arc_print_operand_address): Likewise.
5074 (arc_legitimate_address_p): Update, use
5075 legitimate_small_data_address_p.
5076 (arc_rewrite_small_data_p): Remove.
5077 (arc_rewrite_small_data_1): Likewise.
5078 (arc_rewrite_small_data): Likewise.
5079 (small_data_pattern): Likewise.
5080 (compact_sda_memory_operand): Update to use
5081 legitimate_small_data_address_p and get_symbol_alignment.
5082 (prepare_move_operands): Don't rewite sdata pattern.
5083 (prepare_extend_operands): Remove.
5084 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
5086 (zero_extendqisi2): Likewise.
5087 (zero_extendhisi2): Likewise.
5088 (extendqihi2): Likewise.
5089 (extendqisi2): Likewise.
5090 (extendhisi2): Likewise.
5094 * config/arc/constraints.md (Usd): Change it to memory constraint.
5096 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5098 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
5099 as source of std instructions.
5100 * config/arc/arc.md (movsi_insn): Update pattern predicate to
5101 allow 6-bit constants as source for store instructions.
5102 (movdi_insn): Update instruction pattern to allow 6-bit constants
5103 as source for store instructions.
5105 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
5107 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
5109 2018-04-30 Nathan Sidwell <nathan@acm.org>
5110 Sandra Loosemore <sandra@codesourcery.com>
5112 * dumpfile.c (dump_open): Allow '-' for stdout.
5113 * doc/invoke.texi (Developer Options): Document dump filename
5114 determination early. Document stdin/stdout selection.
5116 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
5118 Microblaze Target: PIC data text relative
5120 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
5121 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
5123 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
5124 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
5125 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
5126 New addressing mode for data-text relative position indepenedent code.
5127 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
5128 'ADDRESS_SYMBOLIC_TXT_REL'.
5129 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
5130 (microblaze_legitimate_pic_operand): Exclude function calls from
5131 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
5132 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
5134 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
5135 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
5136 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
5137 for 'address + offset'.
5138 (microblaze_expand_prologue): Add new function prologue call for
5140 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
5141 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
5142 table in case of TARGET_PIC_DATA_TEXT_REL.
5143 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
5144 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
5145 Add new macros 'UNSPEC_TEXT',
5146 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
5147 + exclude function calls from 'UNSPEC_PLT' in case of data text
5149 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
5150 new target hook for generating address diff vector tables in case of
5152 * doc/tm.texi : Regenerate.
5153 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
5154 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
5155 of addr diff vector generation.
5156 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
5157 target hook definition.
5158 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
5159 Add default function for generate_pic_addr_diff_vec -> flag_pic.
5160 * doc/invoke.texi (Add new pic option): Add new microblaze pic
5161 option for data text relative.
5163 2018-04-30 Richard Biener <rguenther@suse.de>
5165 * tree-chrec.h (evolution_function_is_constant_p): Remove
5167 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
5169 2018-04-30 Richard Biener <rguenther@suse.de>
5172 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
5174 2018-04-30 Richard Biener <rguenther@suse.de>
5176 PR tree-optimization/28364
5177 PR tree-optimization/85275
5178 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
5179 copying first exit test.
5181 2018-04-28 Mark Wielaard <mark@klomp.org>
5183 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
5185 (dwarf_AT): Handle DW_AT_addr_base.
5186 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
5188 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
5191 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
5192 (*ashl<dwi>3_doubleword_mask_1): Ditto.
5193 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
5194 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
5196 2018-04-28 Richard Biener <rguenther@suse.de>
5198 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
5199 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
5200 to reflect use. Only add interesting stmts.
5202 2018-04-27 Martin Jambor <mjambor@suse.cz>
5205 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
5206 the jump function allows for passing through aggregate values.
5208 2018-04-27 David Malcolm <dmalcolm@redhat.com>
5210 * input.h (in_system_header_at): Convert from macro to inline
5212 (from_macro_expansion_at): Likewise.
5213 (from_macro_definition_at): Likewise.
5215 2018-04-27 Jeff Law <law@redhat.com>
5217 * config.gcc: Mark tile* targets as deprecated/obsolete.
5219 2018-04-27 Richard Biener <rguenther@suse.de>
5221 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
5224 2018-04-27 Richard Biener <rguenther@suse.de>
5226 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
5228 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
5230 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
5231 with Yd constraint. Set "preferred_for_speed" attribute from
5232 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
5234 (*movdi_internal): Ditto.
5235 (movti_interunit splitters): Remove
5236 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
5237 (movdi_interunit splitters): Ditto.
5238 * config/i386/constraints.md (Ye): Remove.
5239 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
5241 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5244 * config/aarch64/constraints.md (Usg): Limit to 31.
5247 2018-04-27 Jakub Jelinek <jakub@redhat.com>
5249 PR tree-optimization/85529
5250 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
5251 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
5252 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
5253 zero extension or masking of the MSB bit.
5254 (optimize_range_tests): Add FIRST_BB argument, pass it through
5255 to optimize_range_tests_var_bound.
5256 (maybe_optimize_range_tests, reassociate_bb): Adjust
5257 optimize_range_tests callers.
5259 2018-04-26 Richard Biener <rguenther@suse.de>
5260 Jakub Jelinek <jakub@redhat.com>
5262 * cgraph.h (symbol_table): Just declare debug method here.
5263 * symtab.c (symbol_table::debug): Define.
5265 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
5267 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
5269 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
5271 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
5272 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
5273 (*movdi_internal): Substitute Yi and Yj constraint with x
5274 and Ym and Yn constraint with y constraint. Update "isa"
5275 attribute and set "preferred_for_speed" attribute from
5276 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
5277 (*movsi_internal): Ditto.
5278 (*movdf_internal): Ditto.
5279 (*movsf_internal): Ditto.
5280 (*zero_extendsidi2): Ditto.
5281 * config/i386/sse.md (vec_set<mode>_0): Ditto.
5282 (sse2_loadld): Ditto.
5283 (*vec_extract<ssevecmodelower>_0): Ditto.
5284 (*vec_extractv4si_0_zext_sse4): Ditto.
5285 (vec_concatv2di): Ditto.
5286 (*vec_dup<mode>): Ditto.
5287 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
5288 * config/i386/constraints.md (Yi): Remove.
5293 2018-04-26 Nathan Sidwell <nathan@acm.org>
5295 * dumpfile.c (dump_open): New.
5296 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
5297 (dump_finish): Detect stdio/stderr by value not name.
5299 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
5301 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
5303 2018-04-26 Tom de Vries <tom@codesourcery.com>
5306 * config/nvptx/nvptx.c (verify_neutering_jumps)
5307 (verify_neutering_labels): New function
5308 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
5310 2018-04-26 Tom de Vries <tom@codesourcery.com>
5313 * config/nvptx/nvptx.c (needs_neutering_p): New function.
5314 (nvptx_single): Use needs_neutering_p to skip over insns that do not
5317 2018-04-26 Richard Biener <rguenther@suse.de>
5318 Tom de Vries <tom@codesourcery.com>
5321 * lto-streamer-out.c (output_function): Fixup loops if required to match
5322 discovery done in the reader.
5324 2018-04-26 Richard Biener <rguenther@suse.de>
5326 PR tree-optimization/85116
5327 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
5328 have a loop exit from the single latch predecessor. Remove
5329 case of header with just condition.
5330 (ch_base::copy_headers): Exclude infinite loops from any
5332 (pass_ch::execute): Record exits.
5334 2018-04-26 Richard Biener <rguenther@suse.de>
5336 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
5337 prologue cost vector and pass it to vect_get_load_cost.
5338 (vect_get_peeling_costs_all_drs): Likewise.
5339 (vect_peeling_hash_get_lowest_cost): Likewise.
5340 (vect_enhance_data_refs_alignment): Likewise.
5342 2018-04-26 Richard Biener <rguenther@suse.de>
5345 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
5346 checking of integer<->pointer conversions.
5347 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
5348 sign-/zero-extending pointer types.
5349 (expand_omp_for_static_chunk): Likewise.
5351 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
5352 Jean Lee <xiaoyur347@gmail.com>
5354 * config/mips/mips.c (mips_asan_shadow_offset): New function.
5355 (TARGET_ASAN_SHADOW_OFFSET): Define.
5356 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
5357 true for -fsanitize=address.
5359 2018-04-25 Mark Wielaard <mark@klomp.org>
5361 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
5364 2018-04-25 Jakub Jelinek <jakub@redhat.com>
5366 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
5367 than "alu", remove explicit "memory" and "imm_disp" attributes.
5368 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
5371 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
5372 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
5373 gen_lowpart_no_emit.
5375 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
5378 * config/i386/i386.c (ix86_expand_builtin): Change memory
5379 operand to XI, extend p0 to Pmode.
5380 * config/i386/i386.md: Change unspec volatile and operand
5381 1 mode to XI, change operand 0 mode to P.
5383 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5385 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
5386 GET_MODE_MASK before any checking.
5387 (nds32_can_use_bset_p): Likewise.
5388 (nds32_can_use_btgl_p): Likewise.
5390 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5392 * config/nds32/nds32-doubleword.md: New define_split pattern for
5393 illegal register number.
5395 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5397 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
5399 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5401 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
5403 2018-04-25 Richard Biener <rguenther@suse.de>
5405 * lto-streamer.h (LTO_major_version): Bump to 8.
5407 2018-04-25 Jakub Jelinek <jakub@redhat.com>
5409 * BASE-VER: Set to 9.0.0.
5411 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
5413 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
5414 in __abskf2 and __powikf2.
5416 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5419 * config/aarch64/constraints.md (Usg, Usj): New constraints.
5420 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
5421 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
5422 Use the above on operand 2. Reindent.
5423 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
5425 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
5428 * common/config/i386/i386-common.c (ix86_handle_option): Don't
5430 * config/i386/i386.opt (mcet): Removed.
5431 * doc/install.texi: Remove -mcet documentation.
5432 * doc/invoke.texi: Likewise.
5434 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
5437 * doc/install.texi: Remove -mcet from bootstrap-cet.
5439 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5442 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
5443 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
5447 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
5448 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
5449 containing a CONST_VECTOR.
5451 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
5453 * doc/install.texi: Update newlib dependency for nvptx.
5455 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5458 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
5459 instead of INTVAL when shifting x left.
5461 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
5463 PR tree-optimization/85478
5464 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
5465 vect_grouped_store_supported for single element vectors.
5467 2018-04-24 Richard Biener <rguenther@suse.de>
5470 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
5471 load cost increase to the case of non-constant step.
5473 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5476 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
5477 destination if any_malformed_asm.
5479 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
5482 * expr.c (store_field): In the bitfield case, if the value comes from
5483 a function call and is returned in registers by means of a PARALLEL,
5484 do not change the mode of the temporary unless BLKmode and VOIDmode.
5486 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
5488 PR rtl-optimization/85423
5489 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
5490 dependencies to debug insns when the previous insn is non-debug.
5492 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
5494 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
5495 enums into a single definition.
5496 (fls): Fix predicates and printing.
5499 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
5501 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
5502 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
5503 and short u6 immediate.
5504 (check_if_valid_sleep_operand): Remove.
5505 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
5507 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5509 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
5510 flag_always_save_lp condition.
5511 * config/nds32/nds32.opt (malways-save-lp): New option.
5513 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5515 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
5516 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
5517 * config/nds32/nds32.h
5518 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
5519 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
5521 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5523 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
5524 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
5526 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5527 Chung-Ju Wu <jasonwucj@gmail.com>
5529 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
5531 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
5532 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
5534 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5536 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
5538 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5540 * config/nds32/nds32-protos.h (nds32_data_alignment,
5541 nds32_local_alignment): Declare.
5542 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
5543 nds32_local_alignment): New functions.
5544 (TARGET_CONSTANT_ALIGNMENT): Define.
5545 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
5547 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5549 * config/nds32/nds32.c
5550 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
5551 (TARGET_MODES_TIEABLE_P): Likewise.
5553 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5555 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
5558 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
5559 Chung-Ju Wu <jasonwucj@gmail.com>
5561 * config/nds32/constants.md (unspec_volatile_element): Add enum values
5562 for unaligned access.
5563 * config/nds32/nds32-intrinsic.c: Implementation of expanding
5565 * config/nds32/nds32-intrinsic.md: Likewise.
5566 * config/nds32/nds32_intrinsic.h: Likewise.
5567 * config/nds32/nds32.h (nds32_builtins): Likewise.
5568 * config/nds32/nds32.opt (munaligned-access): New option.
5569 * config/nds32/nds32.c (nds32_asm_file_start): Display
5570 flag_unaligned_access status.
5572 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
5574 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
5575 -mno-relax is present.
5576 * config/riscv/linux.h (LINK_SPEC): Ditto.
5578 2018-04-20 Martin Sebor <msebor@redhat.com>
5581 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
5583 (gimple_fold_builtin_stxcpy_chk): Same.
5584 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
5586 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
5589 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
5590 __powikf2 when long double is IEEE 128-bit.
5592 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
5594 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
5595 step to make sure stack always aligned.
5597 2018-04-20 Carl Love <cel@us.ibm.com>
5600 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
5601 size check for arg0.
5603 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
5604 Tom de Vries <tom@codesourcery.com>
5607 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
5608 Emit insns for calls too.
5609 (nvptx_find_par): Always look for worker-level predecessor insn.
5610 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
5612 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
5613 (nvptx_process_pars): Propagate frames for calls.
5615 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
5618 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
5620 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
5621 (ix86_handle_option): Don't handle OPT_mibt.
5622 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
5624 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
5626 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
5627 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
5628 (ix86_target_macros): Define __CET__ with flag_cf_protection
5629 for -fcf-protection.
5630 * config/i386/i386.c (isa2_opts): Remove -mibt.
5631 * config/i386/i386.h (TARGET_IBT): Removed.
5632 (TARGET_IBT_P): Likewise.
5633 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
5634 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
5635 * config/i386/i386.opt (mcet): Update help message.
5638 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
5639 -mcet as an alias for -mshstk.
5641 2018-04-20 Richard Biener <rguenther@suse.de>
5644 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
5645 complexity by forcing a single use of the multiply operand.
5647 2018-04-20 Martin Jambor <mjambor@suse.cz>
5650 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
5651 recursion dependency to only apply to non-clones.
5653 2018-04-20 Martin Jambor <mjambor@suse.cz>
5656 * ipa-cp.c (create_specialized_node): Check that clones of
5657 self-recursive edges exist during IPA-CP.
5659 2018-04-19 Toon Moene <toon@moene.org>
5661 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
5664 2018-04-19 Jakub Jelinek <jakub@redhat.com>
5666 PR tree-optimization/85467
5667 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
5668 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
5669 VECTOR_CST element to type.
5671 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
5674 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
5675 * config/i386/i386.md (builtin_setjmp_setup): Removed.
5676 (builtin_longjmp): Likewise.
5677 (save_stack_nonlocal): New pattern.
5678 (restore_stack_nonlocal): Likewise.
5680 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
5683 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
5684 Replace ASM_OUTPUT_LABEL with fprintf.
5686 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
5689 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
5690 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
5691 * config/i386/i386-c.c (ix86_target_macros_internal): Also
5692 define __IBT__ and __SHSTK__ for -fcf-protection.
5693 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
5695 (ix86_trampoline_init): Likewise.
5696 (x86_output_mi_thunk): Likewise.
5697 (ix86_notrack_prefixed_insn_p): Likewise.
5698 (ix86_option_override_internal): Don't disallow -fcf-protection.
5699 * config/i386/i386.md (rdssp<mode>): Also enable for
5701 (incssp<mode>): Likewise.
5702 (nop_endbr): Likewise.
5703 * config/i386/i386.opt (mcet): Change help message to built-in
5707 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
5708 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
5709 enable CET built-in functions.
5711 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
5713 * common/config/i386/i386-common.c
5714 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
5715 OPTION_MASK_ISA_MOVDIRI_UNSET,
5716 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
5717 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
5718 * config.gcc (movdirintrin.h): New header.
5719 * config/i386/cpuid.h (bit_MOVDIRI,
5720 bit_MOVDIR64B): New bits.
5721 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
5723 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
5724 (VOID, PVOID, PCVOID)): New function types.
5725 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
5726 __builtin_ia32_directstoreu_u64,
5727 __builtin_ia32_movdir64b): New builtins.
5728 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
5729 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
5731 (ix86_valid_target_attribute_inner_p): Ditto.
5732 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
5733 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
5734 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
5735 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
5736 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
5737 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
5738 (movdiri<mode>, movdir64b_<mode>): New.
5739 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
5740 * config/i386/immintrin.h: Include movdirintrin.h.
5741 * config/i386/movdirintrin.h: New file.
5742 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
5744 2018-04-19 Richard Biener <rguenther@suse.de>
5747 * cfg.c (clear_bb_flags): When loop state says we have
5748 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
5750 2018-04-19 Richard Biener <rguenther@suse.de>
5752 PR tree-optimization/84737
5753 * tree-vect-data-refs.c (vect_copy_ref_info): New function
5754 copying restrict info.
5755 (vect_setup_realignment): Use it.
5756 * tree-vectorizer.h (vect_copy_ref_info): Declare.
5757 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
5758 the first DR to all generated stores.
5759 (vectorizable_load): Likewise for loads.
5761 2018-04-19 Jakub Jelinek <jakub@redhat.com>
5763 PR tree-optimization/85446
5764 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
5765 the integral and pointer types to have the same precision.
5767 * doc/install.texi: Document --disable-cet being the default and
5770 2018-04-18 Martin Liska <mliska@suse.cz>
5772 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
5775 2018-04-18 Martin Liska <mliska@suse.cz>
5778 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5781 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5782 arguments if they are comparable.
5784 2018-04-18 Martin Liska <mliska@suse.cz>
5787 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
5790 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
5793 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
5796 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
5797 ENDBR after calling __morestack.
5799 2018-04-18 David Malcolm <dmalcolm@redhat.com>
5802 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
5803 by using gcc_base_ver to generate a gcc_driver_version, and use
5804 it when generating GCC_DRIVER_NAME.
5805 * configure: Regenerate.
5807 2018-04-18 Jakub Jelinek <jakub@redhat.com>
5810 * config.gcc: Obsolete powerpc*-*-*spe*.
5812 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5815 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
5816 (stabstr_D): Change type of unum from unsigned int to
5817 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
5820 2018-04-17 Jim Wilson <jimw@sifive.com>
5823 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
5824 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
5825 Set arg_pointer_offset after using pretend_args_size.
5827 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5829 PR rtl-optimization/85431
5830 * dse.c (record_store): Ignore zero width stores.
5833 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
5834 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
5835 __builtin_stack_restore rather than after it.
5836 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
5837 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
5838 argument instead of virtual_dynamic_stack_rtx.
5840 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
5842 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
5844 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
5845 Add note to error message to explain internal mapping of overloaded
5846 built-in function name to non-overloaded built-in function name.
5847 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
5850 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
5853 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
5854 where the inputs overlap with the output.
5856 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5858 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
5859 (=v, v) alternative and explicit "memory" attribute.
5860 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
5861 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5863 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
5864 "sselog1" type instead of "sselog".
5865 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
5866 "sselog". Remove explicit "memory" attribute.
5867 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
5868 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5870 (vec_extract_hi_v32hi): Merge all alternatives into one, use
5871 "sselog1" type instead of "sselog". Remove explicit "memory"
5873 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
5874 use "sselog1" type instead of "sselog". Remove explicit "memory"
5876 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
5877 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5879 (vec_extract_hi_v64qi): Merge all alternatives into one, use
5880 "sselog1" type instead of "sselog". Remove explicit "memory"
5882 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
5883 use "sselog1" type instead of "sselog". Remove explicit "memory"
5887 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
5890 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
5893 2018-04-17 Martin Jambor <mjambor@suse.cz>
5896 * ipa-cp.c (create_specialized_node): Call
5897 expand_all_artificial_thunks if necessary.
5899 2018-04-17 Martin Liska <mliska@suse.cz>
5902 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
5903 in message, remote space in between '_G' and '('.
5905 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5908 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
5909 avx512f_vmcmp<mode>3<round_saeonly_name>,
5910 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
5911 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
5912 avx512f_rndscale<mode><round_saeonly_name>,
5913 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
5914 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
5915 Use %<iptr>2 instead of %2 for -masm=intel.
5916 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
5917 avx512f_vcvttss2usi<round_saeonly_name>,
5918 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
5920 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
5921 avx512f_vcvttsd2usi<round_saeonly_name>,
5922 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
5923 Use %q1 instead of %1 for -masm=intel.
5924 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
5925 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
5926 of %3 for -masm=intel.
5927 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
5929 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
5931 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
5933 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
5934 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
5936 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
5938 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
5939 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
5940 %g1 and one with %0 and %1.
5941 (avx512er_vmrcp28<mode><round_saeonly_name>,
5942 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
5944 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
5945 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
5946 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
5947 of %0 and %{%4%} for -masm=intel.
5948 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
5949 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
5950 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
5951 order of %0 and %{%5%}%{z%} for -masm=intel.
5953 2018-04-17 Jan Hubicka <jh@suse.cz>
5956 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
5958 2018-04-17 Martin Liska <mliska@suse.cz>
5961 * multiple_target.c (create_dispatcher_calls): Set apostrophes
5962 for target_clone error message. Make default implementation
5963 clone to be a local declaration.
5964 (separate_attrs): Add new argument and check for an empty
5966 (expand_target_clones): Handle it.
5967 (ipa_target_clone): Make redirection just for target_clones
5970 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
5971 Tom de Vries <tom@codesourcery.com>
5974 * omp-expand.c (expand_oacc_for): Add dummy false branch for
5975 tiled basic blocks without omp continue statements.
5977 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
5980 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
5981 vec_extract expression as having side effects to make sure it gets
5984 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
5987 * config/i386/i386.c (get_builtin_code_for_version): Check
5990 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
5993 * gcc/config.gcc: Support "skylake".
5994 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
5996 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
5997 (processor_target_table): Add "skylake".
5998 (ix86_option_override_internal): Add "skylake".
5999 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
6000 PROCESSOR_CANNONLAKE.
6001 (get_builtin_code_for_version): Fix priority for
6002 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
6003 PROCESSOR_SKYLAKE-AVX512.
6004 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
6005 (processor_type): Add PROCESSOR_SKYLAKE.
6007 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
6008 Jason Merrill <jason@redhat.com>
6011 * convert.c (convert_to_integer_1): Use direct recursion for
6012 enumeral types and types with a precision less than the number
6013 of bits in their mode.
6015 2018-04-16 Julia Koval <julia.koval@intel.com>
6018 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
6019 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
6021 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
6024 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
6025 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
6026 and -mno-direct-move.
6028 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
6031 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
6032 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
6033 Ensure negative shifts result in {0}.
6035 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
6037 PR rtl-optimization/79916
6038 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
6039 regs (if any) to define how to gnerate SD moves when LRA is in
6042 2018-04-13 Jakub Jelinek <jakub@redhat.com>
6044 PR rtl-optimization/85393
6045 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
6046 * except.c (expand_dw2_landing_pad_for_region): Make static.
6047 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
6048 a label and unconditional jump to old_bb, rather than
6049 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
6052 PR rtl-optimization/85376
6053 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
6054 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
6055 instead of a specific value.
6057 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
6058 Bin Cheng <bin.cheng@arm.com>
6060 PR tree-optimization/82965
6061 PR tree-optimization/83991
6062 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
6063 by_profile_only parameter.
6064 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
6065 information if the loop was predicted to iterate too many times.
6066 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
6068 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
6071 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
6074 2018-04-13 Martin Liska <mliska@suse.cz>
6075 Jakub Jelinek <jakub@redhat.com>
6078 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
6079 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
6080 * builtins.c (expand_builtin_memory_copy_args): Use
6081 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
6082 handle dest_addr == pc_rtx.
6084 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
6087 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
6088 asked to not generate direct moves.
6089 (fix_trunc<mode>si2_stfiwx): Similar.
6090 (fix_trunc<mode>si2_internal): Similar.
6092 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6095 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
6096 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
6097 lookup if dest in some wider mode is known to be const0_rtx and
6098 if so, record permanent equivalence for it to be ZERO_EXTEND of
6099 the narrower mode destination.
6101 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
6103 * lto-streamer-out.c (output_function): Revert 259346.
6104 * omp-expand.c (expand_oacc_for): Likewise.
6106 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
6108 PR rtl-optimization/85354
6109 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
6110 * sel-sched.c (sel_global_init): ... here.
6112 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
6115 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
6116 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
6117 mode for PE-COFF targets.
6118 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
6119 (i386_pe_asm_lto_end): Likewise.
6120 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
6121 (TARGET_ASM_LTO_END): Likewise.
6122 * config/i386/winnt.c (saved_debug_info_level): New static variable.
6123 (i386_pe_asm_lto_start): New function.
6124 (i386_pe_asm_lto_end): Likewise.
6126 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
6127 Richard Biener <rguenther@suse.de>
6130 * lto-streamer-out.c (output_function): Fix CFG loop state before
6132 * omp-expand.c (expand_oacc_for): Handle calls to internal
6133 functions like regular functions.
6135 2018-04-12 Richard Biener <rguenther@suse.de>
6138 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
6139 for the early LTO debug to properly generate references to it
6140 during DIE emission. Do not re-use that for the skeleton for
6142 (dwarf2out_early_finish): Likewise.
6144 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6147 * config/i386/sse.md
6148 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
6149 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
6150 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
6151 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
6152 and output is a reg, avoid creating invalid lowpart subreg, but
6153 instead split into a 512-bit move. Don't split if not AVX512VL,
6154 input is xmm16+ reg and output is a mem.
6155 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
6156 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
6157 xmm16+ reg and output is a mem.
6159 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6161 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
6162 also for flag_dwarf2_cfi_asm.
6164 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6166 PR rtl-optimization/85342
6167 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
6168 a bool scalar var inside of the loop instead. Don't try to update
6169 recog_data.operand after failed apply_change_group.
6171 2018-04-12 Tom de Vries <tom@codesourcery.com>
6174 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
6175 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
6176 array with flexible array member as array without given dimension.
6177 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
6178 argument for undefined param to true.
6180 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
6183 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
6184 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
6185 from PowerPC section.
6186 * config/rs6000/sysv4.opt (mcall-): Improve help text.
6187 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
6188 help text that is too long.
6189 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
6190 help text that is too long.
6191 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
6192 help text that is too long.
6194 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
6196 * config/alpha/alpha.md (stack_probe_internal): Rename
6197 from "probe_stack". Update all callers.
6199 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6201 PR rtl-optimization/84566
6202 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
6203 sched_macro_fuse_insns.
6205 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6208 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
6209 (compute_block_dependences): ... from here.
6211 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6213 PR tree-optimization/85331
6214 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
6215 from int to HOST_WIDE_INT.
6217 2018-04-11 Martin Jambor <mjambor@suse.cz>
6220 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
6221 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
6222 not the same as the source val.
6223 (cgraph_edge_brings_value_p): New parameter.
6224 (gather_edges_for_value): Pass destination value to
6225 cgraph_edge_brings_value_p.
6226 (perhaps_add_new_callers): Likewise.
6227 (get_info_about_necessary_edges): Likewise and exclude values brought
6228 only by self-recursive edges.
6229 (create_specialized_node): Redirect only clones of self-calling edges.
6230 (+self_recursive_pass_through_p): New function.
6231 (find_more_scalar_values_for_callers_subset): Use it.
6232 (find_aggregate_values_for_callers_subset): Likewise.
6233 (known_aggs_to_agg_replacement_list): Removed.
6234 (decide_whether_version_node): Re-calculate known constants for all
6235 remaining context clones.
6237 2018-04-11 Richard Biener <rguenther@suse.de>
6240 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
6241 from early DWARF output.
6242 (dwarf2out_early_finish): Output line info unconditionally into
6243 early DWARF and add reference to it.
6245 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6248 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
6249 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
6250 other than V2DFmode using iptr mode attribute.
6251 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
6253 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6255 PR rtl-optimization/84659
6256 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
6258 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6261 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
6263 (output_loc_list): Pass address of a dummy size variable even in the
6264 locview handling loop.
6265 (index_location_lists): Add comment on why skip_loc_list_entry can't
6268 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
6271 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
6274 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
6277 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
6278 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
6279 and -mstring-compare-inline-limit.
6281 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6284 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
6285 for stack clash protection in a register whenever we need it to be in
6288 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6290 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
6291 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
6293 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6296 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
6298 (mlong-double-): Ditto.
6299 * config/rs6000/sysv4.opt (msdata=): Ditto.
6300 (mtls-size=): Ditto.
6302 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
6304 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
6305 erroneous entries for
6306 "vector int vec_ldl (int, long int *)", and
6307 "vector unsigned int vec_ldl (int, unsigned long int *)".
6308 Add comments and entries for
6309 "vector bool char vec_ldl (int, bool char *)",
6310 "vector bool short vec_ldl (int, bool short *)",
6311 "vector bool int vec_ldl (int, bool int *)",
6312 "vector bool long long vec_ldl (int, bool long long *)",
6313 "vector pixel vec_ldl (int, pixel *)",
6314 "vector long long vec_ldl (int, long long *)",
6315 "vector unsigned long long vec_ldl (int, unsigned long long *)".
6316 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
6317 type tree bool_long_long_type_node and correct definition of
6318 bool_V2DI_type_node to make reference to this new type tree.
6319 (rs6000_mangle_type): Replace erroneous reference to
6320 bool_long_type_node with bool_long_long_type_node.
6321 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
6322 comments to emphasize sign distinctions for char and int types and
6323 replace RS6000_BTI_bool_long constant with
6324 RS6000_BTI_bool_long_long constant. Also add comment to restrict
6325 use of RS6000_BTI_pixel.
6326 (bool_long_type_node): Remove this macro definition.
6327 (bool_long_long_type_node): New macro definition
6329 2018-04-10 Jakub Jelinek <jakub@redhat.com>
6331 PR rtl-optimization/85300
6332 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
6333 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
6334 simplify_unary_operation fails.
6336 2018-04-10 Martin Liska <mliska@suse.cz>
6338 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
6339 cgraph_edge and ipa_ref.
6341 2018-04-10 Jakub Jelinek <jakub@redhat.com>
6345 * config/i386/sse.md
6346 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
6347 computation of the VEC_MERGE selector from mask.
6348 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
6349 Fix decoding of the VEC_MERGE selector into mask.
6351 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
6353 PR tree-optimization/85286
6354 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
6356 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
6358 * final.c (final_1): Set insn_last_address as well as
6359 insn_current_address.
6361 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6364 * explow.c (emit_stack_probe): Call validize_mem on memory location
6365 before passing it to gen_probe_stack. Create address operand and
6366 legitimize it for the probe_stack_address case.
6368 2018-04-09 Jan Hubicka <jh@suse.cz>
6371 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
6372 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
6373 * tree.c (free_lang_data_in_type): Fix handling of binfos;
6375 (free_lang_data): Rebuild type inheritance graph.
6377 2018-04-09 Martin Sebor <msebor@redhat.com>
6379 * invoke.texi (-finline-small-functions): Mention other optimization
6381 (-findirect-inlining, -fpartial-inlining): Same.
6382 (-finline-functions-called-once): Same.
6383 (-freorder-blocks-and-partition): Same.
6385 2018-04-09 Jan Hubicka <jh@suse.cz>
6388 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
6389 jumps; choose last target that matches the criteria (i.e.
6390 no partition changes for non-crossing jumps).
6391 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
6392 support for redirecting crossing jumps to non-crossing.
6394 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
6396 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
6397 also for naked functions.
6399 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
6401 * config/arc/arc.md (add_shift): New pattern.
6402 (add_shift2): Likewise.
6403 (sub_shift): Likewise.
6404 (sub_shift_cmp0_noout): Likewise.
6405 (compare_si_ashiftsi): Likewise.
6406 (xbfu_cmp0_noout): New combine pattern.
6407 (xbfu_cmp0"): Likewise.
6408 (movsi_set_cc_insn): Place the predicable variant first.
6409 (commutative_binary_cmp0_noout): Remove clobber.
6410 (commutative_binary_cmp0): New pattern.
6411 (noncommutative_binary_cmp0): Likewise.
6412 (noncommutative_binary_cmp0_noout): Likewise.
6413 (noncommutative_binary_comparison_result_used): Removed.
6414 (rsub_cmp0): New pattern.
6415 (rsub_cmp0_noout): Likewise.
6416 (extzvsi): Changed, keep only meaningful variants.
6417 (SQH, SEZ): New iterators.
6418 (SQH_postfix): New mode attribute.
6419 (SEZ_prefix): New code attribute.
6420 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
6421 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
6422 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
6424 (noncommutative_operator): Check the availability of barrel
6427 2018-04-09 Richard Biener <rguenther@suse.de>
6429 PR tree-optimization/85284
6430 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
6431 Only use the niter constraining form of simple_iv when the exit
6434 2018-04-09 Tom de Vries <tom@codesourcery.com>
6437 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
6438 (define_expand "*memory_barrier"): New define_expand.
6439 (define_insn "memory_barrier"): New insn.
6441 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6443 PR rtl-optimization/80463
6444 PR rtl-optimization/83972
6445 PR rtl-optimization/83480
6447 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
6448 correct producer for the insn.
6449 (tidy_control_flow): Fixup seqnos in case of debug insns.
6451 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6453 PR rtl-optimization/83913
6455 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
6456 different sched-times when merging exprs.
6458 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6460 PR rtl-optimization/83962
6462 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
6463 tidy_fallthru_edge and tidy_control_flow.
6465 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6467 PR rtl-optimization/83530
6469 * sel-sched.c (force_next_insn): New global variable.
6470 (remove_insn_for_debug): When force_next_insn is true, also leave only
6471 next insn in the ready list.
6472 (sel_sched_region): When the region wasn't scheduled, make another pass
6473 over it with force_next_insn set to 1.
6475 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
6477 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
6479 * config/nds32/constants.md (unspec_volatile_element): Add enum values
6480 for interrupt control.
6481 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
6482 functions for interrupt control.
6483 * config/nds32/nds32-intrinsic.md: Likewise.
6484 * config/nds32/nds32_intrinsic.h: Likewise.
6485 * config/nds32/nds32.h (nds32_builtins): Likewise.
6487 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
6489 * config/nds32/nds32.c (nds32_init_machine_status,
6490 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
6491 strict_aligned_p field.
6492 (nds32_expand_to_rtl_hook): New function.
6493 (TARGET_EXPAND_TO_RTL_HOOK): Define.
6494 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
6496 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6497 Chung-Ju Wu <jasonwucj@gmail.com>
6499 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
6500 * config/nds32/nds32-n7.md: New file.
6501 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
6502 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
6504 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
6505 * config/nds32/nds32.md (pipeline_model): Add n7.
6506 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
6507 * config/nds32/pipelines.md: Include n7 settings.
6509 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6510 Chung-Ju Wu <jasonwucj@gmail.com>
6512 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
6513 * config/nds32/nds32-e8.md: New file.
6514 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
6515 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
6517 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
6518 * config/nds32/nds32.md (pipeline_model): Add e8.
6519 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
6520 * config/nds32/pipelines.md: Include e8 settings.
6522 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6523 Chung-Ju Wu <jasonwucj@gmail.com>
6525 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
6526 * config/nds32/nds32-n8.md: New file.
6527 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
6528 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
6530 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
6531 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
6532 * config/nds32/nds32.md (pipeline_model): Add n8.
6533 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
6534 * config/nds32/pipelines.md: Include n8 settings.
6536 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6537 Chung-Ju Wu <jasonwucj@gmail.com>
6539 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
6540 * config/nds32/nds32-n9-2r1w.md: New file.
6541 * config/nds32/nds32-n9-3r2w.md: New file.
6542 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
6543 nds32_register_ports): New or modify for cpu n9.
6544 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
6546 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
6547 * config/nds32/nds32-utils.c: New file.
6548 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
6549 TARGET_MUL_SLOW): Define.
6550 * config/nds32/nds32.md (pipeline_model): New attribute.
6551 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
6552 New options that support cpu n9.
6553 * config/nds32/pipelines.md: Include n9 settings.
6554 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
6556 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
6558 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
6559 information if necessary.
6560 (output_cond_branch_compare_zero): Likewise.
6561 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
6562 (nds32_target_alignment): Refine for alignment.
6563 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
6564 (FUNCTION_BOUNDARY): Modify.
6565 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
6567 * config/nds32/nds32.opt (malways-align, malign-functions): New.
6569 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
6571 * config/nds32/constants.md (unspec_volatile_element): Add values for
6572 TLB operation and data prefetch.
6573 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
6574 functions for TLB operation and data prefetch.
6575 * config/nds32/nds32-intrinsic.md: Likewise.
6576 * config/nds32/nds32_intrinsic.h: Likewise.
6577 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
6578 (nds32_print_operand): Likewise.
6579 * config/nds32/nds32.h (nds32_builtins): Likewise.
6581 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
6582 Andrew Pinski <pinsika@gcc.gnu.org>
6585 * match.pd: Use constant_boolean_node of correct type instead of
6586 boolean_true_node or boolean_false_node for simplifying
6587 pointer comparisons to zero.
6589 2018-04-07 Jakub Jelinek <jakub@redhat.com>
6591 PR tree-optimization/80021
6592 * tree.c (verify_type_variant): Make error call in verify_variant_match
6593 translatable and remove final full stop.
6595 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6597 * config/nds32/constants.md (unspec_volatile_element): Add
6598 UNSPEC_VOLATILE_EH_RETURN.
6599 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
6600 nds32_output_stack_pop): Support dwarf exception handling process.
6601 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
6602 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
6603 exception handling process.
6604 (nds32_compute_stack_frame): Likewise.
6605 (nds32_return_addr_rtx): Likewise.
6606 (nds32_initial_elimination_offset): Likewise.
6607 (nds32_expand_prologue): Likewise.
6608 (nds32_expand_epilogue): Likewise.
6609 (nds32_dynamic_chain_address): New function.
6610 * config/nds32/nds32.h (machine_function): Add fields for dwarf
6612 (DYNAMIC_CHAIN_ADDRESS): Define.
6613 (EH_RETURN_DATA_REGNO): Define.
6614 (EH_RETURN_STACKADJ_RTX): Define.
6615 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
6616 patterns for dwarf exception handling.
6618 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6620 * config/nds32/nds32.h: Clean up obsolete macros.
6622 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6624 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6625 Add enum values for particular instructions.
6626 * config/nds32/nds32-intrinsic.c: Implementation of expanding
6627 particular intrinsic functions.
6628 * config/nds32/nds32-intrinsic.md: Likewise.
6629 * config/nds32/nds32_intrinsic.h: Likewise.
6630 * config/nds32/nds32.h (nds32_builtins): Likewise.
6631 * config/nds32/nds32.md (type): Add pbsad and pbsada.
6632 (btst, ave): New patterns for particular instructions.
6634 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6636 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6637 Add enum values for atomic load/store and memory sync.
6638 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
6640 * config/nds32/nds32-intrinsic.md: Likewise.
6641 * config/nds32/nds32_intrinsic.h: Likewise.
6642 * config/nds32/nds32.h (nds32_builtins): Likewise.
6644 2018-04-07 Jakub Jelinek <jakub@redhat.com>
6646 PR tree-optimization/85257
6647 * fold-const.c (native_encode_vector): If not all elts could fit
6648 and off is -1, return 0 rather than offset.
6649 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
6650 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
6651 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
6652 adjust buffer in native_interpret_expr call.
6654 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6656 * config/nds32/constants.md (unspec_volatile_element): Add cache
6657 control enum values.
6658 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
6659 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
6660 * config/nds32/nds32.c (nds32_cctl_names): New.
6661 (nds32_print_operand): Handle cache control register names.
6662 * config/nds32/nds32.h (nds32_builtins): New enum values.
6663 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
6665 * config/nds32/nds32.md (type): Add mmu.
6666 * config/nds32/pipelines.md (simple_insn): Add mmu.
6668 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6670 * config/nds32/nds32.md (type): Remove call.
6671 * config/nds32/pipelines.md (simple_insn): Likewise.
6673 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6675 * config/nds32/constants.md (unspec_volatile_element): Add
6676 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
6677 UNSPEC_VOLATILE_FMFCFG.
6678 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
6679 description for fmfcfg and fmfcsr.
6680 (bdesc_1arg): Add fmtcsr.
6681 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
6682 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
6683 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
6684 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
6685 unspec_fmfcfg): New patterns.
6686 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
6687 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
6688 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
6689 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
6690 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
6691 __nds32__fmfcfg): Define.
6693 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6695 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
6696 intrinsic register names.
6697 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
6698 intrinsic register enum values and macros.
6700 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6702 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
6703 for load/store addressing form.
6704 (nds32_print_operand_address): Likewise.
6706 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
6709 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
6710 based on LABEL_REF. Remove useless assertion.
6711 (pic_address_needs_scratch): Fix formatting.
6712 (sparc_legitimize_pic_address): Minor tweaks.
6713 (sparc_delegitimize_address): Adjust assertion accordingly.
6714 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
6715 into symbolic_operand.
6716 (movsi_high_pic_label_ref): Likewise.
6717 (movsi_lo_sum_pic_label_ref): Likewise.
6718 (movdi_pic_label_ref): Likewise.
6719 (movdi_high_pic_label_ref): Likewise.
6720 (movdi_lo_sum_pic_label_ref): Likewise.
6722 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
6724 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
6725 custom LIB_SPEC setup.
6727 2018-04-06 Ruslan Bukin <br@bsdpad.com>
6728 Kito Cheng <kito.cheng@gmail.com>
6730 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
6731 * config/riscv/freebsd.h: New.
6733 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6735 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
6736 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
6739 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6740 Kito Cheng <kito.cheng@gmail.com>
6742 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
6743 nds32_output_call, nds32_symbol_binds_local_p): New functions.
6744 * config/nds32/nds32-protos.h (nds32_output_call,
6745 nds32_output_return): Declare.
6746 * config/nds32/nds32.md: Refine all the call and return patterns.
6748 2018-04-06 Jakub Jelinek <jakub@redhat.com>
6751 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
6752 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
6754 PR rtl-optimization/84872
6755 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
6756 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
6759 2018-04-06 Tamar Christina <tamar.christina@arm.com>
6761 * expr.c (copy_blkmode_to_reg): Revert 254862.
6762 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
6764 2018-04-06 Richard Biener <rguenther@suse.de>
6767 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
6768 after seeing a component reference with an adjacent field. Treat
6769 refs to arrays at struct end of external decls similar to
6770 refs to unconstrained commons.
6772 2018-04-06 Jakub Jelinek <jakub@redhat.com>
6775 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
6776 look through SAVE_EXPRs with non-side-effects argument. Adjust
6778 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
6781 2018-04-06 Richard Biener <rguenther@suse.de>
6784 * alias.c (find_base_term): New wrapper around find_base_term
6785 unwinding CSELIB_VAL_PTR changes.
6786 (find_base_term): Do not restore CSELIB_VAL_PTR during the
6789 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6791 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
6793 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
6794 constant definitions.
6795 ("nop"): lr 0,0 -> nopr r0
6796 ("nop_lr0", "nop_lr1"): New insn definitions.
6798 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6800 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
6801 NDS32_V3PUSH_AVAILABLE_P macro.
6803 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
6804 Chung-Ju Wu <jasonwucj@gmail.com>
6806 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
6807 (nds32*-*-*): Add float and fpu_config into supported_defaults.
6808 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
6809 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
6810 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
6811 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
6812 * config/nds32/constraints.md: New constraints and checking for hard
6813 float configuration.
6814 * config/nds32/iterators.md: New mode iterator and attribute for hard
6815 float configuration.
6816 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
6818 * config/nds32/nds32-fpu.md: New file.
6819 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
6820 deal with hard float code generation.
6821 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
6823 (abi_type, float_reg_number): New enum type.
6824 * config/nds32/nds32-predicates.c: New predicates for hard float.
6825 * config/nds32/nds32-protos.h: Declare functions for hard float.
6826 * config/nds32/nds32.c: Implementation for hard float configuration.
6827 * config/nds32/nds32.h: Definitions for hard float configuration.
6828 * config/nds32/nds32.md: Include hard float machine description and
6829 modify patterns for hard float configuration.
6830 * config/nds32/nds32.opt: New options for hard float configuration.
6831 * config/nds32/predicates.md: New predicates for hard float
6834 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6836 * common/config/nds32/nds32-common.c
6837 (nds32_option_optimization_table): Enable -mreleax-hint by default.
6839 2018-04-05 Jakub Jelinek <jakub@redhat.com>
6842 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
6843 CONSTRUCTOR_ELT (ctor, ...)->value.
6845 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
6848 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
6850 2018-04-05 Tom de Vries <tom@codesourcery.com>
6853 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
6856 2018-04-05 Shiva Chen <shiva0217@gmail.com>
6857 Kito Cheng <kito.cheng@gmail.com>
6859 * config/nds32/constraints.md (U33): Fine-tune checking condition.
6860 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
6861 * config/nds32/nds32.h (nds32_16bit_address_type): Add
6862 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
6864 2018-04-05 Shiva Chen <shiva0217@gmail.com>
6865 Kito Cheng <kito.cheng@gmail.com>
6867 * config/nds32/constraints.md (Ufe): New memory constraint.
6868 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
6869 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
6870 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
6872 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
6873 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
6875 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6877 * config/nds32/nds32.md: Use optimize_size in the condition for
6878 alu-shift instructions.
6880 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6882 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
6884 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6886 * config/nds32/nds32.md (negsi2): Refine pattern.
6888 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
6889 Chung-Ju Wu <jasonwucj@gmail.com>
6891 * config/nds32/iterators.md (shift_rotate): New code iterator.
6892 (shift): New code attribute.
6893 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
6894 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
6895 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
6896 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
6897 bit-wise operations.
6898 (andsi3, *andsi3): Ditto.
6899 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
6900 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
6901 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
6902 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
6903 nds32_ior_operand, nds32_xor_operand): New predicates.
6905 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6907 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
6908 (addsi3, subsi3): ... this.
6910 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6912 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
6914 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6916 * config/nds32/nds32.md: Adjust indention.
6918 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
6920 * config/nds32/nds32.md (feature): New attribute.
6922 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6924 * config/nds32/nds32.md (subtype): New attribute.
6926 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
6929 * config/arm/arm-builtins.c (arm_expand_builtin): Change
6930 expansion to perform a bitwise AND of the argument followed by a
6931 boolean negation of the result.
6933 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
6935 PR rtl-optimization/84878
6936 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
6937 the basic block. Assert the use reference is not artificial and that
6938 it has an associated insn.
6940 2018-04-04 Michael Matz <matz@suse.de>
6942 * builtins.c (compute_objsize): Pass correct operand
6943 to array_at_struct_end_p.
6945 2018-04-04 Richard Biener <rguenther@suse.de>
6948 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
6949 from contexts for DINFO_LEVEL_TERSE and below.
6951 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
6953 * config/nds32/nds32-doubleword.md (move_<mode>): Require
6954 resiter_operand condition.
6955 * config/nds32/nds32.md (*move<mode>): Ditto.
6957 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
6958 Monk Chiang <sh.chiang04@gmail.com>
6960 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
6962 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6964 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
6966 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6967 Kito Cheng <kito.cheng@gmail.com>
6969 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
6970 nds32_cond_code_str, output_cond_branch,
6971 output_cond_branch_compare_zero, nds32_expand_cbranch,
6972 nds32_expand_cstore, nds32_expand_movcc,
6973 nds32_output_cbranchsi4_equality_zero,
6974 nds32_output_cbranchsi4_equality_reg,
6975 nds32_output_cbranchsi4_equality_reg_or_const_int,
6976 nds32_output_cbranchsi4_greater_less_zero: New functions.
6977 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
6978 nds32_expand_cstore, nds32_expand_movcc,
6979 nds32_output_cbranchsi4_equality_zero,
6980 nds32_output_cbranchsi4_equality_reg,
6981 nds32_output_cbranchsi4_equality_reg_or_const_int,
6982 nds32_output_cbranchsi4_greater_less_zero): Declare.
6983 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
6984 nds32_rimm11s_operand): New predicates.
6985 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
6986 * config/nds32/nds32.md: Rewrite all the branch and conditional move
6989 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
6991 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
6992 * config/nds32/nds32.md: Ditto.
6993 * config/nds32/pipelines.md: Ditto.
6995 2018-04-04 Richard Biener <rguenther@suse.de>
6997 PR tree-optimization/85168
6998 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
6999 propagating abnormals.
7001 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7003 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
7005 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7006 Kito Cheng <kito.cheng@gmail.com>
7008 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
7009 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
7010 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
7011 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
7012 * config/nds32/nds32.md (sibcall_internal): New.
7013 (sibcall_register): Remove.
7014 (sibcall_immediate): Remove.
7015 (sibcall_value_internal): New.
7016 (sibcall_value_register): Remove.
7017 (sibcall_value_immediate): Remove.
7018 * config/nds32/predicates.md (nds32_general_register_operand): New.
7019 (nds32_call_address_operand): New.
7021 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7023 PR rtl-optimization/85167
7024 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
7025 bb_defs if *split_p, instead preinitialize it to NULL.
7027 PR tree-optimization/85156
7028 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
7029 evaluating the argument multiple times.
7031 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
7033 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
7035 (_mm_cvtpd_ps): Likewise.
7036 (_mm_cvttpd_epi32): Likewise.
7037 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
7038 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
7039 vector, pixel, and bool following altivec.h include.
7041 2018-04-03 Martin Sebor <msebor@redhat.com>
7043 * doc/extend.texi (Common Function Attributes): Clarify.
7044 (const attribute): Likewise.
7045 (pure attribute): Likewise.
7047 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7050 * config/i386/i386.c (ix86_expand_vector_set): Use
7051 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
7053 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
7055 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
7056 instructions when changing rounding bits to preserve precision bits
7057 in the x87 control word.
7059 2018-04-03 Martin Liska <mliska@suse.cz>
7061 PR tree-optimization/82491
7062 * rtl.h (strip_offset_and_add): Replace += suboffset with
7063 poly_uint64 () + suboffset.
7065 2018-03-29 Martin Liska <mliska@suse.cz>
7066 Martin Jambor <mjambor@suse.cz>
7069 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
7070 param_type is not an integral or pointer type.
7072 2018-04-03 Richard Biener <rguenther@suse.de>
7074 * sese.h (recompute_all_dominators): Remove.
7076 2018-04-02 Martin Sebor <msebor@redhat.com>
7078 * doc/invoke.texi (-Wrestrict): Fix typos.
7080 2018-04-02 Jim Wilson <jimw@sifive.com>
7082 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
7083 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
7084 (<optab>di3, <optab>si3_extend): Likewise.
7085 (<optab>si3_mask, <optab>si3_mask_1): New.
7086 (<optab>di3_mask, <optab>di3_mask_1): New.
7087 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
7088 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
7089 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
7091 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
7093 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
7096 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
7098 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
7099 (nds32_canonicalize_comparison): New function.
7101 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7102 Kito Cheng <kito.cheng@gmail.com>
7103 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7105 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
7106 * config/nds32/constants.md (unspec_volatile_element): Add
7107 UNSPEC_VOLATILE_RELAX_GROUP.
7108 * config/nds32/nds32-relax-opt.c: New file.
7109 * config/nds32/nds32-predicates.c
7110 (nds32_symbol_load_store_p): New function.
7111 * config/nds32/nds32-protos.h
7112 (nds32_symbol_load_store_p): Declare function.
7113 (make_pass_nds32_relax_opt): Declare new rtl pass function.
7114 * config/nds32/nds32.c
7115 (nds32_register_pass): New function to register pass.
7116 (nds32_register_passes): New function to register passes.
7117 * config/nds32/nds32.md (relax_group): New pattern.
7118 * config/nds32/nds32.opt (mrelax-hint): New option.
7119 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
7121 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
7123 * config/nds32/t-nds32: Modify files dependency.
7125 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7127 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
7128 (PROFILE_HOOK): Define its implementation.
7130 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7132 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
7133 type and 32-bit size.
7135 2018-04-01 Jakub Jelinek <jakub@redhat.com>
7138 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
7139 (V_128_256): New mode iterator.
7140 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
7141 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
7142 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
7144 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
7145 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
7147 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
7150 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
7151 NaN inputs correctly.
7153 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
7156 * config/rs6000/vsx.md (??r): New mode attribute.
7157 (*vsx_mov<mode>_64bit): Use it.
7158 (*vsx_mov<mode>_32bit): Likewise.
7160 2018-03-30 Martin Sebor <msebor@redhat.com>
7162 PR tree-optimization/84818
7163 * builtins.c (check_access): Use warning_n.
7165 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7168 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
7170 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
7173 2018-03-30 Julia Koval <julia.koval@intel.com>
7176 * x86-tune.def (movx, partial_reg_dependency): Enable for
7179 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
7182 * lra-constraints.c (process_alt_operands): Move setting
7183 this_alternative_matches below.
7185 2018-03-29 Martin Liska <mliska@suse.cz>
7188 * doc/invoke.texi: Document how LTO works with debug info.
7189 Describe auto-load support of binutils. Mention 'x86-64'
7190 as valid option value of -march option.
7192 2018-03-29 Jakub Jelinek <jakub@redhat.com>
7194 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
7197 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
7198 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
7199 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
7202 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
7205 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
7206 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
7207 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
7208 for RS6000_BTM_POWERPC64.
7209 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
7210 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
7211 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
7216 2018-03-28 Carl Love <cel@us.ibm.com>
7219 2017-09-27 Carl Love <cel@us.ibm.com>
7221 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
7222 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
7223 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
7226 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7228 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
7229 instead of __vector bool.
7230 (_mm_max_pu8): Likewise.
7231 (_mm_min_pi16): Likewise.
7233 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
7236 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
7237 (DIVWEUO): Likewise.
7239 (DIVDEUO): Likewise.
7240 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
7241 DIVWEUO and DIVDEUO.
7242 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
7243 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
7244 (div_extend): Likewise.
7245 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
7247 (__builtin_divweuo): Likewise.
7248 (__builtin_divdeo): Likewise.
7249 (__builtin_divdeuo): Likewise.
7251 2018-03-28 Jakub Jelinek <jakub@redhat.com>
7254 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
7255 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
7257 PR tree-optimization/82004
7258 * gimple-match-head.c (optimize_pow_to_exp): New function.
7259 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
7260 Don't fold to exp if optimize_pow_to_exp is false.
7262 2018-03-28 Martin Liska <mliska@suse.cz>
7265 * calls.c (initialize_argument_information): Fix trailing space.
7266 * common.opt: Fix typo and provide better explanation for
7267 -fsanitize-coverage option.
7268 * config/i386/i386.opt: Fix typo.
7270 2018-03-28 Jakub Jelinek <jakub@redhat.com>
7271 Martin Liska <mliska@suse.cz>
7274 * gimplify.c (asan_poison_variable): Don't do the check for
7275 gimplify_omp_ctxp here.
7276 (gimplify_decl_expr): Do it here.
7277 (gimplify_target_expr): Likewise.
7279 2018-03-28 Martin Liska <mliska@suse.cz>
7282 * config/i386/i386.c (ix86_function_arg_advance): Do not call
7283 chkp_type_bounds_count if MPX is not enabled.
7285 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
7287 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
7289 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
7292 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
7293 function to create the function decl for complex long double
7294 multiply and divide for -mabi=ieeelongdouble.
7295 (init_float128_ieee): Call it.
7297 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
7300 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
7301 -fcf-protection=branch -mibt.
7302 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
7304 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7307 * config/arm/arm.c (arm_valid_symbolic_address): Handle
7308 arm_word_relocations.
7310 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
7313 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
7314 extern array declarations.
7316 2018-03-27 Richard Biener <rguenther@suse.de>
7319 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
7320 explicit single_use checks.
7322 2018-03-27 Richard Biener <rguenther@suse.de>
7324 PR tree-optimization/85082
7325 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
7328 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7330 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
7331 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
7332 Turn on fasynchronous-unwind-tables and funwind-tables.
7334 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
7337 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
7338 (*bmi_blsr_<mode>_ccz): Ditto.
7340 2018-03-26 Tom de Vries <tom@codesourcery.com>
7342 PR tree-optimization/85063
7343 * omp-general.c (offloading_function_p): New function. Factor out
7345 * omp-offload.c (pass_omp_target_link::gate): ... here.
7346 * omp-general.h (offloading_function_p): Declare.
7347 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
7348 with attribute omp declare target for offloading functions.
7350 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
7352 PR tree-optimization/84005
7353 * tree-data-ref.h (get_base_for_alignment): Declare.
7354 * tree-data-ref.c (get_base_for_alignment_1): New function.
7355 (get_base_for_alignment): Likewise.
7356 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
7357 get_base_for_alignment to find a suitable base object, instead
7358 of always using drb->base_address.
7360 2018-03-23 Jakub Jelinek <jakub@redhat.com>
7363 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
7364 known size by default.
7366 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
7369 * lra-constraints.c (process_alt_operands): Don't match BLKmode
7370 and non BLKmode operands.
7372 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7375 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
7376 Clean up attributes.
7378 2018-03-23 Richard Biener <rguenther@suse.de>
7381 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
7382 we are going to emit early debug for LTO.
7384 2018-03-23 Jakub Jelinek <jakub@redhat.com>
7387 * function.c (match_asm_constraints_1): Don't optimize if input
7388 doesn't satisfy general_operand predicate for output's mode.
7391 * alias.c (write_dependence_p): Don't require for x_canonicalized
7392 non-VOIDmode if x has VOIDmode.
7395 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
7396 just don't try to optimize it rather than assert it never happens.
7398 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
7400 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
7401 macro expansions for definition of ST_INTERNAL_<mode> and
7402 LD_INTERNAL_<mode> builtins.
7403 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
7405 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
7407 (altivec_expand_st_builtin): Likewise.
7408 (altivec_expand_builtin): Remove calls to deleted functions.
7409 (rs6000_address_for_altivec): Delete this function.
7410 * config/rs6000/vector.md: Remove expands for
7411 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
7413 2018-03-22 Sudakshina Das <sudi.das@arm.com>
7416 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
7417 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
7418 re-computing once computed.
7419 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
7420 (arm_init_machine_status): Initialize
7421 machine->static_chain_stack_bytes.
7423 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
7426 * doc/extend.texi: Add four new prototypes for vec_ld.
7427 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
7428 definitions for more logical presentation.
7429 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
7430 entries for V1TI variants of __builtin_altivec_ld builtin.
7431 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
7432 handling of V1TI variant of LVX icode pattern.
7433 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
7434 (rs6000_gimple_fold_builtin): Likewise.
7435 (altivec_init_builtins): Add code to define
7436 __builtin_altivec_lvx_v1ti function.
7438 2018-03-22 Jakub Jelinek <jakub@redhat.com>
7441 * function.c (match_asm_constraints_1): Don't do the optimization
7442 if input isn't a REG, SUBREG, MEM or constant.
7444 2018-03-22 Tom de Vries <tom@codesourcery.com>
7446 PR tree-optimization/84956
7447 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
7448 bb_has_abnormal_pred.
7450 2018-03-22 Jakub Jelinek <jakub@redhat.com>
7453 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
7454 DECL_INITIAL (decl) to decl at the end.
7455 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
7458 2018-03-21 Joseph Myers <joseph@codesourcery.com>
7460 * doc/extend.texi (__builtin_tgmath): Document when complex
7461 integer types are treated as _Complex _Float64.
7463 2018-03-21 Tom de Vries <tom@codesourcery.com>
7465 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
7467 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7469 PR tree-optimization/84960
7470 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
7471 if it is ENTRY block, move them into single succ of ENTRY in that case.
7473 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
7475 PR tree-optimization/84811
7476 * poly-int.h (poly_span_traits): Remove the T3 parameter and
7477 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
7478 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
7479 (known_subrange_p): Update accordingly. Cast each value involved
7480 in the size comparison, rather than casting the result of the
7483 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7485 PR tree-optimization/84982
7486 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
7487 by flipping the least significant bit rather than all bits from
7488 bitpos to bitpos + bitsize - 1.
7490 2018-03-21 Nathan Sidwell <nathan@acm.org>
7492 * doc/extend.texi (Deprecated Features): Remove mention of
7493 long-deleted deprecations.
7495 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7498 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
7499 * configure: Regenerate.
7501 2018-03-21 Tom de Vries <tom@codesourcery.com>
7503 PR tree-optimization/83126
7504 * tree-parloops.c (num_phis): New function.
7505 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
7507 2018-03-21 Nathan Sidwell <nathan@acm.org>
7509 * doc/extend.texi (Deprecated Features): Update deprecated flags,
7510 mention anon-struct/union members and trailing attributes.
7512 2018-03-21 Bin Cheng <bin.cheng@arm.com>
7514 PR tree-optimization/84969
7515 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
7516 builtin memset partitions if they set different rhs values.
7518 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7520 PR rtl-optimization/84989
7521 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
7522 VEC_DUPLICATE with scalar result mode.
7524 2018-03-21 Martin Liska <mliska@suse.cz>
7527 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
7528 not intended return statement.
7530 2018-03-21 Martin Liska <mliska@suse.cz>
7533 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
7534 (chkp_find_bound_slots_1): Limit number of iterations.
7536 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
7539 * Minor grammar fixes for x86 options.
7541 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7544 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
7545 holding REG_CFA_RESTORE notes, instead turn them into a USE.
7547 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
7550 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
7551 (altivec_lvx_<mode>_1op): Likewise.
7552 (altivec_stvx_<mode>_2op): Likewise.
7553 (altivec_stvx_<mode>_1op): Likewise.
7554 (altivec_lvx_<VM2:mode>): New define_expand.
7555 (altivec_stvx_<VM2:mode>): Likewise.
7556 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
7557 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
7558 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
7559 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
7560 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
7561 (rs6000_gen_lvx): Likewise.
7562 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
7563 (altivec_expand_stv_builtin): Likewise.
7564 (altivec_expand_builtin): Likewise.
7565 * config/rs6000/vector.md: Likewise.
7567 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7570 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
7573 2018-03-20 Richard Biener <rguenther@suse.de>
7576 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
7577 sign-conversions as zero, fall back to standard scalar_stmt
7580 2018-03-20 Martin Liska <mliska@suse.cz>
7583 * predict.c (rebuild_frequencies): Handle case when we have
7584 PROFILE_ABSENT, but flag_guess_branch_prob is false.
7586 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7589 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
7590 flag_section_anchors.
7591 * varasm.c (use_blocks_for_decl_p): Remove hack for
7592 dw2_force_const_mem.
7595 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
7597 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
7598 be created, use lowpart_subreg of operands[0] rather than operands[0]
7600 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
7601 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
7602 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
7603 and n constraint instead of aarch64_shift_imm_di and Usd.
7604 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
7605 (*aarch64_<optab>_reg_minus<mode>3): ... this.
7607 2018-03-20 Sudakshina Das <sudi.das@arm.com>
7610 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
7611 to favor GPR over NEON registers.
7612 (<shift>di3_neon): Likewise.
7614 2018-03-20 Tom de Vries <tom@codesourcery.com>
7617 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
7618 (nvptx_process_pars): Emit bar.sync asap and alap.
7620 2018-03-20 Tom de Vries <tom@codesourcery.com>
7623 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
7624 seen_label if seen_label is already set.
7626 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7629 * config/i386/i386.c (fold_builtin_cpu): For features above 31
7630 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
7631 Use 1U instead of 1. Formatting fixes.
7634 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
7635 instead of TREE_TYPE (s1) for the return value.
7637 2018-03-19 Jakub Jelinek <jakub@redhat.com>
7639 PR tree-optimization/84946
7640 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
7641 bitsize + bitsize in poly_uint64 rather than poly_int64.
7644 * dwarf2asm.c: Include fold-const.c.
7645 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
7646 of decl rather than decl itself.
7648 PR rtl-optimization/84643
7649 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
7651 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
7654 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
7655 calling assemble_variable.
7657 2018-03-19 Sudakshina Das <sudi.das@arm.com>
7660 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
7661 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
7663 2018-03-19 Jim Wilson <jimw@sifive.com>
7666 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
7667 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
7668 (riscv_first_stack_step): Likewise.
7669 (riscv_option_override): Use STACK_BOUNDARY instead of
7671 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
7673 (BIGGEST_ALIGNMENT): Set to 128.
7674 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
7675 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
7678 2018-03-19 Richard Biener <rguenther@suse.de>
7680 PR tree-optimization/84933
7681 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
7682 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
7684 2018-03-19 Richard Biener <rguenther@suse.de>
7686 PR tree-optimization/84859
7687 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
7688 (cond_if_else_store_replacement): Perform sinking operation on
7689 single-store BBs regardless of MAX_STORES_TO_SINK setting.
7690 Generalize what a BB with a single eligible store is.
7692 2018-03-19 Richard Biener <rguenther@suse.de>
7694 PR tree-optimization/84929
7695 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
7696 chrec_is_positive against non-chrec arg.
7698 2018-03-19 Tamar Christina <tamar.christina@arm.com>
7701 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
7703 2018-03-18 Martin Liska <mliska@suse.cz>
7705 PR rtl-optimization/84635
7706 * regrename.c (build_def_use): Use matches_mode only when
7709 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
7711 PR tree-optimization/84913
7712 * tree-vect-loop.c (vectorizable_reduction): Don't try to
7713 vectorize chains of COND_EXPRs.
7715 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
7717 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
7719 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
7721 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
7723 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
7725 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
7727 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
7728 Kito Cheng <kito.cheng@gmail.com>
7730 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
7731 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
7732 (nds32_adjust_reg_alloc_order): New function.
7733 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
7735 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
7737 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
7738 nds32_print_operand, nds32_print_operand_address): Use
7739 HOST_WIDE_INT_PRINT_DEC instead.
7741 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
7743 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
7745 2018-03-17 Jakub Jelinek <jakub@redhat.com>
7748 * config/i386/i386.c (initial_ix86_tune_features,
7749 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
7751 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
7752 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
7753 rather than 1u << ix86_tune. Formatting fix.
7754 (ix86_option_override_internal): Change ix86_arch_mask from
7755 unsigned int to unsigned HOST_WIDE_INT, initialize to
7756 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
7757 (ix86_function_specific_restore): Likewise.
7759 2018-03-16 Jakub Jelinek <jakub@redhat.com>
7762 * postreload.c (reload_combine_recognize_pattern): Perform
7763 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
7764 truncate_int_for_mode the result for the destination's mode.
7767 * hsa-gen.c (mem_type_for_type): Fix comment typo.
7768 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
7770 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
7773 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
7776 * lra-assigns.c (lra_split_hard_reg_for): Don't use
7777 regno_allocno_class_array and sorted_pseudos.
7778 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
7779 insns where regno is used.
7781 2018-03-16 Martin Liska <mliska@suse.cz>
7784 * multiple_target.c (create_dispatcher_calls): Redirect
7785 reference in the symbol table.
7787 2018-03-16 Martin Liska <mliska@suse.cz>
7790 * multiple_target.c (create_dispatcher_calls): Redirect also
7793 2018-03-16 Jakub Jelinek <jakub@redhat.com>
7797 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
7798 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
7799 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
7801 2018-03-16 Julia Koval <julia.koval@intel.com>
7803 * doc/invoke.texi (Skylake Server): Add CLWB.
7804 Cannonlake): Remove CLWB.
7806 2018-03-16 Jakub Jelinek <jakub@redhat.com>
7808 PR tree-optimization/84841
7809 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
7811 (FLOAT_ONE_CONST_TYPE): Define.
7812 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
7813 (sort_by_operand_rank): Put entries with higher constant_type last
7814 rather than first to match comments.
7816 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
7818 * config/nios2/nios2.md (movsi_internal): Fix thinko in
7821 2018-03-15 Jakub Jelinek <jakub@redhat.com>
7824 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
7825 check and use address of target always.
7827 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
7830 * config/i386/i386.c (indirect_thunk_needed): Update comments.
7831 (indirect_thunk_bnd_needed): Likewise.
7832 (indirect_thunks_used): Likewise.
7833 (indirect_thunks_bnd_used): Likewise.
7834 (indirect_return_needed): New.
7835 (indirect_return_bnd_needed): Likewise.
7836 (output_indirect_thunk_function): Add a bool argument for
7838 (output_indirect_thunk_function): Don't generate alias for
7839 function return thunk.
7840 (ix86_code_end): Call output_indirect_thunk_function to generate
7841 function return thunks.
7842 (ix86_output_function_return): Set indirect_return_bnd_needed
7843 and indirect_return_needed instead of indirect_thunk_bnd_needed
7844 and indirect_thunk_needed.
7846 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
7848 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
7849 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
7850 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
7852 2018-03-15 David Malcolm <dmalcolm@redhat.com>
7853 Paul Hua <paul.hua.gm@gmail.com>
7856 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
7858 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
7860 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
7861 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
7864 2018-03-15 Tamar Christina <tamar.christina@arm.com>
7867 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
7868 instead of GET_MODE_SIZE when comparing Units.
7870 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
7873 * varasm.c (hash_section): Return an unchangeble hash value
7874 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
7875 Return !aarch64_can_use_per_function_literal_pools_p ().
7877 2018-03-15 Jakub Jelinek <jakub@redhat.com>
7880 * optabs.c (emit_conditional_move): Pass address of cmode's copy
7881 rather than address of cmode as last argument to prepare_cmp_insn.
7883 2018-03-15 Julia Koval <julia.koval@intel.com>
7885 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
7886 F_AVX512VNNI, F_AVX512BITALG): New.
7888 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
7891 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
7892 insn for floating-point loads and stores.
7894 2018-03-14 Carl Love <cel@us.ibm.com>
7896 * config/rs6000/rs6000-c.c: Add macro definitions for
7897 ALTIVEC_BUILTIN_VEC_PERMXOR.
7898 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
7899 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
7900 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
7901 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
7903 * config/doc/extend.texi: Add prototypes for vec_permxor.
7905 2018-03-14 David Malcolm <dmalcolm@redhat.com>
7908 * diagnostic-show-locus.c (class layout_point): Convert m_line
7909 from int to linenum_type.
7910 (line_span::comparator): Use linenum "compare" function when
7911 comparing line numbers.
7912 (test_line_span): New function.
7913 (layout_range::contains_point): Convert param "row" from int to
7915 (layout_range::intersects_line_p): Likewise.
7916 (layout::will_show_line_p): Likewise.
7917 (layout::print_source_line): Likewise.
7918 (layout::should_print_annotation_line_p): Likewise.
7919 (layout::print_annotation_line): Likewise.
7920 (layout::print_leading_fixits): Likewise.
7921 (layout::annotation_line_showed_range_p): Likewise.
7922 (struct line_corrections): Likewise for field m_row.
7923 (line_corrections::line_corrections): Likewise for param "row".
7924 (layout::print_trailing_fixits): Likewise.
7925 (layout::get_state_at_point): Likewise.
7926 (layout::get_x_bound_for_row): Likewise.
7927 (layout::print_line): Likewise.
7928 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
7929 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
7930 * input.c (selftest::test_linenum_comparisons): New function.
7931 (selftest::input_c_tests): Call it.
7932 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
7933 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
7934 * selftest.h (ASSERT_GT): New macro.
7935 (ASSERT_GT_AT): New macro.
7936 (ASSERT_LT): New macro.
7937 (ASSERT_LT_AT): New macro.
7939 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
7941 PR rtl-optimization/84780
7942 * combine.c (distribute_links): Don't make a link based on pc_rtx.
7944 2018-03-14 Martin Liska <mliska@suse.cz>
7946 * tree.c (record_node_allocation_statistics): Use
7947 get_stats_node_kind.
7948 (get_stats_node_kind): New function extracted from
7949 record_node_allocation_statistics.
7950 (free_node): Use get_stats_node_kind.
7952 2018-03-14 Richard Biener <rguenther@suse.de>
7954 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
7955 that the value-set of ANTIC_IN doesn't grow.
7958 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
7960 (BB_VISITED_WITH_VISITED_SUCCS): New define.
7961 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
7963 2018-03-14 Julia Koval <julia.koval@intel.com>
7965 * config.gcc (icelake-client, icelake-server): New.
7967 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
7968 (initial_ix86_arch_features): Ditto.
7969 (PTA_SKYLAKE): Add SGX.
7970 (PTA_ICELAKE): Remove.
7971 (PTA_ICELAKE_CLIENT): New.
7972 (PTA_ICELAKE_SERVER): New.
7973 (ix86_option_override_internal): Split up icelake on icelake client and
7975 (get_builtin_code_for_version): Ditto.
7976 (fold_builtin_cpu): Ditto.
7977 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
7978 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
7979 * config/i386/i386.h (processor_type): Ditto.
7980 * doc/invoke.texi: Ditto.
7982 2018-03-14 Jakub Jelinek <jakub@redhat.com>
7985 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
7986 INTEGER_CST offset, add it together with bitpos / 8 and
7987 sign extend based on POINTER_SIZE.
7991 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
7994 * config/i386/constraints.md (Yc): New register constraint.
7995 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
7996 Use Yc constraint for alternative 2 of operand 0. Remove
7997 preferred_for_speed attribute.
7999 2018-03-14 Richard Biener <rguenther@suse.de>
8001 PR tree-optimization/84830
8002 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
8003 with the old one to avoid oscillations.
8005 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
8008 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
8010 (assign_by_spills): Return a flag of reload assignment failure.
8011 Do not process the reload assignment failures. Do not spill other
8012 reload pseudos if they has the same reg class. Update n if
8014 (lra_assign): Add a return arg. Set up from the result of
8015 assign_by_spills call.
8016 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8017 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8018 usage_insns if it is not NULL.
8019 (spill_hard_reg_in_range): New function.
8020 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8021 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8022 function prototypes.
8023 (lra_assign): Change prototype.
8024 * lra.c (lra): Add code to deal with fails by splitting hard reg
8027 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
8029 * config/riscv/riscv.opt (mrelax): New option.
8030 * config/riscv/riscv.c (riscv_file_start): Emit ".option
8031 "norelax" when riscv_mrelax is disabled.
8032 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
8034 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8037 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
8038 reassociation for int modes.
8040 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8042 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
8043 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
8045 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
8046 * config/aarch64/aarch64-sve.md
8047 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
8048 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
8049 (*extend<mode><Vwide>2): Rename to...
8050 (aarch64_sve_extend<mode><Vwide>2): ...this.
8051 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
8052 renaming the old pattern to...
8053 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
8055 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
8056 define_expand, renaming the old pattern to...
8057 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
8058 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
8059 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
8060 account when deciding which SVE instruction the optab should use.
8061 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
8063 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8065 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
8066 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
8067 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
8068 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
8069 (tlsdesc_small_<mode>): Turn a define_expand and use
8070 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
8071 (tlsdesc_small_advsimd_<mode>): ...this.
8072 (tlsdesc_small_sve_<mode>): New pattern.
8074 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8076 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
8077 (UNSPEC_UMUL_HIGHPART): New constants.
8078 (MUL_HIGHPART): New int iteraor.
8079 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
8080 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
8082 (*<su>mul<mode>3_highpart): New define_insn.
8084 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
8087 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
8090 2018-03-13 Martin Liska <mliska@suse.cz>
8093 * (sem_item_optimizer::sem_item_optimizer): Initialize new
8095 (sem_item_optimizer::~sem_item_optimizer): Release it.
8096 (sem_item_optimizer::merge_classes): Register variable aliases.
8097 (sem_item_optimizer::fixup_pt_set): New function.
8098 (sem_item_optimizer::fixup_points_to_sets): Likewise.
8099 * ipa-icf.h: Declare new variables and functions.
8101 2018-03-13 Jakub Jelinek <jakub@redhat.com>
8104 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
8105 integer_pow2p@2 and test integer_pow2p in condition.
8106 (A < 0 ? C : 0): Similarly for @1.
8109 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
8110 characters starting at p contain '\0' character, don't look beyond
8114 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
8115 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
8118 * reg-stack.c (change_stack): Change update_end var from int to
8119 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
8120 also call set_block_for_insn on the newly added insns and rescan.
8123 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
8124 on the last operand.
8127 * tree.c (stabilize_reference_1): Return save_expr (e) for
8128 STATEMENT_LIST even if it doesn't have side-effects.
8130 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
8132 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
8134 2018-03-12 Renlin Li <renlin.li@arm.com>
8136 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
8137 aarch64_output_scalar_simd_mov_immediate.
8139 2018-03-12 Martin Sebor <msebor@redhat.com>
8141 PR tree-optimization/83456
8142 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
8143 for perfectly overlapping calls to memcpy.
8144 (gimple_fold_builtin_memory_chk): Same.
8145 (gimple_fold_builtin_strcpy): Handle no-warning.
8146 (gimple_fold_builtin_stxcpy_chk): Same.
8147 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
8149 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
8151 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
8152 parameter. Use it for SFmode.
8153 (rs6000_function_arg_advance_1): Adjust.
8154 (rs6000_function_arg): Adjust.
8155 (rs6000_gimplify_va_arg): Pass false for that new parameter.
8157 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
8159 PR rtl-optimization/84169
8160 PR rtl-optimization/84780
8161 * combine.c (can_combine_p): Check for a 2-insn combination whether
8162 the destination register is used between the two insns, too.
8164 2018-03-12 Richard Biener <rguenther@suse.de>
8166 PR tree-optimization/84803
8167 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
8168 for refs DR analysis didn't process.
8170 2018-03-12 Richard Biener <rguenther@suse.de>
8172 PR tree-optimization/84777
8173 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
8174 force-vectorize loops ignore whether we are optimizing for size.
8176 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
8178 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
8179 (TARGET_MD_ASM_ADJUST): Define.
8181 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
8182 Kito Cheng <kito.cheng@gmail.com>
8183 Chung-Ju Wu <jasonwucj@gmail.com>
8185 * config/nds32/nds32.c (nds32_compute_stack_frame,
8186 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
8187 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
8188 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
8189 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
8190 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
8191 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
8192 * config/nds32/nds32.md (prologue, epilogue): Use macro
8193 NDS32_V3PUSH_AVAILABLE_P to do checking.
8195 2018-03-11 Jakub Jelinek <jakub@redhat.com>
8198 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
8199 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
8200 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
8201 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
8202 addition of most attributes on !orig_type_die or the attribute not
8203 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
8205 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8206 Chung-Ju Wu <jasonwucj@gmail.com>
8208 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
8210 * config/nds32/nds32.opt (mvh): New option.
8212 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8213 Chung-Ju Wu <jasonwucj@gmail.com>
8215 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
8217 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
8218 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
8221 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8222 Chung-Ju Wu <jasonwucj@gmail.com>
8224 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
8226 * config/nds32/nds32-multiple.md (strlensi): New pattern.
8227 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
8229 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
8230 Kito Cheng <kito.cheng@gmail.com>
8231 Chung-Ju Wu <jasonwucj@gmail.com>
8233 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
8234 UNSPEC_FFMISM and UNSPEC_FLMISM.
8235 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
8236 for ffb, ffmism and flmism.
8237 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
8238 (unspec_ffmism): Ditto.
8239 (unspec_flmism): Ditto.
8240 (nds32_expand_builtin_impl): Check if string extension is available.
8241 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
8242 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
8244 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
8247 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
8250 * lra-assigns.c (assign_by_spills): Return a flag of reload
8251 assignment failure. Do not process the reload assignment
8252 failures. Do not spill other reload pseudos if they has the same
8254 (lra_assign): Add a return arg. Set up from the result of
8255 assign_by_spills call.
8256 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8257 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8258 usage_insns if it is not NULL.
8259 (spill_hard_reg_in_range): New function.
8260 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8261 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8262 function prototypes.
8263 (lra_assign): Change prototype.
8264 * lra.c (lra): Add code to deal with fails by splitting hard reg
8267 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
8270 * config/i386/i386.opt: Replace Enforcment with Enforcement.
8272 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
8275 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
8276 (dw_val_node): Add val_symbolic_view.
8277 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
8278 (symview_upper_bound): New.
8279 (new_line_info_table): Initialize symviews_since_reset.
8280 (dwarf2out_source_line): Count symviews_since_reset and set
8281 symview_upper_bound.
8282 (dw_val_equal_p): Handle symview.
8283 (add_AT_symview): New.
8284 (print_dw_val): Handle symview.
8285 (attr_checksum, attr_checksum_ordered): Likewise.
8286 (same_dw_val_p, size_of_die): Likewise.
8287 (value_format, output_die): Likewise.
8288 (add_high_low_attributes): Use add_AT_symview for entry_view.
8289 (dwarf2out_finish): Reset symview_upper_bound, clear
8292 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
8295 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
8296 Add strict argument and use it.
8297 (rs6000_split_multireg_move): Update for new strict argument.
8298 (mem_operand_gpr): Disallow all non-offsettable addresses.
8299 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
8301 2018-03-09 Jakub Jelinek <jakub@redhat.com>
8304 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
8305 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
8306 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
8309 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
8310 decl, use remap_type if we want to use the type.
8312 2018-03-09 Martin Sebor <msebor@redhat.com>
8314 PR tree-optimization/84526
8315 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
8317 (builtin_access::generic_overlap): Be prepared to handle non-array
8320 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
8322 PR rtl-optimization/84682
8323 * lra-constraints.c (process_address_1): Check is_address flag
8324 for address constraints.
8325 (process_alt_operands): Likewise.
8326 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
8327 preprocess_constraints.
8328 * recog.h (preprocess_constraints): Add oploc parameter.
8330 * recog.c (preprocess_constraints): Test address_operand for
8331 CT_ADDRESS constraints.
8333 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
8336 * lra-assigns.c (assign_by_spills): Return a flag of reload
8337 assignment failure. Do not process the reload assignment
8338 failures. Do not spill other reload pseudos if they has the same
8340 (lra_assign): Add a return arg. Set up from the result of
8341 assign_by_spills call.
8342 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8343 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8344 usage_insns if it is not NULL.
8345 (spill_hard_reg_in_range): New function.
8346 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8347 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8348 function prototypes.
8349 (lra_assign): Change prototype.
8350 * lra.c (lra): Add code to deal with fails by splitting hard reg
8353 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8356 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
8357 Accept complain bool parameter. Only emit errors if it is true.
8358 (arm_parse_cpu_option_name): Likewise.
8359 (arm_target_thumb_only): Adjust callers of the above.
8360 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
8361 prototype to take a default true bool parameter.
8362 (arm_parse_arch_option_name): Likewise.
8364 2018-03-09 David Malcolm <dmalcolm@redhat.com>
8365 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
8369 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
8370 * configure: Regenerate.
8371 * configure.ac ("linker --version-script option"): New.
8372 ("linker soname option"): New.
8374 2018-03-09 Richard Biener <rguenther@suse.de>
8376 PR tree-optimization/84775
8377 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
8378 immediate uses of predicate stmts and mark them modified.
8381 PR tree-optimization/84178
8382 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
8384 (version_loop_for_if_conversion): Delay update_ssa call.
8385 (tree_if_conversion): Delay update_ssa until after predicate
8388 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
8391 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
8392 when the function accesses prior frames.
8394 2018-03-08 Jakub Jelinek <jakub@redhat.com>
8397 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
8398 gen_llsym, otherwise call maybe_gen_llsym.
8401 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
8402 has ',' character inside of it.
8404 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8407 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
8408 as clobbering CC_REGNUM.
8410 2018-03-08 Richard Biener <rguenther@suse.de>
8413 * tree-scalar-evolution.c: Include tree-into-ssa.h.
8414 (follow_copies_to_constant): Do not follow SSA names registered
8417 2018-03-08 Richard Biener <rguenther@suse.de>
8419 PR tree-optimization/84178
8420 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
8422 (version_loop_for_if_conversion): Delay update_ssa call.
8423 (tree_if_conversion): Delay update_ssa until after predicate
8426 2018-03-08 David Malcolm <dmalcolm@redhat.com>
8428 PR tree-optimization/84178
8429 * tree-if-conv.c (release_bb_predicate): Remove the
8430 the assertion that the stmts have NULL use_ops.
8431 Discard the statements, asserting that they haven't
8432 yet been added to a BB.
8434 2018-03-08 Richard Biener <rguenther@suse.de>
8436 PR tree-optimization/84746
8437 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
8438 (phi_translate): Pass in destination ANTIC_OUT set.
8439 (phi_translate_1): Likewise. For a simplified result lookup
8440 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
8441 (phi_translate_set): Adjust.
8442 (do_pre_regular_insertion): Likewise.
8443 (do_pre_partial_partial_insertion): Likewise.
8445 2018-03-08 Martin Liska <mliska@suse.cz>
8447 PR gcov-profile/84735
8448 * doc/gcov.texi: Document usage of profile files.
8449 * gcov-io.h: Document changes in the format.
8451 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
8455 * dwarf2out.c (struct dw_line_info_table): Update comments for
8457 (FORCE_RESET_NEXT_VIEW): New.
8458 (FORCE_RESETTING_VIEW_P): New.
8459 (RESETTING_VIEW_P): Check for -1 too.
8460 (ZERO_VIEW_P): Likewise.
8461 (new_line_info_table): Force-reset next view.
8462 (dwarf2out_begin_function): Likewise.
8463 (dwarf2out_source_line): Simplify zero_view_p initialization.
8464 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
8465 view directly. Omit view when omitting .loc at line 0.
8467 2018-03-08 Jakub Jelinek <jakub@redhat.com>
8469 PR tree-optimization/84740
8470 * tree-switch-conversion.c (process_switch): Call build_constructors
8471 only if info.phi_count is non-zero.
8473 PR tree-optimization/84739
8474 * tree-tailcall.c (find_tail_calls): Check call arguments against
8475 DECL_ARGUMENTS (current_function_decl) rather than
8476 DECL_ARGUMENTS (func) when checking for tail recursion.
8478 2018-03-07 Jakub Jelinek <jakub@redhat.com>
8480 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
8481 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
8482 Volker Reichelt's entry and add entries for people that perform
8483 GCC fuzzy testing and report numerous bugs.
8485 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
8488 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
8489 readonly data in sdata, if that is disabled.
8490 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
8491 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
8492 -mreadonly-in-sdata option.
8494 2018-03-07 Martin Sebor <msebor@redhat.com>
8496 PR tree-optimization/84468
8497 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
8498 basic block when looking for nul assignment.
8500 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
8503 * except.h (output_function_exception_table): Adjust prototype.
8504 * except.c (output_function_exception_table): Remove FNNAME parameter
8505 and add SECTION parameter. Ouput one part of the table at a time.
8506 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
8507 the first part of the exception table and emit unwind directives.
8508 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
8509 (i386_pe_seh_cold_init): Likewise.
8510 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
8511 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
8512 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
8513 (ix86_output_call_insn): Emit a nop in one more case for SEH.
8514 * config/i386/winnt.c: Include except.h.
8515 (struct seh_frame_state): Add reg_offset, after_prologue and
8516 in_cold_section fields.
8517 (i386_pe_seh_end_prologue): Set seh->after_prologue.
8518 (i386_pe_seh_cold_init): New function.
8519 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
8520 to seh->in_cold_section.
8521 (seh_emit_push): Record the offset of the push.
8522 (seh_emit_save): Record the offet of the save.
8523 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
8524 Test seh->after_prologue to disregard the epilogue.
8525 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
8526 (i386_pe_end_cold_function): New function.
8528 2018-03-07 Jakub Jelinek <jakub@redhat.com>
8531 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
8532 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
8535 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
8537 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
8538 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
8541 * multiple_target.c: Include tree-inline.h and intl.h.
8542 (expand_target_clones): Diagnose and fail if node->definition and
8543 !tree_versionable_function_p (node->decl).
8545 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
8547 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
8549 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
8550 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
8551 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
8553 2018-03-06 Jakub Jelinek <jakub@redhat.com>
8556 * combine.c (try_combine): Use reg_or_subregno instead of handling
8557 just paradoxical SUBREGs and REGs.
8559 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
8561 * config/arc/arc.c (arc_finalize_pic): Remove function.
8562 (arc_must_save_register): We use single base PIC register, remove
8563 checks to save/restore the PIC register.
8564 (arc_expand_prologue): Likewise.
8565 * config/arc/arc-protos.h (arc_set_default_type_attributes):
8567 (arc_verify_short): Likewise.
8568 (arc_attr_type): Likewise.
8569 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
8570 (walk_stores): Likewise.
8571 (arc_address_cost): Make it static.
8572 (arc_verify_short): Likewise.
8573 (branch_dest): Likewise.
8574 (arc_attr_type): Likewise.
8575 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
8576 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
8577 (arc_final_prescan_insn): Remove inserting the nops due to
8578 hardware hazards. It is done in reorg step.
8579 (insn_length_variant_t): Remove.
8580 (insn_length_parameters_t): Likewise.
8581 (arc_insn_length_parameters): Likewise.
8582 (arc_get_insn_variants): Likewise.
8583 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
8585 2018-03-06 Jakub Jelinek <jakub@redhat.com>
8588 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
8591 PR tree-optimization/84687
8592 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
8594 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
8596 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8598 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
8599 Rename to ppc_speculation_barrier.
8600 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
8601 __builtin_ppc_speculation_barrier.
8603 2018-03-05 Jakub Jelinek <jakub@redhat.com>
8606 * combine.c (combine_simplify_rtx): Don't try to simplify if
8607 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
8610 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
8612 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
8613 to 32 bytes when compiling for POWER9.
8615 2018-03-05 Jakub Jelinek <jakub@redhat.com>
8618 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
8619 regparm >= 3 with no arg reg available also for calls with
8620 flag_force_indirect_call. Pass decl to ix86_function_regparm.
8623 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
8625 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
8627 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
8630 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
8632 2018-03-05 Richard Biener <rguenther@suse.de>
8634 PR tree-optimization/84486
8635 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
8636 When inserting a __builtin_assume_aligned call set the LHS
8637 SSA name alignment info accordingly.
8639 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
8641 PR tree-optimization/84114
8642 * config/aarch64/aarch64.c (aarch64_reassociation_width)
8643 Avoid reassociation of FLOAT_MODE addition.
8645 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
8647 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
8648 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
8649 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
8650 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
8651 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
8652 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
8653 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
8655 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
8656 __builtin_ia32_wbinvd): New builtins.
8657 (SPECIAL_ARGS2): New.
8658 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
8659 (SPECIAL_ARGS2): New.
8660 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
8661 (ix86_valid_target_attribute_inner_p): Ditto.
8662 (ix86_init_mmx_sse_builtins): Add special_args2.
8663 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
8664 TARGET_WBNOINVD_P): New.
8665 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
8666 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
8667 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
8668 * config/i386/immintrin.h (_wbinvd): New intrinsic.
8669 * config/i386/pconfigintrin.h: New file.
8670 * config/i386/wbnoinvdintrin.h: Ditto.
8671 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
8673 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
8675 2018-03-05 Richard Biener <rguenther@suse.de>
8677 PR tree-optimization/84670
8678 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
8680 (BB_VISITED_WITH_VISITED_SUCCS): New define.
8681 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
8682 (compute_antic_aux): Only assert the number of values in ANTIC_IN
8683 doesn't grow if all successors (recursively) were visited at least
8686 2018-03-05 Richard Biener <rguenther@suse.de>
8688 PR tree-optimization/84650
8689 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
8690 if executed in the loop pipeline.
8692 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
8694 * doc/configfiles.texi (Configuration Files): Move info about
8695 conditionalizing $target-protos.h to...
8696 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
8697 differs from $target-protos.h.
8699 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
8700 Chung-Ju Wu <jasonwucj@gmail.com>
8702 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
8703 * config/nds32/nds32-multiple.md (setmemsi): Define.
8704 * config/nds32/nds32-memory-manipulation.c
8705 (nds32_gen_dup_4_byte_to_word_value): New.
8706 (emit_setmem_word_loop): New.
8707 (emit_setmem_byte_loop): New.
8708 (nds32_expand_setmem_loop): New.
8709 (nds32_expand_setmem_loop_v3m): New.
8710 (nds32_expand_setmem_unroll): New.
8711 (nds32_expand_setmem): New.
8713 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
8714 Chung-Ju Wu <jasonwucj@gmail.com>
8716 * config/nds32/nds32-memory-manipulation.c
8717 (nds32_emit_load_store): New.
8718 (nds32_emit_post_inc_load_store): New.
8719 (nds32_emit_mem_move): New.
8720 (nds32_emit_mem_move_block): New.
8721 (nds32_expand_movmemsi_loop_unknown_size): New.
8722 (nds32_expand_movmemsi_loop_known_size): New.
8723 (nds32_expand_movmemsi_loop): New.
8724 (nds32_expand_movmemsi_unroll): New.
8725 (nds32_expand_movmemqi): Rename ...
8726 (nds32_expand_movmemsi): ... to this.
8727 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
8728 (movmemsi): ... to this.
8729 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
8730 (nds32_expand_movmemsi): ... to this.
8732 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
8733 Monk Chiang <sh.chiang04@gmail.com>
8734 Chung-Ju Wu <jasonwucj@gmail.com>
8736 * config/nds32/nds32-protos.h
8737 (nds32_expand_load_multiple): New arguments.
8738 (nds32_expand_store_multiple): Ditto.
8739 (nds32_valid_multiple_load_store): Rename ...
8740 (nds32_valid_multiple_load_store_p): ... to this.
8741 * config/nds32/nds32-memory-manipulation.c
8742 (nds32_expand_load_multiple): Refine implementation.
8743 (nds32_expand_store_multiple): Ditto.
8744 * config/nds32/nds32-multiple.md
8745 (load_multiple): Update nds32_expand_load_multiple interface.
8746 (store_multiple): Update nds32_expand_store_multiple interface.
8747 * config/nds32/nds32-predicates.c
8748 (nds32_valid_multiple_load_store): Rename ...
8749 (nds32_valid_multiple_load_store_p): ... to this and refine
8751 * config/nds32/predicates.md
8752 (nds32_load_multiple_and_update_address_operation): New predicate.
8753 (nds32_store_multiple_and_update_address_operation): New predicate.
8755 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
8756 Chung-Ju Wu <jasonwucj@gmail.com>
8758 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
8759 (combo): New attribute.
8760 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
8762 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
8764 * config/nds32/nds32.opt: Change -mcmodel= default value.
8766 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
8767 Monk Chiang <sh.chiang04@gmail.com>
8768 Chung-Ju Wu <jasonwucj@gmail.com>
8770 * config/nds32/constants.md (unspec_element): New enum.
8771 * config/nds32/constraints.md (Umw): New constraint.
8772 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
8773 * config/nds32/nds32-intrinsic.md: Likewise.
8774 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
8775 (nds32_valid_smw_lwm_base_p): New.
8776 (nds32_output_smw_single_word): New.
8777 (nds32_output_lmw_single_word): New.
8778 (nds32_expand_unaligned_load): New.
8779 (nds32_expand_unaligned_store): New.
8780 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
8781 (nds32_output_smw_single_word): Declare.
8782 (nds32_output_lmw_single_word): Declare.
8783 (nds32_expand_unaligned_load): Declare.
8784 (nds32_expand_unaligned_store): Declare.
8785 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
8786 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
8787 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
8788 NDS32_BUILTIN_UASTORE_DW.
8789 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
8792 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
8793 Kito Cheng <kito.cheng@gmail.com>
8794 Chung-Ju Wu <jasonwucj@gmail.com>
8796 * config/nds32/nds32-intrinsic.c
8797 (nds32_expand_builtin_null_ftype_reg): Delete.
8798 (nds32_expand_builtin_reg_ftype_imm): Ditto.
8799 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
8800 (nds32_read_argument): New.
8801 (nds32_legitimize_target): Ditto.
8802 (nds32_legitimize_argument): Ditto.
8803 (nds32_check_constant_argument): Ditto.
8804 (nds32_expand_unop_builtin): Ditto.
8805 (nds32_expand_unopimm_builtin): Ditto.
8806 (nds32_expand_binop_builtin): Ditto.
8807 (nds32_builtin_decl_impl): Ditto.
8808 (builtin_description): Ditto.
8809 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
8810 (nds32_init_builtins_impl): Ditto.
8811 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
8812 (nds32_builtin_decl): New.
8813 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
8814 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
8816 2018-03-02 Jeff Law <law@redhat.com>
8818 * reorg.c (stop_search_p): Handle DEBUG_INSN.
8819 (redundant_insn, fill_simple_delay_slots): Likewise.
8820 (fill_slots_from_thread): Likewise.
8821 * resource.c (mark_referenced_resources): Likewise.
8822 (mark_set_resources, find_dead_or_set_registers): Likewise.
8824 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8826 * substring-locations.h (format_warning_va): Formatting fix for
8828 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
8830 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
8831 * substring-locations.c: Include intl.h.
8832 (format_warning_va): Turned into small wrapper around
8833 format_warning_n_va, renamed to ...
8834 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
8835 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
8837 (format_warning_at_substring_n): New function.
8838 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
8839 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
8840 format_warning_at_substring with just a shorter name instead of
8841 const function pointer.
8842 (fmtwarn_n): New function.
8843 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
8844 appropriate, get rid of all the fmtstr temporaries, move conditionals
8845 with G_() wrapped string literals directly into fmtwarn arguments,
8846 cast dir.len to (int), formatting fixes.
8848 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
8850 * doc/invoke.texi: Remove "Cilk Plus" references.
8852 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8853 Richard Biener <rguenther@suse.de>
8856 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
8857 for error or warning attributes if CALL_FROM_THUNK_P is set.
8860 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8863 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
8864 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
8867 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
8868 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
8870 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
8871 instead of -1U in last predictors element's probability member.
8873 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
8876 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
8877 arguments if they are comparable.
8879 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
8881 PR tree-optimization/84634
8882 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
8883 masks and masked_loop_p with a single loop_masks, making sure it's
8884 null for bb vectorization.
8886 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
8888 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
8889 (vect_analyze_data_ref_access): Use loop->safe_len rather than
8890 loop->force_vectorize to check whether there is no alias.
8892 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8895 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
8897 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
8899 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
8900 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
8901 instead of a loop around prev_real_insn.
8902 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
8906 * config/i386/i386.c (ix86_print_operand): Use conditional
8907 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
8910 2018-03-02 Richard Biener <rguenther@suse.de>
8912 PR tree-optimization/84427
8913 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
8914 (bitmap_set_subtract_values): Rewrite to handle multiple
8917 (prune_clobbered_mems): Likewise.
8918 (phi_translate): Take edge instead of pred/phiblock.
8919 (phi_translate_1): Likewise.
8920 (phi_translate_set): Likewise. Insert all translated
8921 exprs for a value into the set, keeping possibly multiple
8922 expressions per value.
8923 (compute_antic_aux): Adjust for phi_translate changes.
8924 When intersecting union the expressions and prune those
8925 not in the final value set, keeping possibly multiple
8926 expressions per value. Do not use value-insertion
8927 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
8928 all expressions. Add verification that the value-sets
8929 only shrink during iteration.
8930 (compute_partial_antic_aux): Adjust for the phi_translate changes.
8931 (do_pre_regular_insertion): Likewise.
8932 (do_pre_partial_partial_insertion): Likewise.
8934 2018-03-02 Richard Biener <rguenther@suse.de>
8937 * config/darwin.c (saved_debug_info_level): New static global.
8938 (darwin_asm_lto_start): Disable debug info generation for LTO out.
8939 (darwin_asm_lto_end): Restore debug info generation settings.
8941 2018-03-01 Martin Liska <mliska@suse.cz>
8944 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
8947 2018-03-01 Richard Biener <rguenther@suse.de>
8950 * dwarf2out.c (gen_variable_die): Properly handle late VLA
8951 type annotation with LTO when debug was disabled at compile-time.
8953 2018-03-01 Matthew Fortune <mfortune@gmail.com>
8955 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
8957 (mips_final_postscan_insn): Likewise.
8959 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
8961 PR rtl-optimization/84528
8962 * alias.c (init_alias_target): Add commentary.
8963 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
8964 a unique base value if the frame pointer is not eliminated
8965 to the stack pointer.
8967 2018-03-01 Tom de Vries <tom@codesourcery.com>
8969 PR rtl-optimization/83327
8970 * lra-int.h (hard_regs_spilled_into): Declare.
8971 * lra.c (hard_regs_spilled_into): Define.
8972 (init_reg_info): Init hard_regs_spilled_into.
8973 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
8974 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
8975 (process_bb_lives): Handle hard_regs_spilled_into.
8976 (lra_create_live_ranges_1): Before doing liveness propagation, clear
8977 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
8979 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
8981 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
8982 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
8983 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
8984 * config/rs6000/aix72.h: New file.
8986 2018-02-28 Jakub Jelinek <jakub@redhat.com>
8988 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
8989 instead of warning_at with conditional singular and plural messages
8993 * stor-layout.c (update_alignment_for_field): For
8994 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
8995 && !DECL_PACKED (field), do the alignment update, just use
8996 only desired_align instead of MAX (type_align, desired_align)
8998 (place_field): Don't do known_align < desired_align handling
8999 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
9000 is non-NULL, instead do it after rli->prev_field handling and
9001 only if not within a bitfield word. For DECL_PACKED (field)
9002 use type_align of BITS_PER_UNIT.
9004 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
9006 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
9007 superfluous parentheses and trailing spaces.
9009 2018-02-28 Richard Biener <rguenther@suse.de>
9011 PR tree-optimization/84584
9012 * graphite-scop-detection.c (scop_detection::add_scop): Discard
9013 SCoPs with fake exit edge.
9015 2018-02-28 Martin Liska <mliska@suse.cz>
9018 * timevar.c (timer::print): Fix format to properly print 100%
9021 2018-02-28 Richard Biener <rguenther@suse.de>
9024 * genmatch.c (capture_info::walk_match): Do not mark
9025 captured expressions without operands as expr_p given
9026 they act more like predicates and should be subject to
9027 "lost tail" side-effect preserving.
9029 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
9031 PR rtl-optimization/81611
9032 * auto-inc-dec.c (attempt_change): Move dead note from
9033 mem_insn if it's the next use of regno
9034 (find_address): Take address use of reg holding
9035 non-incremented value. Add parm to limit search to the named
9037 (merge_in_block): Attempt to use a mem insn that is the next
9038 use of the original regno.
9040 2018-02-27 Martin Sebor <msebor@redhat.com>
9043 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
9044 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
9046 2018-02-27 Martin Sebor <msebor@redhat.com>
9048 PR translation/84207
9049 * diagnostic-core.h (warning_n, error_n, inform_n): Change
9050 n argument to unsigned HOST_WIDE_INT.
9051 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
9052 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
9053 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
9054 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
9056 2018-02-27 Richard Biener <rguenther@suse.de>
9058 PR tree-optimization/84512
9059 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
9060 Do not use the estimate returned from record_stmt_cost for
9061 the scalar iteration cost but sum properly using add_stmt_cost.
9063 2018-02-27 Richard Biener <rguenther@suse.de>
9065 PR tree-optimization/84466
9066 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
9067 Adjust last change to less strictly validate use operands.
9069 2018-02-27 Martin Liska <mliska@suse.cz>
9071 PR gcov-profile/84548
9072 * gcov.c (process_file): Allow partial overlap and consider it
9073 also as group functions.
9074 (output_lines): Properly calculate range of lines for a group.
9076 2018-02-27 Martin Liska <mliska@suse.cz>
9078 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
9079 'ggc' suffixes. Change first column width.
9080 (timer::print): Fix formatting of the column.
9082 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
9084 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
9085 preserve inline entry blocks for the sake of debug inline
9086 entry point markers alone.
9087 (remove_unused_locals): Suggest in comments a better place to
9088 force the preservation of inline entry blocks that are
9089 otherwise unused, but do not preserve them.
9091 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9093 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
9095 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9098 * config/i386/constraints.md (Bs): Replace
9099 ix86_indirect_branch_register with
9100 TARGET_INDIRECT_BRANCH_REGISTER.
9102 * config/i386/i386.md (indirect_jump): Likewise.
9103 (tablejump): Likewise.
9104 (*sibcall_memory): Likewise.
9105 (*sibcall_value_memory): Likewise.
9106 Peepholes of indirect call and jump via memory: Likewise.
9107 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
9108 (*sibcall_value_GOT_32): Likewise.
9109 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9110 (GOT_memory_operand): Likewise.
9111 (call_insn_operand): Likewise.
9112 (sibcall_insn_operand): Likewise.
9113 (GOT32_symbol_operand): Likewise.
9114 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
9116 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
9118 PR rtl-optimization/83496
9119 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
9120 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
9121 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
9122 redundant insn, if any.
9123 (relax_delay_slots): Likewise.
9124 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
9126 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
9128 PR tree-optimization/83965
9129 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
9130 that grouped statements are part of a reduction chain. Return
9131 true if the statement is not marked as a reduction itself but
9133 (vect_recog_dot_prod_pattern): Don't check whether the statement
9134 is part of a group here.
9135 (vect_recog_sad_pattern): Likewise.
9136 (vect_recog_widen_sum_pattern): Likewise.
9138 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
9141 * final.c (rest_of_clean_state): Also look for calls inside sequences.
9143 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9146 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
9148 (ix86_output_indirect_function_return): New prototype.
9149 (ix86_split_simple_return_pop_internal): Likewise.
9150 * config/i386/i386.c (indirect_return_via_cx): New.
9151 (indirect_return_via_cx_bnd): Likewise.
9152 (indirect_thunk_name): Handle return va CX_REG.
9153 (output_indirect_thunk_function): Create alias for
9154 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
9155 (ix86_output_indirect_jmp): Remove the bool argument.
9156 (ix86_output_indirect_function_return): New function.
9157 (ix86_split_simple_return_pop_internal): Likewise.
9158 * config/i386/i386.md (*indirect_jump): Don't pass false
9159 to ix86_output_indirect_jmp.
9160 (*tablejump_1): Likewise.
9161 (simple_return_pop_internal): Change it to define_insn_and_split.
9162 Call ix86_split_simple_return_pop_internal to split it for
9164 (simple_return_indirect_internal): Call
9165 ix86_output_indirect_function_return instead of
9166 ix86_output_indirect_jmp.
9168 2018-02-26 Jakub Jelinek <jakub@redhat.com>
9171 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
9172 memset and value initialization afterwards.
9174 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
9176 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
9178 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9181 * common/config/aarch64/aarch64-common.c
9182 (aarch_option_optimization_table[]): Switch
9183 off fomit-frame-pointer
9185 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
9186 Chung-Ju Wu <jasonwucj@gmail.com>
9188 * config/nds32/nds32-multiple.md (load_multiple): Disallow
9190 (store_multiple): Ditto.
9192 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
9194 * config.gcc: Add --with-cpu support for nds32 target.
9195 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
9196 * config/nds32/nds32.opt: Add -mcpu= option.
9198 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
9200 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
9201 isel=yes): Warn for these deprecated options.
9203 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
9205 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
9206 ISA_2_5_MASKS_EMBEDDED.
9208 2018-02-23 Jakub Jelinek <jakub@redhat.com>
9210 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
9211 p->max as pointers rather than using iterative_hash_expr.
9213 2018-02-23 Carl Love <cel@us.ibm.com>
9215 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
9216 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
9218 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
9219 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
9220 P8V_BUILTIN_VEC_VUNSIGNED2.
9222 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
9225 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
9226 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
9228 * lra-constraints.c (curr_insn_transform): Set up
9229 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
9231 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
9232 * lra-eliminations.c (spill_pseudos): Ditto.
9233 (process_insn_for_elimination): Ditto.
9234 * lra-lives.c (reg_early_clobber_p): Use the new macros.
9235 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
9236 LRA_NON_CLOBBERED_ALT.
9238 2018-02-22 Martin Sebor <msebor@redhat.com>
9240 PR tree-optimization/84480
9241 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
9242 to maybe_diag_stxncpy_trunc. Call it.
9243 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
9244 from gimple_fold_builtin_strcpy. Print inlining stack.
9245 (handle_builtin_stxncpy): Print inlining stack.
9246 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
9248 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
9251 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
9252 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
9253 and -fcheck-pointer-bounds are used together.
9254 (indirect_thunk_prefix): New enum.
9255 (indirect_thunk_need_prefix): New function.
9256 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
9257 "_nt" instead of "_bnd" for NOTRACK prefix.
9258 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
9259 (output_indirect_thunk_function): Likewise.
9261 (ix86_code_end): Update output_indirect_thunk_function calls.
9262 (ix86_output_indirect_branch_via_reg): Replace
9263 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
9264 (ix86_output_indirect_branch_via_push): Likewise.
9265 (ix86_output_function_return): Likewise.
9266 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
9267 incompatible with -fcf-protection=branch and
9268 -fcheck-pointer-bounds.
9270 2018-02-22 Steve Ellcey <sellcey@cavium.com>
9273 * config/aarch64/aarch64.c (aarch64_print_address_internal):
9274 Change gcc_assert call to output_operand_lossage.
9276 2018-02-22 Steve Ellcey <sellcey@cavium.com>
9278 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
9280 2018-02-22 DJ Delorie <dj@redhat.com>
9281 Sebastian Perta <sebastian.perta@renesas.com>
9282 Oleg Endo <olegendo@gcc.gnu.org>
9284 * config/rx/rx.c (rx_rtx_costs): New function.
9285 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
9287 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
9289 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
9291 2018-02-22 Martin Liska <mliska@suse.cz>
9294 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
9295 Add "native" as a possible value.
9297 2018-02-22 Martin Liska <mliska@suse.cz>
9300 * config/i386/i386.c (ix86_option_override_internal):
9301 Add "native" as a possible value for -march and -mtune.
9303 2018-02-22 Jakub Jelinek <jakub@redhat.com>
9306 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
9307 to all type variants.
9309 PR tree-optimization/84503
9310 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
9311 width as info->bitpos + info->bitsize - start.
9312 (merged_store_group::merge_overlapping): Simplify width computation.
9313 (check_no_overlap): New function.
9314 (imm_store_chain_info::try_coalesce_bswap): Compute expected
9315 start + width and last_order of the group, fail if check_no_overlap
9317 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
9318 to group if check_no_overlap fails.
9320 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
9322 * config/rs6000/altivec.md: Delete contraint arguments to
9323 define_expand, define_split, and define_peephole2, and in
9324 define_insn_and_split if always unused.
9325 * config/rs6000/darwin.md: Ditto.
9326 * config/rs6000/dfp.md: Ditto.
9327 * config/rs6000/rs6000.md: Ditto.
9328 * config/rs6000/sync.md: Ditto.
9329 * config/rs6000/vector.md: Ditto.
9330 * config/rs6000/vsx.md: Ditto.
9332 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
9334 * config/rs6000/altivec.md: Write output control strings as braced
9335 blocks instead of double-quoted strings.
9336 * config/rs6000/darwin.md: Ditto.
9337 * config/rs6000/rs6000.md: Ditto.
9338 * config/rs6000/vector.md: Ditto.
9339 * config/rs6000/vsx.md: Ditto.
9341 2018-02-21 Jason Merrill <jason@redhat.com>
9343 PR c++/84314 - ICE with templates and fastcall attribute.
9344 * attribs.c (build_type_attribute_qual_variant): Remove assert.
9346 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
9348 * ipa-cp.c (determine_versionability): Fix comment typos.
9350 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
9353 * ipa-cp.c (determine_versionability): Do not version functions caling
9356 2018-02-21 Martin Liska <mliska@suse.cz>
9359 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
9360 Add "native" as a possible value.
9361 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
9362 the macro when native cpu detection is available.
9364 2018-02-21 Martin Liska <mliska@suse.cz>
9367 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
9368 Add "native" as a possible value.
9369 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
9370 when native cpu detection is available.
9372 2018-02-21 Jakub Jelinek <jakub@redhat.com>
9373 Martin Sebor <msebor@redhat.com>
9375 PR tree-optimization/84478
9376 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
9378 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
9379 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
9380 support which is conservatively correct, for 2 only stay conservative
9381 for maxlen. Formatting and comment capitalization fixes. Add STRICT
9382 argument to the 2 argument get_range_strlen, adjust 6 arg
9383 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
9385 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
9386 (gimple_fold_builtin_strlen): Pass true as last argument to
9389 2018-02-20 Martin Sebor <msebor@redhat.com>
9392 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
9393 (builtin_memref::set_base_and_offset): Same. Handle inner references.
9394 (builtin_memref::builtin_memref): Factor out parts into
9395 set_base_and_offset and call it.
9397 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
9400 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
9401 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
9402 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
9403 search at the associated MODE_INT.
9405 2018-02-20 Jeff Law <law@redhat.com>
9408 PR tree-optimization/81592
9410 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
9411 for range data rather than using global data.
9412 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
9413 range data rather than using global data.
9414 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
9415 pass it to children as needed.
9416 (struct directive::fmtresult): Similarly.
9417 (struct directive::set_width): Similarly.
9418 (struct directive::set_precision): Similarly.
9419 (format_integer, format_directive, parse_directive): Similarly.
9420 (format_none): Accept unnamed vr_values parameter.
9421 (format_percent, format_floating, format_character): Similarly.
9422 (format_string, format_plain): Similarly.
9423 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
9424 the EVRP range analyzer for range data rather than using global data.
9425 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
9426 gimple-ssa-evrp-analyze.h
9427 (class sprintf_dom_walker): Add after_dom_children member function.
9428 Add evrp_range_analyzer member.
9429 (sprintf_dom_walker::before_dom_children): Call into the EVRP
9430 range analyzer as needed.
9431 (sprintf_dom_walker::after_dom_children): New member function.
9432 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
9434 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
9435 (evrp_range_analyzer::pop_to_marker): Likewise.
9437 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
9439 PR tree-optimization/84419
9440 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
9441 with the required type if its current type is compatible but
9444 2018-02-20 Jakub Jelinek <jakub@redhat.com>
9447 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
9448 after vectorization.
9450 2018-02-20 Martin Liska <mliska@suse.cz>
9453 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
9454 possible values if we don't have a hint.
9456 2018-02-20 Martin Liska <mliska@suse.cz>
9460 * final.c (shorten_branches): Build align_tab array with one
9462 * opts.c (finish_options): Add alignment option limit check.
9463 (MAX_CODE_ALIGN): Likewise.
9464 (MAX_CODE_ALIGN_VALUE): Likewise.
9465 * doc/invoke.texi: Document maximum allowed option value for
9466 all -falign-* options.
9468 2018-02-19 Jakub Jelinek <jakub@redhat.com>
9471 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
9472 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
9473 * var-tracking.c (emit_note_insn_var_location): Remove all references
9474 to NOTE_INSN_CALL_ARG_LOCATION.
9475 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
9476 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
9477 Use copy_rtx_if_shared.
9478 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
9479 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
9480 (dwarf2out_var_location): Remove handling of
9481 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
9483 * final.c (final_scan_insn): Remove all references to
9484 NOTE_INSN_CALL_ARG_LOCATION.
9485 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
9486 before dumping final insns.
9487 * except.c (emit_note_eh_region_end): Remove all references to
9488 NOTE_INSN_CALL_ARG_LOCATION.
9489 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
9490 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
9491 * config/arc/arc.c (hwloop_optimize): Likewise.
9492 * config/arm/arm.c (create_fix_barrier): Likewise.
9493 * config/s390/s390.c (s390_chunkify_start): Likewise.
9494 * config/sh/sh.c (find_barrier): Likewise.
9495 * config/i386/i386.c (rest_of_insert_endbranch,
9496 ix86_seh_fixup_eh_fallthru): Likewise.
9497 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
9498 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
9499 * config/frv/frv.c (frv_function_prologue): Likewise.
9500 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
9502 (note_outside_basic_block_p): Remove all references to
9503 NOTE_INSN_CALL_ARG_LOCATION.
9504 * gengtype.c (adjust_field_rtx_def): Likewise.
9505 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
9507 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
9508 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
9511 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
9514 PR tree-optimization/84452
9515 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
9516 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
9519 2018-02-19 Martin Liska <mliska@suse.cz>
9522 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
9524 2018-02-19 Martin Liska <mliska@suse.cz>
9525 Richard Sandiford <richard.sandiford@linaro.org>
9527 PR tree-optimization/82491
9528 * gimple-fold.c (get_base_constructor): Make earlier bail out
9531 2018-02-19 Carl Love <cel@us.ibm.com>
9533 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
9534 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
9536 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
9537 P8V_BUILTIN_VEC_NEG.
9539 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
9541 * config/rl78/rl78.md (movdf): New define expand.
9543 2018-02-19 Martin Liska <mliska@suse.cz>
9546 * doc/invoke.texi: Fix typo.
9547 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
9549 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
9551 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
9552 handle rs6000_single_float and rs6000_double_float specially for
9555 2018-02-16 Jeff Law <law@redhat.com>
9557 * config/rx/rx.c (add_pop_cfi_notes): New function.;
9560 2018-02-16 Jakub Jelinek <jakub@redhat.com>
9563 * ipa-inline.c (inline_small_functions): Fix a typo.
9565 2018-02-16 Nathan Sidwell <nathan@acm.org>
9567 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
9569 2018-02-16 Carl Love <cel@us.ibm.com>
9571 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
9572 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
9573 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
9574 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
9575 expansion to P8V_BUILTIN_VEC_FLOAT2.
9577 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
9579 PR rtl-optimization/70023
9580 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
9581 src_regno into account.
9583 2018-02-16 Carl Love <cel@us.ibm.com>
9585 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
9586 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
9587 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
9588 * config/rs6000/rs6000.c: Remove case statements for
9589 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
9590 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
9591 and P9V_BUILTIN_VEC_VINSERT4B.
9592 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
9593 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
9594 * config/rs6000/vsx.md:
9595 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
9598 2018-02-16 Carl Love <cel@us.ibm.com>
9600 * config/rs6000/altivec.h: Add builtin names vec_extract4b
9602 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
9604 * config/rs6000/rs6000-c.c: Add the definitions for
9605 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
9606 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
9607 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
9608 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
9609 definition for insert4b and define insn *insert3b_internal.
9610 * doc/extend.texi: Add documentation for vec_extract4b.
9612 2018-02-16 Nathan Sidwell <nathan@acm.org>
9614 * doc/extend.texi (Backwards Compatibility): Mention friend
9615 injection. Note for-scope is deprecated.
9616 * doc/invoke.texi (-ffriend-injection): Deprecate.
9618 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
9620 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
9621 that moved to I2, also allow destinations that are a paradoxical
9622 subreg (instead of a normal reg).
9624 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
9627 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
9630 2018-02-16 Richard Biener <rguenther@suse.de>
9632 PR tree-optimization/84037
9633 PR tree-optimization/84016
9635 * config/i386/i386.c (ix86_builtin_vectorization_cost):
9636 Adjust vec_construct for the fact we need additional higher latency
9637 128bit inserts for AVX256 and AVX512 vector builds.
9638 (ix86_add_stmt_cost): Scale vector construction cost for
9641 2018-02-16 Richard Biener <rguenther@suse.de>
9643 PR tree-optimization/84417
9644 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
9645 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
9646 (non_rewritable_lvalue_p): Likewise, use poly-ints.
9648 2018-02-16 Martin Liska <mliska@suse.cz>
9651 * internal-fn.def (ASAN_CHECK): Set proper flags.
9652 (ASAN_MARK): Likewise.
9654 2018-02-16 Julia Koval <julia.koval@intel.com>
9656 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
9657 from PTA_CANNONLAKE.
9659 2018-02-16 Jakub Jelinek <jakub@redhat.com>
9662 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
9663 Use ++iter rather than iter++ for std::list iterators.
9664 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
9665 defer deleting them until all nodes in the forest are processed. Do
9666 free even leaf nodes. Change to_process into auto_vec.
9669 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
9670 * vec.h (vec_default_construct): Use memset instead of placement new
9671 if BROKEN_VALUE_INITIALIZATION is defined.
9672 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
9673 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
9676 PR rtl-optimization/83723
9677 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
9678 * lra.c (lra_substitute_pseudo): Likewise. If true, use
9679 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
9681 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
9683 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
9685 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
9687 PR rtl-optimization/81443
9688 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
9689 from inner REGs to paradoxical SUBREGs.
9691 2018-02-16 Richard Biener <rguenther@suse.de>
9693 PR tree-optimization/84399
9694 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
9695 For operands we can analyze at their definition make sure we can
9696 analyze them at each use as well.
9698 2018-02-16 Richard Biener <rguenther@suse.de>
9700 PR tree-optimization/84190
9701 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
9702 volatile accesses if the decl isn't volatile.
9704 2018-02-15 Jason Merrill <jason@redhat.com>
9706 PR c++/84314 - ICE with templates and fastcall attribute.
9707 * attribs.c (build_type_attribute_qual_variant): Don't clobber
9708 TYPE_CANONICAL on an existing type.
9710 2018-02-15 Jakub Jelinek <jakub@redhat.com>
9712 PR tree-optimization/84383
9713 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
9714 dstoff nor call operand_equal_p if dstbase is NULL.
9716 PR tree-optimization/84334
9717 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
9718 also a CONSTANT_CLASS_P, punt.
9720 2018-02-14 Jim Wilson <jimw@sifive.com>
9722 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
9723 first SMALL_OPERAND check. New local min_second_step. Move assert
9724 to where locals are set. Add TARGET_RVC support.
9725 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
9727 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
9729 * doc/invoke.texi: Correct -Wformat-overflow code sample.
9731 2018-02-14 Martin Sebor <msebor@redhat.com>
9733 PR tree-optimization/83698
9734 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
9735 arrays constrain the offset range to their bounds.
9736 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
9737 (builtin_access::overlap): Avoid setting the size of overlap if it's
9739 (maybe_diag_overlap): Also consider arrays when deciding what values
9740 of offsets to include in diagnostics.
9742 2018-02-14 Martin Sebor <msebor@redhat.com>
9745 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
9746 that correspond to the kind of a declaration.
9748 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
9751 * config/pa/pa.md: Load address of PIC label using the linkage table
9752 if the label is nonlocal.
9754 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
9756 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
9757 warning message if user requests -maltivec=be.
9758 * doc/invoke.texi: Document deprecation of -maltivec=be.
9760 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
9763 * config/rs6000/rs6000-c.c: Update definitions for
9764 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
9765 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
9767 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9770 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
9771 add _get_ssp intrinsics. Remove argument from
9772 __builtin_ia32_rdssp[d|q].
9773 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
9774 * config/i386/i386-builtin.def: Remove argument from
9775 __builtin_ia32_rdssp[d|q].
9776 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
9777 ix86_expand_special_args_builtin for _rdssp[d|q].
9778 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
9779 Clear register before usage.
9780 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
9781 Add documentation for new _get_ssp and _inc_ssp intrinsics.
9783 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
9785 PR tree-optimization/84357
9786 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
9787 operand 1 of an ARRAY_REF too.
9789 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
9792 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
9793 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
9795 (set_of_reg): New struct.
9796 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
9797 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
9798 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
9800 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
9801 Split into bitclr, bitset, bitinvert patterns if appropriate.
9802 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
9803 use rx_fuse_in_memory_bitop.
9804 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
9805 to named insn, correct maximum insn length.
9807 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
9810 * machmode.def: Define a complex mode for PARTIAL_INT.
9811 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
9813 * doc/rtl.texi: Document CSPImode.
9814 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
9816 (msp430_hard_regno_nregs_with_padding): Likewise.
9818 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
9821 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
9823 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
9825 PR rtl-optimization/84169
9826 * combine.c (try_combine): New variable split_i2i3. Set it to true if
9827 we generated a parallel as new i3 and we split that to new i2 and i3
9828 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
9829 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
9830 those to i2, not i1. Partially rewrite this scan code.
9832 2018-02-13 Jakub Jelinek <jakub@redhat.com>
9835 * stor-layout.c (place_field): For variable length fields, adjust
9836 offset_align afterwards not just based on the field's alignment,
9837 but also on the size.
9840 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
9841 of exps and logs in the use_exp2 case.
9843 2018-02-13 Jeff Law <law@redhat.com>
9845 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
9848 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
9851 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
9855 * common.opt (gas-loc-support, gas-locview-support): New.
9856 (ginline-points, ginternal-reset-location-views): New.
9857 * doc/invoke.texi: Document them. Use @itemx where intended.
9858 (gvariable-location-views): Adjust.
9859 * target.def (reset_location_view): New.
9860 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
9861 (TARGET_RESET_LOCATION_VIEW): New.
9862 * doc/tm.texi: Rebuilt.
9863 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
9864 (dwarf2out_default_as_locview_support): New.
9865 (output_asm_line_debug_info): Use option variables.
9866 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
9867 (output_loc_list): Likewise.
9868 (add_high_low_attributes): Check option variables.
9869 Don't output entry view attribute in strict mode.
9870 (gen_inlined_subroutine_die): Check option variables.
9871 (dwarf2out_inline_entry): Likewise.
9872 (init_sections_and_labels): Likewise.
9873 (dwarf2out_early_finish): Likewise.
9874 (maybe_reset_location_view): New, from...
9875 (dwarf2out_var_location): ... here. Call it.
9876 * debug.h (dwarf2out_default_as_loc_support): Declare.
9877 (dwarf2out_default_as_locview_support): Declare.
9878 * hooks.c (hook_int_rtx_insn_0): New.
9879 * hooks.h (hook_int_rtx_insn_0): Declare.
9880 * toplev.c (process_options): Take -gas-loc-support and
9881 -gas-locview-support from dwarf2out. Enable
9882 -gvariable-location-views by default only with locview
9883 assembler support. Enable -ginternal-reset-location-views by
9884 default only if the target defines the corresponding hook.
9885 Enable -ginline-points by default if location views are
9886 enabled; force it disabled if statement frontiers are
9888 * tree-inline.c (expand_call_inline): Check option variables.
9889 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
9891 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
9893 PR tree-optimization/84321
9894 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
9895 handling. Also check whether the anti-range contains any values
9896 that satisfy the mask; switch to a VR_RANGE if not.
9898 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
9901 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
9903 2018-02-13 Martin Jambor <mjambor@suse.cz>
9906 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
9907 of call statements, also set location of a load to a temporary.
9909 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
9911 * config/rl78/rl78.c (add_vector_labels): New function.
9912 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
9913 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
9914 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
9915 which checks that no arguments are passed.
9916 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
9917 * doc/extend.texi: Documentation for the new attribute.
9919 2018-02-13 Andreas Schwab <schwab@suse.de>
9921 * config/riscv/linux.h (CPP_SPEC): Define.
9923 2018-02-13 Jakub Jelinek <jakub@redhat.com>
9926 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
9927 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
9928 OPTION_MASK_ISA_AES as first argument to def_builtin_const
9929 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
9930 instead of OPTION_MASK_ISA_PCLMUL as first argument to
9931 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
9932 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
9933 temporarily for AES and PCLMUL builtins.
9935 PR tree-optimization/84339
9936 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
9937 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
9941 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
9942 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
9943 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
9945 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
9947 * omp-simd-clone.h: New file.
9948 * omp-simd-clone.c: Include omp-simd-clone.h.
9949 (expand_simd_clones): No longer static.
9950 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
9951 cgraph.h and omp-simd-clone.h.
9952 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
9953 (vect_recog_widen_shift_pattern): Formatting fix.
9954 (vect_pattern_recog_1): Don't check optab for calls.
9957 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
9958 operands[2] into a REG before using gen_lowpart on it.
9960 2018-02-12 Jeff Law <law@redhat.com>
9963 * config/sh/sh.c (find_barrier): Consider a sibling call
9966 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
9967 successfully back substituting a reg.
9969 2018-02-12 Richard Biener <rguenther@suse.de>
9971 PR tree-optimization/84037
9972 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
9973 parameter, move visited init to caller.
9974 (vect_slp_analyze_operations): Separate cost from validity
9975 check, initialize visited once for all instances.
9976 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
9978 * tree-vect-stmts.c (vect_model_simple_cost): Make early
9980 (vect_model_promotion_demotion_cost): Likewise.
9981 (vectorizable_bswap): Guard cost modeling with !slp_node
9982 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
9984 (vectorizable_call): Likewise.
9985 (vectorizable_conversion): Likewise.
9986 (vectorizable_assignment): Likewise.
9987 (vectorizable_shift): Likewise.
9988 (vectorizable_operation): Likewise.
9989 (vectorizable_store): Likewise.
9990 (vectorizable_load): Likewise.
9991 (vectorizable_condition): Likewise.
9992 (vectorizable_comparison): Likewise.
9994 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
9997 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
9998 (ASAN_MARK): Fix fnspec to account for return value, change pointer
9999 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
10001 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
10003 PR middle-end/83665
10004 * params.def (inline-min-speedup): Increase from 8 to 15.
10005 (max-inline-insns-auto): Decrease from 40 to 30.
10006 * ipa-split.c (consider_split): Add some buffer for function to
10007 be considered inlining candidate.
10008 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
10011 2018-02-12 Richard Biener <rguenther@suse.de>
10013 PR tree-optimization/84037
10014 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
10015 matched stmts if we cannot swap the non-matched ones.
10017 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10019 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
10020 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
10021 _mm_maskz_scalef_round_ss): New intrinsics.
10022 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
10023 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
10024 __builtin_ia32_scalefss_round): Remove.
10025 (__builtin_ia32_scalefsd_mask_round,
10026 __builtin_ia32_scalefss_mask_round): New intrinsics.
10027 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
10028 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
10029 ((match_operand:VF_128 2 "<round_nimm_predicate>"
10030 "<round_constraint>")): Changed to ...
10031 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
10032 "<round_scalar_constraint>")): ... this.
10033 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
10034 %0, %1, %2<round_op3>}"): Changed to ...
10035 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
10036 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
10037 %2<round_scalar_mask_op3>}"): ... this.
10038 * config/i386/subst.md (round_scalar_nimm_predicate): New.
10040 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10042 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
10043 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
10044 (_mm_maskz_sqrt_round_ss): New intrinsics.
10045 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
10046 (__builtin_ia32_sqrtsd_mask_round)
10047 (__builtin_ia32_sqrtss_mask_round): New builtins.
10048 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
10049 (__builtin_ia32_sqrtss_round): Remove.
10050 (__builtin_ia32_sqrtsd_mask_round)
10051 (__builtin_ia32_sqrtss_mask_round): New builtins.
10052 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
10053 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
10054 ((match_operand:VF_128 1 "vector_operand"
10055 "xBm,<round_constraint>")): Changed to ...
10056 ((match_operand:VF_128 1 "vector_operand"
10057 "xBm,<round_scalar_constraint>")): ... this.
10058 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
10059 %0, %2, %<iptr>1<round_op3>}): Changed to ...
10060 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
10061 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
10062 %<iptr>1<round_scalar_mask_op3>}): ... this.
10063 ((set_attr "prefix" "<round_prefix>")): Changed to ...
10064 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
10066 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
10069 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
10070 Cast vec_cmpeq result to correct type.
10071 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
10072 Cast vec_cmpgt result to correct type.
10074 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
10076 * final.c (final_scan_insn_1): Renamed from...
10077 (final_scan_insn): ... this. New wrapper, to recover
10078 seen from the outermost call in recursive ones.
10079 * config/sparc/sparc.c (output_return): Drop seen from call.
10080 (output_sibcall): Likewise.
10081 * config/visium/visium.c (output_branch): Likewise.
10083 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
10085 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
10088 2018-02-10 Alan Modra <amodra@gmail.com>
10091 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
10092 Specify LR as an input.
10094 2018-02-10 Jakub Jelinek <jakub@redhat.com>
10097 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
10098 remove_member_access_dummy_vars): New functions.
10099 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
10100 lower_omp_1, execute_lower_omp): Use them.
10102 PR rtl-optimization/84308
10103 * shrink-wrap.c (spread_components): Release todo vector.
10105 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
10107 PR rtl-optimization/57193
10108 * ira-color.c (struct allocno_color_data): Add member
10109 conflict_allocno_hard_prefs.
10110 (update_conflict_allocno_hard_prefs): New.
10111 (bucket_allocno_compare_func): Add a preference based on
10112 conflict_allocno_hard_prefs.
10113 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
10114 (color_allocnos): Remove a dead code. Initiate
10115 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
10117 2018-02-09 Jakub Jelinek <jakub@redhat.com>
10120 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
10121 constraint from =wa to wa. Avoid a subreg on the output operand,
10122 instead use a pseudo and subreg it in a move.
10123 (p9_xxbrd_<mode>): Changed to ...
10124 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
10125 (p9_xxbrd_v2df): New expander.
10126 (p9_xxbrw_<mode>): Changed to ...
10127 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
10128 (p9_xxbrw_v4sf): New expander.
10130 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
10132 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
10134 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
10137 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
10138 multiply in 32-bit mode.
10139 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
10140 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
10143 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
10145 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
10146 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
10147 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
10148 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
10150 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
10153 * dwarf2out.c (is_trivial_indirect_ref): New function.
10154 (dwarf2out_late_global_decl): Do not generate a location
10155 attribute for variables that have a non-trivial DECL_VALUE_EXPR
10156 and that are not defined in the current unit.
10158 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
10160 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
10161 instead of a libcall for UNORDERED.
10163 2018-02-09 Tamar Christina <tamar.christina@arm.com>
10166 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
10167 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
10169 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10172 * config/s390/s390.c (s390_set_current_function): Invoke
10173 s390_indirect_branch_settings also if fndecl didn't change.
10175 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
10177 * config/rs6000/rs6000.md (blockage): Set length to zero.
10179 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
10181 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
10183 2018-02-09 Jakub Jelinek <jakub@redhat.com>
10186 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
10187 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
10191 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
10192 PARALLEL incoming that failed vt_get_decl_and_offset check.
10194 PR middle-end/84237
10195 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
10196 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
10198 (get_variable_section): For decls in named .bss* sections pass true as
10199 second argument to bss_initializer_p.
10201 2018-02-09 Marek Polacek <polacek@redhat.com>
10202 Jakub Jelinek <jakub@redhat.com>
10205 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
10206 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
10207 Sync some changes from cxx_fold_indirect_ref.
10209 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
10211 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
10213 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
10214 (BLOCK_INLINE_ENTRY_LABEL): New.
10215 (dwarf2out_var_location): Disregard inline entry markers.
10216 (inline_entry_data): New struct.
10217 (inline_entry_data_hasher): New hashtable type.
10218 (inline_entry_data_hasher::hash): New.
10219 (inline_entry_data_hasher::equal): New.
10220 (inline_entry_data_table): New variable.
10221 (add_high_low_attributes): Add DW_AT_entry_pc and
10222 DW_AT_GNU_entry_view attributes if a pending entry is found
10223 in inline_entry_data_table. Add old entry_pc attribute only
10224 if debug nonbinding markers are disabled.
10225 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
10226 markers are enabled.
10227 (block_within_block_p, dwarf2out_inline_entry): New.
10228 (dwarf2out_finish): Check that no entries remained in
10229 inline_entry_data_table.
10230 * final.c (reemit_insn_block_notes): Handle inline entry notes.
10231 (final_scan_insn, notice_source_line): Likewise.
10232 (rest_of_clean_state): Skip inline entry markers.
10233 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
10235 * gimple.c (gimple_build_debug_inline_entry): New.
10236 * gimple.h (enum gimple_debug_subcode): Add
10237 GIMPLE_DEBUG_INLINE_ENTRY.
10238 (gimple_build_debug_inline_entry): Declare.
10239 (gimple_debug_inline_entry_p): New.
10240 (gimple_debug_nonbind_marker_p): Adjust.
10241 * insn-notes.def (INLINE_ENTRY): New.
10242 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
10243 inline entry marker notes.
10244 (print_insn): Likewise.
10245 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
10246 (INSN_DEBUG_MARKER_KIND): Likewise.
10247 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
10248 * tree-inline.c (expand_call_inline): Build and insert
10249 debug_inline_entry stmt.
10250 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
10251 inline entry blocks early, if nonbind markers are enabled.
10252 (dump_scope_block): Dump fragment info.
10253 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
10254 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
10255 (gimple_build_debug_inline_entry): New.
10256 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
10257 Enable/disable inline entry points too.
10258 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
10259 (DEBUG_INSN): Describe inline entry markers.
10261 * common.opt (gvariable-location-views): New.
10262 (gvariable-location-views=incompat5): New.
10263 * config.in: Rebuilt.
10264 * configure: Rebuilt.
10265 * configure.ac: Test assembler for view support.
10266 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
10267 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
10268 * dwarf2out.c (var_loc_view): New typedef.
10269 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
10270 (dwarf2out_locviews_in_attribute): New.
10271 (dwarf2out_locviews_in_loclist): New.
10272 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
10273 (enum dw_line_info_opcode): Add LI_adv_address.
10274 (struct dw_line_info_table): Add view.
10275 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
10276 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
10277 (zero_view_p): New variable.
10278 (ZERO_VIEW_P): New macro.
10279 (output_asm_line_debug_info): New.
10280 (struct var_loc_node): Add view.
10281 (add_AT_view_list, AT_loc_list): New.
10282 (add_var_loc_to_decl): Add view param. Test it against last.
10283 (new_loc_list): Add view params. Record them.
10284 (AT_loc_list_ptr): Handle loc and view lists.
10285 (view_list_to_loc_list_val_node): New.
10286 (print_dw_val): Handle dw_val_class_view_list.
10287 (size_of_die): Likewise.
10288 (value_format): Likewise.
10289 (loc_list_has_views): New.
10290 (gen_llsym): Set vl_symbol too.
10291 (maybe_gen_llsym, skip_loc_list_entry): New.
10292 (dwarf2out_maybe_output_loclist_view_pair): New.
10293 (output_loc_list): Output view list or entries too.
10294 (output_view_list_offset): New.
10295 (output_die): Handle dw_val_class_view_list.
10296 (output_dwarf_version): New.
10297 (output_compilation_unit_header): Use it.
10298 (output_skeleton_debug_sections): Likewise.
10299 (output_rnglists, output_line_info): Likewise.
10300 (output_pubnames, output_aranges): Update version comments.
10301 (output_one_line_info_table): Output view numbers in asm comments.
10302 (dw_loc_list): Determine current endview, pass it to new_loc_list.
10303 Call maybe_gen_llsym.
10304 (loc_list_from_tree_1): Adjust.
10305 (add_AT_location_description): Create view list attribute if
10306 needed, check it's absent otherwise.
10307 (convert_cfa_to_fb_loc_list): Adjust.
10308 (maybe_emit_file): Call output_asm_line_debug_info for test.
10309 (dwarf2out_var_location): Reset views as needed. Precompute
10310 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
10311 attribute. Set view.
10312 (new_line_info_table): Reset next view.
10313 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
10314 (dwarf2out_source_line): Likewise. Output view resets and labels to
10315 the assembler, or select appropriate line info opcodes.
10316 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
10317 (optimize_string_length): Catch it. Adjust.
10318 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
10319 dw_val_class_view_list, and remove it if no longer needed.
10320 (hash_loc_list): Hash view numbers.
10321 (loc_list_hasher::equal): Compare them.
10322 (optimize_location_lists): Check whether a view list symbol is
10323 needed, and whether the locview attribute is present, and
10324 whether they match. Remove the locview attribute if no longer
10326 (index_location_lists): Call skip_loc_list_entry for test.
10327 (dwarf2out_finish): Call output_asm_line_debug_info for test.
10328 Use output_dwarf_version.
10329 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
10330 (struct dw_val_node): Add val_view_list.
10331 * final.c (SEEN_NEXT_VIEW): New.
10332 (set_next_view_needed): New.
10333 (clear_next_view_needed): New.
10334 (maybe_output_next_view): New.
10335 (final_start_function): Rename to...
10336 (final_start_function_1): ... this. Take pointer to FIRST,
10337 add SEEN parameter. Emit param bindings in the initial view.
10338 (final_start_function): Reintroduce SEEN-less interface.
10339 (final): Rename to...
10340 (final_1): ... this. Take SEEN parameter. Output final pending
10341 next view at the end.
10342 (final): Reintroduce seen-less interface.
10343 (final_scan_insn): Output pending next view before switching
10344 sections or ending a block. Mark the next view as needed when
10345 outputting variable locations. Notify debug backend of section
10346 changes, and of location view changes.
10347 (rest_of_handle_final): Adjust.
10348 * toplev.c (process_options): Autodetect value for debug variable
10349 location views option. Warn on incompat5 without -gdwarf-5.
10350 * doc/invoke.texi (gvariable-location-views): New.
10351 (gvariable-location-views=incompat5): New.
10352 (gno-variable-location-views): New.
10354 2018-02-08 David Malcolm <dmalcolm@redhat.com>
10356 PR tree-optimization/84136
10357 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
10358 that the result of find_edge is non-NULL.
10360 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
10363 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
10364 storing integer register in SImode. Fix cost of 256 and 512
10365 byte aligned SSE register store.
10367 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
10369 * config/i386/i386.c (ix86_multiplication_cost): Fix
10370 multiplication cost for TARGET_AVX512DQ.
10372 2018-02-08 Marek Polacek <polacek@redhat.com>
10374 PR tree-optimization/84238
10375 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
10378 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10380 PR tree-optimization/84265
10381 * tree-vect-stmts.c (vectorizable_store): Don't treat
10382 VMAT_CONTIGUOUS accesses as grouped.
10383 (vectorizable_load): Likewise.
10385 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10387 PR tree-optimization/81635
10388 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
10389 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
10390 (test_round_for_mask): New functions.
10391 (wide_int_cc_tests): Call test_round_for_mask.
10392 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
10393 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
10394 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
10395 range returned by get_range_info.
10397 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
10400 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
10401 * symtab.c: Include builtins.h
10402 (symtab_node::output_to_lto_symbol_table_p): Move here
10403 from lto-streamer-out.c:output_symbol_p.
10404 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
10405 (output_symbol_p): Move all logic to symtab.c
10406 (produce_symtab): Update.
10408 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10410 * config/s390/s390-opts.h (enum indirect_branch): Define.
10411 * config/s390/s390-protos.h (s390_return_addr_from_memory)
10412 (s390_indirect_branch_via_thunk)
10413 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
10414 (enum s390_indirect_branch_type): Define.
10415 * config/s390/s390.c (struct s390_frame_layout, struct
10416 machine_function): Remove.
10417 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
10418 (indirect_branch_table_label_no, indirect_branch_table_name):
10420 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
10421 (enum s390_indirect_branch_option): Define.
10422 (s390_return_addr_from_memory): New function.
10423 (s390_handle_string_attribute): New function.
10424 (s390_attribute_table): Add new attribute handler.
10425 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
10426 (s390_indirect_branch_via_thunk): New function.
10427 (s390_indirect_branch_via_inline_thunk): New function.
10428 (s390_function_ok_for_sibcall): When jumping via thunk disallow
10429 sibling call optimization for non z10 compiles.
10430 (s390_emit_call): Force indirect branch target to be a single
10431 register. Add r1 clobber for non-z10 compiles.
10432 (s390_emit_epilogue): Emit return jump via return_use expander.
10433 (s390_reorg): Handle JUMP_INSNs as execute targets.
10434 (s390_option_override_internal): Perform validity checks for the
10435 new command line options.
10436 (s390_indirect_branch_attrvalue): New function.
10437 (s390_indirect_branch_settings): New function.
10438 (s390_set_current_function): Invoke s390_indirect_branch_settings.
10439 (s390_output_indirect_thunk_function): New function.
10440 (s390_code_end): Implement target hook.
10441 (s390_case_values_threshold): Implement target hook.
10442 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
10444 * config/s390/s390.h (struct s390_frame_layout)
10445 (struct machine_function): Move here from s390.c.
10446 (TARGET_INDIRECT_BRANCH_NOBP_RET)
10447 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
10448 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
10449 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
10450 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
10451 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
10452 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
10453 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
10454 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
10455 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
10456 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
10457 (mnemonic attribute): Add values which aren't recognized
10459 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
10460 pattern for branch conversion. Fix mnemonic attribute.
10461 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
10462 indirect branch via thunk if requested.
10463 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
10464 ("*indirect_jump"): Disable for branch conversion using out of
10466 ("indirect_jump_via_thunk<mode>_z10")
10467 ("indirect_jump_via_thunk<mode>")
10468 ("indirect_jump_via_inlinethunk<mode>_z10")
10469 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
10470 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
10471 ("casesi_jump_via_inlinethunk<mode>_z10")
10472 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
10473 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
10474 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
10475 ("*indirect2_jump"): Disable for branch conversion.
10476 ("casesi_jump"): Turn into expander and expand patterns for branch
10478 ("return_use"): New expander.
10479 ("*return"): Emit return via thunk and rename it to ...
10480 ("*return<mode>"): ... this one.
10481 * config/s390/s390.opt: Add new options and and enum for the
10484 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10486 * lra-constraints.c (match_reload): Unconditionally use
10487 gen_lowpart_SUBREG, rather than selecting between that
10488 and equivalent gen_rtx_SUBREG code.
10490 2018-02-08 Richard Biener <rguenther@suse.de>
10492 PR tree-optimization/84233
10493 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
10494 changed flag instead of boguously re-using phi_inserted.
10496 2018-02-08 Martin Jambor <mjambor@suse.cz>
10498 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
10499 static local variables.
10501 2018-02-08 Richard Biener <rguenther@suse.de>
10503 PR tree-optimization/84278
10504 * tree-vect-stmts.c (vectorizable_store): When looking for
10505 smaller vector types to perform grouped strided loads/stores
10506 make sure the mode is supported by the target.
10507 (vectorizable_load): Likewise.
10509 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
10511 * config/aarch64/aarch64.c (aarch64_components_for_bb):
10512 Increase LDP/STP opportunities by adding adjacent callee-saves.
10514 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
10516 PR rtl-optimization/84068
10517 PR rtl-optimization/83459
10518 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
10520 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
10522 PR tree-optimization/84224
10523 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
10524 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
10525 non-zero arguments.
10527 2018-02-07 Iain Sandoe <iain@codesourcery.com>
10530 * config/rs6000/altivec.md (*restore_world): Remove LR use.
10531 * config/rs6000/predicates.md (restore_world_operation): Adjust op
10532 count, remove one USE.
10534 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
10536 * doc/install.texi (Configuration): Document the
10537 --with-long-double-format={ibm,ieee} PowerPC configuration
10541 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
10542 Convert from define_expand to be define_insn_and_split. Rework
10543 float/double/_Float128 conversions to QI/HI/SImode to work with
10544 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
10545 conversions to QI/HImode types did a store and then a load to
10546 truncate the value. For conversions to VSX registers, don't split
10547 the insn, instead emit the code directly. Use the code iterator
10548 any_fix to combine signed and unsigned conversions.
10549 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
10550 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
10551 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
10552 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
10553 (fix_<mode>di2_hw): Likewise.
10554 (fixuns_<mode>di2_hw): Likewise.
10555 (fix_<mode>si2_hw): Likewise.
10556 (fixuns_<mode>si2_hw): Likewise.
10557 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
10558 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
10559 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
10560 fix<uns>_trunc<SFDF:mode>si2_p8.
10561 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
10562 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
10563 (fix<uns>_<mode>_mem): Likewise.
10564 (fctiw<u>z_<mode>_mem): Likewise.
10565 (fix<uns>_<mode>_mem): Likewise.
10566 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
10567 the register allocator from doing a direct move to the GPRs to do
10568 a store, and instead use the ISA 3.0 store byte/half-word from
10569 vector register instruction. For IEEE 128-bit floating point,
10570 also optimize stores of 32-bit ints.
10571 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
10573 2018-02-07 Alan Hayward <alan.hayward@arm.com>
10575 * genextract.c (push_pathstr_operand): New function to support
10577 (walk_rtx): Call push_pathstr_operand.
10578 (print_path): Support [a-zA-Z].
10580 2018-02-07 Richard Biener <rguenther@suse.de>
10582 PR tree-optimization/84037
10583 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
10584 (cse_and_gimplify_to_preheader): Declare.
10585 (vect_get_place_in_interleaving_chain): Likewise.
10586 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10588 (_loop_vec_info::~_loop_vec_info): Delete it.
10589 (cse_and_gimplify_to_preheader): New function.
10590 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
10591 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
10592 (vectorizable_load): Likewise. For grouped stores always base
10593 the IV on the first element.
10594 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
10595 condition before gimplifying.
10597 2018-02-07 Jakub Jelinek <jakub@redhat.com>
10599 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
10600 *DIV_EXPR and *MOD_EXPR.
10602 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
10605 * config/i386/i386.c (ix86_option_override_internal): Mask out
10606 the CF_SET bit when checking -fcf-protection.
10608 2018-02-07 Tom de Vries <tom@codesourcery.com>
10611 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
10614 2018-02-07 Richard Biener <rguenther@suse.de>
10616 PR tree-optimization/84204
10617 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
10620 PR tree-optimization/84205
10621 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
10622 special-case isl_ast_op_zdiv_r.
10624 PR tree-optimization/84223
10625 * graphite-scop-detection.c (gather_bbs::before_dom_children):
10626 Only add conditions from within the region.
10627 (gather_bbs::after_dom_children): Adjust.
10629 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
10632 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
10633 * config/avr/avr.md: Only post-reload split REG-REG moves if
10634 either register is GENERAL_REG_P.
10636 2018-02-07 Jakub Jelinek <jakub@redhat.com>
10638 PR tree-optimization/84235
10639 * tree-ssa-scopedtables.c
10640 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
10641 if the subtraction is performed in floating point type where NaNs are
10642 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
10643 build 1. Formatting fix.
10645 2018-02-06 Jakub Jelinek <jakub@redhat.com>
10648 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
10649 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
10650 and skip it regardless of bb boundaries. Use CALL_P macro,
10651 don't test INSN_P (insn) together with CALL_P or JUMP_P check
10652 unnecessarily, formatting fix.
10654 2018-02-06 Michael Collison <michael.collison@arm.com>
10656 * config/arm/thumb2.md:
10657 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
10658 (*thumb_mov_notscc): Ditto.
10660 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
10663 * config/rs6000/rs6000.md (su code attribute): Use "u" for
10664 unsigned_fix, not "s".
10666 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10668 * configure.ac (gcc_fn_eh_frame_ro): New function.
10669 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
10670 correct .eh_frame permissions.
10671 * configure: Regenerate.
10673 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
10675 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
10676 irrelevant options.
10678 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10680 * config/rs6000/rs6000.c (rs6000_option_override_internal):
10681 Display warning message for -mno-speculate-indirect-jumps.
10683 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
10685 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
10687 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
10689 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
10691 PR tree-optimization/84225
10692 * tree-eh.c (find_trapping_overflow): Only call
10693 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
10695 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10698 * config/i386/i386.c: Reimplement the check of possible options
10699 -mibt/-mshstk conbination. Change error messages.
10700 * doc/invoke.texi: Fix a typo: remove extra '='.
10702 2018-02-06 Marek Polacek <polacek@redhat.com>
10704 PR tree-optimization/84228
10705 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
10707 2018-02-06 Tamar Christina <tamar.christina@arm.com>
10710 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
10711 emitted arch directives.
10712 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
10713 __ARM_FEATURE_COPROC before changing architectures.
10715 2018-02-06 Richard Biener <rguenther@suse.de>
10717 * config/i386/i386.c (print_reg): Fix typo.
10718 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
10720 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
10722 * configure: Regenerate.
10724 2018-02-05 Martin Sebor <msebor@redhat.com>
10726 PR tree-optimization/83369
10727 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
10730 2018-02-05 Martin Liska <mliska@suse.cz>
10732 * doc/invoke.texi: Cherry-pick upstream r323995.
10734 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
10736 * ira.c (ira_init_register_move_cost): Adjust comment.
10738 2018-02-05 Martin Liska <mliska@suse.cz>
10740 PR gcov-profile/84137
10741 * doc/gcov.texi: Fix typo in documentation.
10743 2018-02-05 Martin Liska <mliska@suse.cz>
10745 PR gcov-profile/83879
10746 * doc/gcov.texi: Document necessity of --dynamic-list-data when
10747 using dlopen functionality.
10749 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
10751 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
10752 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
10753 _mm_maskz_range_ss, _mm_mask_range_round_ss,
10754 _mm_maskz_range_round_ss): New intrinsics.
10755 (__builtin_ia32_rangesd128_round)
10756 (__builtin_ia32_rangess128_round): Remove.
10757 (__builtin_ia32_rangesd128_mask_round,
10758 __builtin_ia32_rangess128_mask_round): New builtins.
10759 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
10760 __builtin_ia32_rangess128_round): Remove.
10761 (__builtin_ia32_rangesd128_mask_round,
10762 __builtin_ia32_rangess128_mask_round): New builtins.
10763 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
10764 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
10765 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
10766 "<round_saeonly_constraint>")): Changed to ...
10767 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
10768 "<round_saeonly_scalar_constraint>")): ... this.
10769 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
10770 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
10771 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
10772 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
10773 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
10775 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
10777 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
10779 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
10780 Remove all values except native, 8540 and 8548.
10782 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
10784 * config/i386/i386.c (ix86_output_function_return): Pass
10785 INVALID_REGNUM, instead of -1, as invalid register number to
10786 indirect_thunk_name and output_indirect_thunk.
10788 2018-02-02 Julia Koval <julia.koval@intel.com>
10790 * config.gcc: Add -march=icelake.
10791 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
10792 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
10793 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
10794 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
10795 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
10796 (processor_target_table): Add icelake.
10797 (ix86_option_override_internal): Handle new PTAs.
10798 (get_builtin_code_for_version): Handle icelake.
10799 (M_INTEL_COREI7_ICELAKE): New.
10800 (fold_builtin_cpu): Handle icelake.
10801 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
10802 * doc/invoke.texi: Add -march=icelake.
10804 2018-02-02 Julia Koval <julia.koval@intel.com>
10806 * config/i386/i386.c (ix86_option_override_internal): Change flags type
10807 to wide_int_bitmask.
10808 * wide-int-bitmask.h: New.
10810 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10813 * config/i386/i386.md: Replace Pmode with word_mode in
10814 builtin_setjmp_setup and builtin_longjmp to support x32.
10816 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
10820 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
10822 (rs6000_supported_cpu_names): New static variable.
10823 (linux_cpu_translation_table): Likewise.
10824 (elf_platform) <cpu>: Define new static variable and use it.
10825 Translate kernel AT_PLATFORM name to canonical name if needed.
10826 Error if platform name is unknown.
10828 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
10831 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
10833 2018-02-01 Jeff Law <law@redhat.com>
10836 * config/i386/i386.c (release_scratch_register_on_entry): Add new
10837 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
10838 the scratch if RELEASE_VIA_POP is false.
10839 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
10840 If we have to save a temporary register, decrement SIZE appropriately.
10841 Pass new arguments to release_scratch_register_on_entry.
10842 (ix86_adjust_stack_and_probe): Likewise.
10843 (ix86_emit_probe_stack_range): Pass new arguments to
10844 release_scratch_register_on_entry.
10846 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
10848 PR rtl-optimization/84157
10849 * combine.c (change_zero_ext): Use REG_P predicate in
10850 front of HARD_REGISTER_P predicate.
10852 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
10854 * config/avr/avr.c (avr_option_override): Move disabling of
10855 -fdelete-null-pointer-checks to...
10856 * common/config/avr/avr-common.c (avr_option_optimization_table):
10859 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10861 PR tree-optimization/81635
10862 * tree-data-ref.c (split_constant_offset_1): For types that
10863 wrap on overflow, try to use range info to prove that wrapping
10866 2018-02-01 Renlin Li <renlin.li@arm.com>
10869 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
10870 TAILCALL_ADDR_REGS.
10871 (aarch64_register_move_cost): Likewise.
10872 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
10873 TAILCALL_ADDR_REGS.
10874 (REG_CLASS_NAMES): Likewise.
10875 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
10876 TAILCALL_ADDR_REGS. Remove IP registers.
10877 * config/aarch64/aarch64.md (Ucs): Update register constraint.
10879 2018-02-01 Richard Biener <rguenther@suse.de>
10881 * domwalk.h (dom_walker::dom_walker): Add additional constructor
10882 for specifying RPO order and allow NULL for that.
10883 * domwalk.c (dom_walker::dom_walker): Likewise.
10884 (dom_walker::walk): Handle NULL RPO order.
10885 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
10887 (rewrite_update_dom_walker): Likewise.
10888 (mark_def_dom_walker): Likewise.
10890 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10892 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
10893 (aarch64_maybe_expand_sve_subreg_move): Declare.
10894 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
10895 * config/aarch64/predicates.md (aarch64_any_register_operand): New
10897 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
10898 that are semantically a reverse operation.
10899 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
10900 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
10901 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
10903 (aarch64_can_change_mode_class): For big-endian, forbid changes
10904 between two SVE modes if they have different element sizes.
10906 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10908 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
10909 the TImode handling for big-endian targets.
10911 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10913 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
10914 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
10916 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
10917 Remove BSWAP handing for big-endian targets and use the form of
10918 LD1RQ appropariate for the mode.
10920 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10922 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
10923 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
10924 duplicated element.
10926 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10929 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
10930 check for operands that need to go through aarch64_sve_reload_be.
10932 2018-02-01 Jakub Jelinek <jakub@redhat.com>
10934 PR tree-optimization/81661
10935 PR tree-optimization/84117
10936 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
10937 * tree-eh.c: Include gimplify.h.
10938 (find_trapping_overflow, replace_trapping_overflow,
10939 rewrite_to_non_trapping_overflow): New functions.
10940 * tree-vect-loop.c: Include tree-eh.h.
10941 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
10942 * tree-data-ref.c: Include tree-eh.h.
10943 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
10945 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
10947 PR rtl-optimization/84123
10948 * combine.c (change_zero_ext): Check if hard register satisfies
10949 can_change_dest_mode before calling gen_lowpart_SUBREG.
10951 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
10954 * ira.c (ira_init_register_move_cost): Remove assert.
10956 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
10958 PR rtl-optimization/84071
10959 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
10960 * doc/tm.texi: Regenerate.
10962 2018-01-31 Richard Biener <rguenther@suse.de>
10964 PR tree-optimization/84132
10965 * tree-data-ref.c (analyze_miv_subscript): Properly
10966 check whether evolution_function_is_affine_multivariate_p
10967 before calling gcd_of_steps_may_divide_p.
10969 2018-01-31 Julia Koval <julia.koval@intel.com>
10972 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
10973 * config/i386/i386.md (rdpid_rex64) New.
10974 (rdpid): Make 32bit only.
10976 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
10979 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
10980 an IDENTIFIER_NODE for FUNCTION_TYPE's.
10982 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
10985 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
10987 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
10989 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
10991 PR rtl-optimization/84071
10992 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
10993 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
10995 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
10997 * config/arc/arc.c (arc_handle_aux_attribute): New function.
10998 (arc_attribute_table): Add 'aux' attribute.
10999 (arc_in_small_data_p): Consider aux like variables.
11000 (arc_is_aux_reg_p): New function.
11001 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
11002 (arc_get_aux_arg): New function.
11003 (prepare_move_operands): Handle aux-register access.
11004 (arc_handle_aux_attribute): New function.
11005 * doc/extend.texi (ARC Variable attributes): Add subsection.
11007 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11009 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
11010 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
11011 (arc_attribute_table): Add 'uncached' attribute.
11012 (arc_print_operand): Print '.di' flag for uncached memory
11014 (arc_in_small_data_p): Do not consider for small data the uncached
11016 (arc_is_uncached_mem_p): New function.
11017 * config/arc/predicates.md (compact_store_memory_operand): Check
11018 for uncached memory accesses.
11019 (nonvol_nonimm_operand): Likewise.
11020 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
11022 2018-01-31 Jakub Jelinek <jakub@redhat.com>
11025 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
11026 falign-loops=): Add Optimization flag.
11028 2018-01-30 Jeff Law <law@redhat.com>
11031 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
11032 INT_REGISTERS_SAVED. Check it prior to calling
11033 get_scratch_register_on_entry.
11034 (ix86_adjust_stack_and_probe): Similarly.
11035 (ix86_emit_probe_stack_range): Similarly.
11036 (ix86_expand_prologue): Corresponding changes.
11038 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11041 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
11042 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
11044 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
11047 * lra-constraints.c (curr_insn_transform): Process AND in the
11050 2018-01-30 Jakub Jelinek <jakub@redhat.com>
11052 PR rtl-optimization/83986
11053 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
11054 dependence against last_pending_memory_flush in addition to
11055 pending_jump_insns.
11057 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
11059 PR tree-optimization/81611
11060 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
11063 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11066 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
11069 2018-01-30 Richard Biener <rguenther@suse.de>
11070 Jakub Jelinek <jakub@redhat.com>
11072 PR tree-optimization/84111
11073 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
11074 inner loops added during recursion, as they don't have up-to-date
11077 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
11080 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
11081 (can_inline_edge_by_limits_p): ... here.
11082 (can_early_inline_edge_p, check_callers,
11083 update_caller_keys, update_callee_keys, recursive_inlining,
11084 add_new_edges_to_heap, speculation_useful_p,
11085 inline_small_functions,
11086 inline_small_functions, flatten_function,
11087 inline_to_all_callers_1): Update.
11089 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
11091 * profile-count.c (profile_count::combine_with_ipa_count): Handle
11094 2018-01-30 Richard Biener <rguenther@suse.de>
11096 PR tree-optimization/83008
11097 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
11098 invariant and constant vector uses in stmts when they need
11099 more than one stmt.
11101 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11104 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
11105 * configure: Regenerate.
11107 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11109 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
11111 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
11112 Use gen_rtx_REG rather than gen_lowpart.
11114 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11116 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
11117 rather than 0 when creating partial subregs.
11119 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11121 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
11124 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
11127 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
11128 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
11129 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
11130 flags. This restores the settings used before the 2017-07-24.
11131 Turning off pre increment/decrement/modify allows IVOPTS to
11132 optimize DF/SF loops where the index is an int.
11134 2018-01-29 Richard Biener <rguenther@suse.de>
11135 Kelvin Nilsen <kelvin@gcc.gnu.org>
11138 * tree-vect-stmts.c (vectorizable_call): Don't call
11139 targetm.vectorize_builtin_md_vectorized_function if callee is
11142 2018-01-22 Carl Love <cel@us.ibm.com>
11144 * doc/extend.tex: Fix typo in second arg in
11145 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
11147 2018-01-29 Richard Biener <rguenther@suse.de>
11149 PR tree-optimization/84086
11150 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
11151 (flush_ssaname_freelist): When SSA names were released reset
11152 the SCEV hash table.
11154 2018-01-29 Richard Biener <rguenther@suse.de>
11156 PR tree-optimization/84057
11157 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
11158 removed paths when removing edges.
11160 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
11162 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
11163 -mfunction-return=@var{choice}.
11165 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
11167 PR diagnostic/84034
11168 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
11169 Handle CR like TAB.
11170 (layout::print_source_line): Likewise.
11171 (test_get_line_width_without_trailing_whitespace): Add test cases.
11173 2018-01-27 Jakub Jelinek <jakub@redhat.com>
11175 PR middle-end/84040
11176 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
11179 2018-01-26 Jim Wilson <jimw@sifive.com>
11181 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
11183 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
11186 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11188 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
11189 and CMP + SUB-immediate -> SUBS.
11191 2018-01-26 Martin Sebor <msebor@redhat.com>
11193 PR tree-optimization/83896
11194 * tree-ssa-strlen.c (get_string_len): Rename...
11195 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
11196 Avoid assuming length is constant.
11197 (handle_char_store): Use HOST_WIDE_INT for string length.
11199 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
11202 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
11203 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
11205 2018-01-26 Richard Biener <rguenther@suse.de>
11207 PR rtl-optimization/84003
11208 * dse.c (record_store): Only record redundant stores when
11209 the earlier store aliases at least all accesses the later one does.
11211 2018-01-26 Jakub Jelinek <jakub@redhat.com>
11213 PR rtl-optimization/83985
11214 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
11215 REG_CFA_RESTORE insns.
11216 (delete_unmarked_insns): Don't ignore separate shrink wrapping
11217 REG_CFA_RESTORE insns here.
11220 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
11221 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
11223 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11225 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
11226 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
11227 (arc_init): Likewise.
11228 (arc_override_options): Likewise.
11229 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
11231 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
11233 * config/arc/arc.h (TARGET_DBNZ): Define.
11234 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
11235 properly set the tune attribute.
11236 (dbnz): Use TARGET_DBNZ guard.
11237 * config/arc/arc.opt (mtune): Add core3 option.
11239 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11241 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
11242 recognize new pic like addresses.
11243 (arc_delegitimize_address): Clean up.
11245 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11247 * config/arc/arc-arches.def: Option mrf16 valid for all
11249 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
11250 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
11251 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
11252 * config/arc/arc-tables.opt: Regenerate.
11253 * config/arc/arc.c (arc_conditional_register_usage): Handle
11254 reduced register file case.
11255 (arc_file_start): Set must have build attributes.
11256 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
11257 mrf16 option value.
11258 * config/arc/arc.opt (mrf16): Add new option.
11259 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
11260 * config/arc/genmultilib.awk: Handle new mrf16 option.
11261 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
11262 * config/arc/t-multilib: Regenerate.
11263 * doc/invoke.texi (ARC Options): Document mrf16 option.
11265 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11267 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
11268 * config/arc/arc.c (arc_handle_secure_attribute): New function.
11269 (arc_attribute_table): Add 'secure_call' attribute.
11270 (arc_print_operand): Print secure call operand.
11271 (arc_function_ok_for_sibcall): Don't optimize tail calls when
11273 (arc_is_secure_call_p): New function. * config/arc/arc.md
11274 (call_i): Add support for sjli instruction.
11275 (call_value_i): Likewise.
11276 * config/arc/constraints.md (Csc): New constraint.
11278 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11279 John Eric Martin <John.Martin@emmicro-us.com>
11281 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
11282 * config/arc/arc.c (_arc_jli_section): New struct.
11283 (arc_jli_section): New type.
11284 (rc_jli_sections): New static variable.
11285 (arc_handle_jli_attribute): New function.
11286 (arc_attribute_table): Add jli_always and jli_fixed attribute.
11287 (arc_file_end): New function.
11288 (TARGET_ASM_FILE_END): Define.
11289 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
11290 (arc_add_jli_section): New function.
11291 (jli_call_scan): Likewise.
11292 (arc_reorg): Call jli_call_scan.
11293 (arc_output_addsi): Remove 'S' from printing asm operand.
11294 (arc_is_jli_call_p): New function.
11295 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
11297 (movhi_insn): Likewise.
11298 (movsi_insn): Likewise.
11299 (movsi_set_cc_insn): Likewise.
11300 (loadqi_update): Likewise.
11301 (load_zeroextendqisi_update): Likewise.
11302 (load_signextendqisi_update): Likewise.
11303 (loadhi_update): Likewise.
11304 (load_zeroextendhisi_update): Likewise.
11305 (load_signextendhisi_update): Likewise.
11306 (loadsi_update): Likewise.
11307 (loadsf_update): Likewise.
11308 (movsicc_insn): Likewise.
11309 (bset_insn): Likewise.
11310 (bxor_insn): Likewise.
11311 (bclr_insn): Likewise.
11312 (bmsk_insn): Likewise.
11313 (bicsi3_insn): Likewise.
11314 (cmpsi_cc_c_insn): Likewise.
11315 (movsi_ne): Likewise.
11316 (movsi_cond_exec): Likewise.
11317 (clrsbsi2): Likewise.
11318 (norm_f): Likewise.
11328 (call_i): Remove 'S' asm letter, add jli instruction.
11329 (call_value_i): Likewise.
11330 * config/arc/arc.op (mjli-always): New option.
11331 * config/arc/constraints.md (Cji): New constraint.
11332 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
11334 (subsf3_fpx): Likewise.
11335 (mulsf3_fpx): Likewise.
11336 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
11338 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
11339 function attrbutes.
11340 * doc/invoke.texi (ARC): Document mjli-always option.
11342 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
11344 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
11345 avoid addition with 0 and use incw and decw where possible.
11347 2018-01-26 Richard Biener <rguenther@suse.de>
11349 PR tree-optimization/81082
11350 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
11351 association if it requires casting to unsigned.
11352 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
11353 from fold_plusminus_mult_expr to catch important cases late when
11354 range info is available.
11356 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11358 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
11359 * configure.ac (hidden_linkonce): New test.
11360 * configure: Regenerate.
11361 * config.in: Regenerate.
11363 2018-01-26 Julia Koval <julia.koval@intel.com>
11365 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
11366 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
11367 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
11368 _mm_mask_bitshuffle_epi64_mask): Fix type.
11369 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
11370 USI_FTYPE_V4DI_V4DI_USI): Remove.
11371 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
11372 __builtin_ia32_vpshufbitqmb256_mask,
11373 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
11374 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
11375 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
11377 2018-01-26 Alan Modra <amodra@gmail.com>
11380 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
11381 UNSPEC_VBPERMQ. Sort other unspecs.
11383 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
11385 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
11387 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
11389 PR middle-end/83055
11390 * predict.c (drop_profile): Do not push/pop cfun; update also
11392 (handle_missing_profiles): Fix logic looking for zero profiles.
11394 2018-01-25 Jakub Jelinek <jakub@redhat.com>
11396 PR middle-end/83977
11397 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
11398 on functions with #pragma omp declare simd or functions with simd
11400 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
11401 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
11402 Remove trailing \n from warning_at calls.
11404 2018-01-25 Tom de Vries <tom@codesourcery.com>
11407 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
11408 for neutered workers.
11410 2018-01-24 Joseph Myers <joseph@codesourcery.com>
11413 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
11414 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
11416 2018-01-24 Jeff Law <law@redhat.com>
11419 * i386.c (get_probe_interval): Move to earlier point.
11420 (ix86_compute_frame_layout): If -fstack-clash-protection and
11421 the frame is larger than the probe interval, then use pushes
11422 to save registers rather than reg->mem moves.
11423 (ix86_expand_prologue): Remove conditional for int_registers_saved
11426 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
11429 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
11430 min/max for never referenced object.
11432 2018-01-24 Jakub Jelinek <jakub@redhat.com>
11434 PR middle-end/83977
11435 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
11437 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
11438 attributes from DECL_ATTRIBUTES (decl) without affecting
11439 DECL_ATTRIBUTES (current_function_decl).
11440 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
11441 functions with non-NULL DECL_ABSTRACT_ORIGIN.
11443 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
11445 PR tree-optimization/83979
11446 * fold-const.c (fold_comparison): Use constant_boolean_node
11447 instead of boolean_{true,false}_node.
11449 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
11451 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
11454 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11456 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11457 Simplify the clause that sets the length attribute.
11458 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11459 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
11460 clause that sets the length attribute.
11461 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11463 2018-01-24 Tom de Vries <tom@codesourcery.com>
11466 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
11467 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
11468 Add strict parameter.
11469 (prevent_branch_around_nothing): Insert dummy insn between branch to
11470 label and label with no ptx insn inbetween.
11471 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
11473 2018-01-24 Tom de Vries <tom@codesourcery.com>
11476 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
11477 for neutered threads in warp.
11478 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
11480 2018-01-24 Richard Biener <rguenther@suse.de>
11482 PR tree-optimization/83176
11483 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
11486 2018-01-24 Richard Biener <rguenther@suse.de>
11488 PR tree-optimization/82819
11489 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
11490 code generating pluses that are no-ops in the target precision.
11492 2018-01-24 Richard Biener <rguenther@suse.de>
11494 PR middle-end/84000
11495 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
11497 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11499 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
11500 to merge probabilities.
11501 * predict.c (probably_never_executed): Also mark as cold functions
11502 with global 0 profile and guessed local profile.
11503 * profile-count.c (profile_probability::combine_with_count): New
11505 * profile-count.h (profile_probability::operator*,
11506 profile_probability::operator*=, profile_probability::operator/,
11507 profile_probability::operator/=): Reduce precision to adjusted
11508 and set value to guessed on contradictory divisions.
11509 (profile_probability::combine_with_freq): Remove.
11510 (profile_probability::combine_wiht_count): Declare.
11511 (profile_count::force_nonzero):: Set to adjusted.
11512 (profile_count::probability_in):: Set quality to adjusted.
11513 * tree-ssa-tail-merge.c (replace_block_by): Use
11514 combine_with_count.
11516 2018-01-23 Andrew Waterman <andrew@sifive.com>
11517 Jim Wilson <jimw@sifive.com>
11519 * config/riscv/riscv.c (riscv_stack_boundary): New.
11520 (riscv_option_override): Set riscv_stack_boundary. Handle
11521 riscv_preferred_stack_boundary_arg.
11522 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
11523 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
11524 (STACK_BOUNDARY): Set to riscv_stack_boundary.
11525 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
11526 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
11527 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
11529 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
11532 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
11533 of struct ix86_frame.
11534 (ix86_expand_epilogue): Likewise. Add a local variable for
11535 the reg_save_offset field in struct ix86_frame.
11537 2018-01-23 Bin Cheng <bin.cheng@arm.com>
11539 PR tree-optimization/82604
11540 * tree-loop-distribution.c (enum partition_kind): New enum item
11541 PKIND_PARTIAL_MEMSET.
11542 (partition_builtin_p): Support above new enum item.
11543 (generate_code_for_partition): Ditto.
11544 (compute_access_range): Differentiate cases that equality can be
11545 proven at all loops, the innermost loops or no loops.
11546 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
11547 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
11548 (finalize_partitions, distribute_loop): Don't fuse partition of
11549 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
11550 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
11551 parloop is enabled.
11553 2018-01-23 Martin Liska <mliska@suse.cz>
11555 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
11556 order to ignore the predictor.
11557 (PRED_POLYMORPHIC_CALL): Likewise.
11558 (PRED_RECURSIVE_CALL): Likewise.
11560 2018-01-23 Martin Liska <mliska@suse.cz>
11562 * tree-profile.c (tree_profiling): Print function header to
11563 aware reader which function we are working on.
11564 * value-prof.c (gimple_find_values_to_profile): Do not print
11565 not interesting value histograms.
11567 2018-01-23 Martin Liska <mliska@suse.cz>
11569 * profile-count.h (enum profile_quality): Add
11570 profile_uninitialized as the first value. Do not number values
11571 as they are zero based.
11572 (profile_count::verify): Update sanity check.
11573 (profile_probability::verify): Likewise.
11575 2018-01-23 Nathan Sidwell <nathan@acm.org>
11577 * doc/invoke.texi (ffor-scope): Deprecate.
11579 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11581 PR tree-optimization/83510
11582 * domwalk.c (set_all_edges_as_executable): New function.
11583 (dom_walker::dom_walker): Convert bool param
11584 "skip_unreachable_blocks" to enum reachability. Move setup of
11585 edge flags to set_all_edges_as_executable and only do it when
11586 reachability is REACHABLE_BLOCKS.
11587 * domwalk.h (enum dom_walker::reachability): New enum.
11588 (dom_walker::dom_walker): Convert bool param
11589 "skip_unreachable_blocks" to enum reachability.
11590 (set_all_edges_as_executable): New decl.
11591 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
11592 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
11594 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
11595 but converting true to REACHABLE_BLOCKS.
11596 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
11598 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
11599 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
11600 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
11602 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
11603 if check_all_array_refs will be called.
11605 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11607 * tree.c (selftest::test_location_wrappers): Add more test
11610 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11612 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
11613 (selftest::test_bit_in_range): Likewise.
11615 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
11618 * doc/sourcebuild.texi (vect_float): Say that the selector
11619 only describes the situation when -funsafe-math-optimizations is on.
11620 (vect_float_strict): Document.
11622 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
11624 PR tree-optimization/83965
11625 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
11626 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
11627 instead of checking only for a reduction.
11628 (vect_recog_widen_sum_pattern): Likewise.
11630 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11632 * predict.c (probably_never_executed): Only use precise profile info.
11633 (compute_function_frequency): Skip after inlining hack since we now
11634 have quality checking.
11636 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11638 * profile-count.h (profile_probability::very_unlikely,
11639 profile_probability::unlikely, profile_probability::even): Set
11640 precision to guessed.
11642 2018-01-23 Richard Biener <rguenther@suse.de>
11644 PR tree-optimization/83963
11645 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
11646 Properly terminate dominator walk when crossing the exit edge not
11647 when visiting its source block.
11649 2018-01-23 Jakub Jelinek <jakub@redhat.com>
11652 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
11653 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
11655 2018-01-22 Jakub Jelinek <jakub@redhat.com>
11657 PR tree-optimization/83957
11658 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
11659 semicolon after for body surrounded by braces.
11661 PR tree-optimization/83081
11662 * profile-count.h (profile_probability::split): New method.
11663 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
11664 Use profile_probability::split.
11665 (do_compare_rtx_and_jump): Fix adjustment of probabilities
11666 when splitting a single conditional jump into 2.
11668 2018-01-22 David Malcolm <dmalcolm@redhat.com>
11670 PR tree-optimization/69452
11671 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
11674 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11676 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
11677 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
11678 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
11680 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11682 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
11684 * config/rl78/rl78.md (movdi): New define_expand.
11685 * config/rl78/rl78.c (rl78_split_movdi): New function.
11687 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
11690 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
11692 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
11693 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
11694 128-bit to produce an UNSPEC move to get the double word with the
11695 signbit and then a shift directly to do signbit.
11696 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
11697 implementation with a new version that just does either a direct
11698 move or a regular move. Move memory interface to separate insns.
11699 Move insns so they are next to the expander.
11700 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
11701 with signbit move. Split big and little endian case.
11702 (signbit<mode>2_dm_mem_le): Likewise.
11703 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
11704 (signbit<mode>2_dm2): Likewise.
11706 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11708 * config/rl78/rl78.md (anddi3): New define_expand.
11710 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11712 * config/rl78/rl78.md (umindi3): New define_expand.
11714 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11716 * config/rl78/rl78.md (smindi3): New define_expand.
11718 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11720 * config/rl78/rl78.md (smaxdi3): New define_expand.
11722 2018-01-22 Carl Love <cel@us.ibm.com>
11724 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
11725 LVX_V1TI): Add macro expansion.
11726 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
11727 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
11728 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
11729 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
11730 Change check to determine if the instruction is a byte reversing
11731 entry. Fix typo in comment.
11732 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
11733 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
11734 Add def_builtin calls for new builtins.
11735 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
11736 Add define_insn expansion.
11738 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11740 * config/rl78/rl78.md (umaxdi3): New define_expand.
11742 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11744 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
11745 for non-QImode registers.
11747 2018-01-22 Richard Biener <rguenther@suse.de>
11749 PR tree-optimization/83963
11750 * graphite-scop-detection.c (scop_detection::get_sese): Delay
11751 including the loop exit block.
11752 (scop_detection::merge_sese): Likewise.
11753 (scop_detection::add_scop): Do it here instead.
11755 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11757 * doc/sourcebuild.texi (arm_softfloat): Document.
11759 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
11762 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
11763 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
11764 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
11766 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11767 David Edelsohn <dje.gcc@gmail.com>
11770 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11771 Change "crset eq" to "crset 2".
11772 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11773 (*call_indirect_aix<mode>_nospec): Likewise.
11774 (*call_value_indirect_aix<mode>_nospec): Likewise.
11775 (*call_indirect_elfv2<mode>_nospec): Likewise.
11776 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
11777 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
11778 change assembly output from . to $.
11779 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11780 (indirect_jump<mode>_nospec): Change assembly output from . to $.
11781 (*tablejump<mode>_internal1_nospec): Likewise.
11783 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
11786 * config/sh/sh_optimize_sett_clrt.cc:
11787 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
11789 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
11791 PR tree-optimization/83940
11792 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
11793 offset_dt to vect_constant_def rather than vect_unknown_def_type.
11794 (vect_check_load_store_mask): Add a mask_dt_out parameter and
11795 use it to pass back the definition type.
11796 (vect_check_store_rhs): Likewise rhs_dt_out.
11797 (vect_build_gather_load_calls): Add a mask_dt argument and use
11798 it instead of a call to vect_is_simple_use.
11799 (vectorizable_store): Update calls to vect_check_load_store_mask
11800 and vect_check_store_rhs. Use the dt returned by the latter instead
11801 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
11802 instead of calls to vect_is_simple_use. Pass the scalar rather
11803 than the vector operand to vect_is_simple_use when handling
11804 second and subsequent copies of an rhs value.
11805 (vectorizable_load): Update calls to vect_check_load_store_mask
11806 and vect_build_gather_load_calls. Use the cached mask_dt and
11807 gs_info.offset_dt instead of calls to vect_is_simple_use.
11809 2018-01-20 Jakub Jelinek <jakub@redhat.com>
11811 PR middle-end/83945
11812 * tree-emutls.c: Include gimplify.h.
11813 (lower_emutls_2): New function.
11814 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
11815 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
11816 it before further processing.
11819 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
11820 UINTVAL (trueop1) instead of INTVAL (op1).
11822 2018-01-19 Jakub Jelinek <jakub@redhat.com>
11826 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
11827 INCOMING_FRAME_SP_OFFSET if not defined.
11828 (scan_trace): Add ENTRY argument. If true and
11829 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
11830 emit a note to adjust the CFA offset.
11831 (create_cfi_notes): Adjust scan_trace callers.
11832 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
11833 INCOMING_FRAME_SP_OFFSET in the CIE.
11834 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
11835 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
11837 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
11838 * doc/tm.texi: Regenerated.
11840 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11842 PR rtl-optimization/83147
11843 * lra-constraints.c (remove_inheritance_pseudos): Use
11844 lra_substitute_pseudo_within_insn.
11846 2018-01-19 Tom de Vries <tom@codesourcery.com>
11847 Cesar Philippidis <cesar@codesourcery.com>
11850 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
11852 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
11855 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
11856 spaces for function labels.
11858 2018-01-19 Martin Liska <mliska@suse.cz>
11860 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
11861 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
11862 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
11863 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
11864 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
11865 (PRED_CONST_RETURN): Change from 69 to 65.
11866 (PRED_NULL_RETURN): Change from 91 to 71.
11867 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
11868 (PRED_LOOP_GUARD): Change from 66 to 73.
11870 2018-01-19 Martin Liska <mliska@suse.cz>
11872 * predict.c (predict_insn_def): Add new assert.
11873 (struct branch_predictor): Change type to signed integer.
11874 (test_prediction_value_range): Amend test to cover
11875 PROB_UNINITIALIZED.
11876 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
11877 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
11878 (PRED_LOOP_ITERATIONS_MAX): Likewise.
11879 (PRED_LOOP_IV_COMPARE): Likewise.
11880 * predict.h (PROB_UNINITIALIZED): Define new constant.
11882 2018-01-19 Martin Liska <mliska@suse.cz>
11884 * predict.c (dump_prediction): Add new format for
11885 analyze_brprob.py script which is enabled with -details
11887 * profile-count.h (precise_p): New function.
11889 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
11891 PR tree-optimization/83922
11892 * tree-vect-loop.c (vect_verify_full_masking): Return false if
11893 there are no statements that need masking.
11894 (vect_active_double_reduction_p): New function.
11895 (vect_analyze_loop_operations): Use it when handling phis that
11896 are not in the loop header.
11898 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
11900 PR tree-optimization/83914
11901 * tree-vect-loop.c (vectorizable_induction): Don't convert
11902 init_expr or apply the peeling adjustment for inductions
11903 that are nested within the vectorized loop.
11905 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11907 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
11910 2018-01-18 Jakub Jelinek <jakub@redhat.com>
11914 * function.h (gimplify_parameters): Add gimple_seq * argument.
11915 * function.c: Include gimple.h and options.h.
11916 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
11917 for the added local temporaries if needed.
11918 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
11919 if there are any parameter cleanups, wrap whole body into a
11920 try/finally with the cleanups.
11922 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
11925 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
11926 Use GET_MODE_CLASS for scalar floating point.
11928 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
11932 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
11933 Fix call of call_cgraph_insertion_hooks.
11935 2018-01-18 Martin Sebor <msebor@redhat.com>
11937 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
11939 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
11942 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
11945 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
11948 * common.opt: (-ffile-prefix-map): New option.
11949 * opts.c (common_handle_option): Defer it.
11950 * opts-global.c (handle_common_deferred_options): Handle it.
11951 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
11952 * file-prefix-map.h: New file.
11953 (remap_debug_filename, add_debug_prefix_map): ...here.
11954 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
11955 * final.c (debug_prefix_map, add_debug_prefix_map
11956 remap_debug_filename): Move to...
11957 * file-prefix-map.c: New file.
11958 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
11959 generalize, get rid of alloca(), use strrchr() instead of strchr().
11960 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
11961 Implement in terms of add_prefix_map().
11962 (remap_macro_filename, remap_debug_filename): Implement in term of
11964 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
11965 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
11966 * dbxout.c: Include file-prefix-map.h.
11967 * varasm.c: Likewise.
11968 * vmsdbgout.c: Likewise.
11969 * xcoffout.c: Likewise.
11970 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
11971 * doc/cppopts.texi (-fmacro-prefix-map): Document.
11972 * doc/invoke.texi (-ffile-prefix-map): Document.
11973 (-fdebug-prefix-map): Update description.
11975 2018-01-18 Martin Liska <mliska@suse.cz>
11977 * config/i386/i386.c (indirect_thunk_name): Document that also
11979 (output_indirect_thunk): Document why both instructions
11980 (pause and lfence) are generated.
11982 2018-01-18 Richard Biener <rguenther@suse.de>
11984 PR tree-optimization/83887
11985 * graphite-scop-detection.c
11986 (scop_detection::get_nearest_dom_with_single_entry): Remove.
11987 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
11988 (scop_detection::merge_sese): Re-implement with a flood-fill
11989 algorithm that properly finds a SESE region if it exists.
11991 2018-01-18 Jakub Jelinek <jakub@redhat.com>
11994 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
11995 pointer_diff optimizations use view_convert instead of convert.
11997 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11999 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12000 Generate different code for -mno-speculate-indirect-jumps.
12001 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12002 (*call_indirect_aix<mode>): Disable for
12003 -mno-speculate-indirect-jumps.
12004 (*call_indirect_aix<mode>_nospec): New define_insn.
12005 (*call_value_indirect_aix<mode>): Disable for
12006 -mno-speculate-indirect-jumps.
12007 (*call_value_indirect_aix<mode>_nospec): New define_insn.
12008 (*sibcall_nonlocal_sysv<mode>): Generate different code for
12009 -mno-speculate-indirect-jumps.
12010 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12012 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
12014 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
12015 long double type, set the flags for noting the default long double
12016 type, even if we don't pass or return a long double type.
12018 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
12021 * ipa-inline.c (flatten_function): Do not overwrite final inlining
12024 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
12026 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
12027 support for merge[hl].
12028 (fold_mergehl_helper): New helper function.
12029 (tree-vector-builder.h): New #include for tree_vector_builder usage.
12030 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
12031 (altivec_vmrglw_direct): Add xxmrglw insn.
12033 2018-01-17 Andrew Waterman <andrew@sifive.com>
12035 * config/riscv/riscv.c (riscv_conditional_register_usage): If
12036 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
12038 2018-01-17 David Malcolm <dmalcolm@redhat.com>
12041 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
12042 call the lto_location_cache before reading the
12043 DECL_SOURCE_LOCATION of the types.
12045 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
12046 Richard Sandiford <richard.sandiford@linaro.org>
12048 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
12049 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
12050 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
12051 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
12052 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
12054 * config/aarch64/constraints.md (aarch64_movti_operand):
12056 * config/aarch64/predicates.md (Uti): Add new constraint.
12058 2018-01-17 Carl Love <cel@us.ibm.com>
12060 * config/rs6000/vsx.md (define_expand xl_len_r,
12061 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
12062 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
12064 (define_expand, define_insn): Move the shift left from the
12065 define_insn to the define_expand for lxvl and stxvl instructions.
12066 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
12067 and XL_LEN_R definitions to PURE.
12069 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
12071 * config/i386/i386.c (indirect_thunk_name): Declare regno
12072 as unsigned int. Compare regno with INVALID_REGNUM.
12073 (output_indirect_thunk): Ditto.
12074 (output_indirect_thunk_function): Ditto.
12075 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
12076 in the call to output_indirect_thunk_function.
12078 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
12080 PR middle-end/83884
12081 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
12082 rather than the size of inner_type to determine the stack slot size
12083 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
12085 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
12088 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
12091 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
12093 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
12094 endian Linux systems to optionally enable multilibs for selecting
12095 the long double type if the user configured an explicit type.
12096 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
12097 have no long double multilibs if not defined.
12098 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
12099 warn if the user used -mabi={ieee,ibm}longdouble and we built
12100 multilibs for long double.
12101 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
12102 appropriate multilib option.
12103 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
12105 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
12106 for building long double multilibs.
12107 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
12109 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
12111 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
12114 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
12116 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
12119 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
12122 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
12125 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
12127 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
12128 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
12130 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
12132 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
12133 different rtl trees depending on TARGET_64BIT.
12134 (rs6000_gen_lvx): Likewise.
12136 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
12138 * config/visium/visium.md (nop): Tweak comment.
12139 (hazard_nop): Likewise.
12141 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12143 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
12144 -mspeculate-indirect-jumps.
12145 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
12146 for -mno-speculate-indirect-jumps.
12147 (*call_indirect_elfv2<mode>_nospec): New define_insn.
12148 (*call_value_indirect_elfv2<mode>): Disable for
12149 -mno-speculate-indirect-jumps.
12150 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
12151 (indirect_jump): Emit different RTL for
12152 -mno-speculate-indirect-jumps.
12153 (*indirect_jump<mode>): Disable for
12154 -mno-speculate-indirect-jumps.
12155 (*indirect_jump<mode>_nospec): New define_insn.
12156 (tablejump): Emit different RTL for
12157 -mno-speculate-indirect-jumps.
12158 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
12159 (tablejumpsi_nospec): New define_expand.
12160 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
12161 (tablejumpdi_nospec): New define_expand.
12162 (*tablejump<mode>_internal1): Disable for
12163 -mno-speculate-indirect-jumps.
12164 (*tablejump<mode>_internal1_nospec): New define_insn.
12165 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
12168 2018-01-16 Artyom Skrobov tyomitch@gmail.com
12170 * caller-save.c (insert_save): Drop unnecessary parameter. All
12173 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12174 Richard Biener <rguenth@suse.de>
12177 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
12178 return early, inline manually is_gimple_sizepos. Make sure if we
12179 call gimplify_expr we don't end up with a gimple constant.
12180 * tree.c (variably_modified_type_p): Don't return true for
12181 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
12182 * gimplify.h (is_gimple_sizepos): Remove.
12184 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12186 PR tree-optimization/83857
12187 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
12188 vectorizable_live_operation for pure SLP statements.
12189 (vectorizable_live_operation): Handle PHIs.
12191 2018-01-16 Richard Biener <rguenther@suse.de>
12193 PR tree-optimization/83867
12194 * tree-vect-stmts.c (vect_transform_stmt): Precompute
12195 nested_in_vect_loop_p since the scalar stmt may get invalidated.
12197 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12200 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
12201 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
12202 If off is not INTEGER_CST, issue a may not be aligned warning
12203 rather than isn't aligned. Use isn%'t rather than isn't.
12204 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
12206 <case MULT_EXPR>: Improve the case when bottom and one of the
12207 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
12208 operand, in that case check if the other operand is multiple of
12209 bottom divided by the INTEGER_CST operand.
12211 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12214 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
12215 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
12216 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
12217 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
12218 * config/pa/pa.c (pa_function_arg_advance): Likewise.
12219 (pa_function_arg, pa_arg_partial_bytes): Likewise.
12220 (pa_function_arg_size): New function.
12222 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12224 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
12225 in a separate statement.
12227 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12229 PR tree-optimization/83847
12230 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
12231 group gathers and scatters.
12233 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12235 PR rtl-optimization/86620
12236 * params.def (max-sched-ready-insns): Bump minimum value to 1.
12238 PR rtl-optimization/83213
12239 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
12240 to last if both are JUMP_INSNs.
12242 PR tree-optimization/83843
12243 * gimple-ssa-store-merging.c
12244 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
12245 store_immediate_info for bswap/nop orig_stores.
12247 2018-01-15 Andrew Waterman <andrew@sifive.com>
12249 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
12251 <UDIV>: Increase cost if !TARGET_DIV.
12253 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
12255 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
12256 (define_attr "cr_logical_3op"): New.
12257 (cceq_ior_compare): Adjust.
12258 (cceq_ior_compare_complement): Adjust.
12259 (*cceq_rev_compare): Adjust.
12260 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
12261 (is_cracked_insn): Adjust.
12262 (insn_must_be_first_in_group): Adjust.
12263 * config/rs6000/40x.md: Adjust.
12264 * config/rs6000/440.md: Adjust.
12265 * config/rs6000/476.md: Adjust.
12266 * config/rs6000/601.md: Adjust.
12267 * config/rs6000/603.md: Adjust.
12268 * config/rs6000/6xx.md: Adjust.
12269 * config/rs6000/7450.md: Adjust.
12270 * config/rs6000/7xx.md: Adjust.
12271 * config/rs6000/8540.md: Adjust.
12272 * config/rs6000/cell.md: Adjust.
12273 * config/rs6000/e300c2c3.md: Adjust.
12274 * config/rs6000/e500mc.md: Adjust.
12275 * config/rs6000/e500mc64.md: Adjust.
12276 * config/rs6000/e5500.md: Adjust.
12277 * config/rs6000/e6500.md: Adjust.
12278 * config/rs6000/mpc.md: Adjust.
12279 * config/rs6000/power4.md: Adjust.
12280 * config/rs6000/power5.md: Adjust.
12281 * config/rs6000/power6.md: Adjust.
12282 * config/rs6000/power7.md: Adjust.
12283 * config/rs6000/power8.md: Adjust.
12284 * config/rs6000/power9.md: Adjust.
12285 * config/rs6000/rs64.md: Adjust.
12286 * config/rs6000/titan.md: Adjust.
12288 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12290 * config/i386/predicates.md (indirect_branch_operand): Rewrite
12291 ix86_indirect_branch_register logic.
12293 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12295 * config/i386/constraints.md (Bs): Update
12296 ix86_indirect_branch_register check. Don't check
12297 ix86_indirect_branch_register with GOT_memory_operand.
12299 * config/i386/predicates.md (GOT_memory_operand): Don't check
12300 ix86_indirect_branch_register here.
12301 (GOT32_symbol_operand): Likewise.
12303 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12305 * config/i386/predicates.md (constant_call_address_operand):
12306 Rewrite ix86_indirect_branch_register logic.
12307 (sibcall_insn_operand): Likewise.
12309 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12311 * config/i386/constraints.md (Bs): Replace
12312 ix86_indirect_branch_thunk_register with
12313 ix86_indirect_branch_register.
12315 * config/i386/i386.md (indirect_jump): Likewise.
12316 (tablejump): Likewise.
12317 (*sibcall_memory): Likewise.
12318 (*sibcall_value_memory): Likewise.
12319 Peepholes of indirect call and jump via memory: Likewise.
12320 * config/i386/i386.opt: Likewise.
12321 * config/i386/predicates.md (indirect_branch_operand): Likewise.
12322 (GOT_memory_operand): Likewise.
12323 (call_insn_operand): Likewise.
12324 (sibcall_insn_operand): Likewise.
12325 (GOT32_symbol_operand): Likewise.
12327 2018-01-15 Jakub Jelinek <jakub@redhat.com>
12329 PR middle-end/83837
12330 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
12331 type rather than type addr's type points to.
12332 (expand_omp_atomic_mutex): Likewise.
12333 (expand_omp_atomic): Likewise.
12335 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12338 * config/i386/i386.c (output_indirect_thunk_function): Use
12339 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
12340 for __x86_return_thunk.
12342 2018-01-15 Richard Biener <rguenther@suse.de>
12344 PR middle-end/83850
12345 * expmed.c (extract_bit_field_1): Fix typo.
12347 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12350 * config/arm/iterators.md (VF): New mode iterator.
12351 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
12352 Remove integer-related logic from pattern.
12353 (neon_vabd<mode>_3): Likewise.
12355 2018-01-15 Jakub Jelinek <jakub@redhat.com>
12357 PR middle-end/82694
12358 * common.opt (fstrict-overflow): No longer an alias.
12359 (fwrapv-pointer): New option.
12360 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
12361 also for pointer types based on flag_wrapv_pointer.
12362 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
12363 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
12364 opts->x_flag_wrapv got set.
12365 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
12366 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
12367 POINTER_TYPE_OVERFLOW_UNDEFINED.
12368 * match.pd: Likewise in address comparison pattern.
12369 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
12371 2018-01-15 Richard Biener <rguenther@suse.de>
12374 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
12375 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
12376 Reset type names to their identifier if their TYPE_DECL doesn't
12377 have linkage (and thus is used for ODR and devirt).
12378 (save_debug_info_for_decl): Remove.
12379 (save_debug_info_for_type): Likewise.
12380 (add_tree_to_fld_list): Adjust.
12381 * tree-pretty-print.c (dump_generic_node): Make dumping of
12382 type names more robust.
12384 2018-01-15 Richard Biener <rguenther@suse.de>
12386 * BASE-VER: Bump to 8.0.1.
12388 2018-01-14 Martin Sebor <msebor@redhat.com>
12391 * builtins.c (check_access): Avoid warning when the no-warning bit
12394 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
12396 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
12397 * ira-color (allocno_hard_regs_compare): Likewise.
12399 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
12402 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
12403 Use .pushsection/.popsection.
12405 2018-01-14 Martin Sebor <msebor@redhat.com>
12408 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
12410 2018-01-14 Jakub Jelinek <jakub@redhat.com>
12412 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
12413 entry from extra_headers.
12414 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
12415 extra_headers, make the list bitwise identical to the i?86-*-* one.
12417 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12419 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
12420 -mcmodel=large with -mindirect-branch=thunk,
12421 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
12422 -mfunction-return=thunk-extern.
12423 * doc/invoke.texi: Document -mcmodel=large is incompatible with
12424 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
12425 -mfunction-return=thunk and -mfunction-return=thunk-extern.
12427 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12429 * config/i386/i386.c (print_reg): Print the name of the full
12430 integer register without '%'.
12431 (ix86_print_operand): Handle 'V'.
12432 * doc/extend.texi: Document 'V' modifier.
12434 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12436 * config/i386/constraints.md (Bs): Disallow memory operand for
12437 -mindirect-branch-register.
12439 * config/i386/predicates.md (indirect_branch_operand): Likewise.
12440 (GOT_memory_operand): Likewise.
12441 (call_insn_operand): Likewise.
12442 (sibcall_insn_operand): Likewise.
12443 (GOT32_symbol_operand): Likewise.
12444 * config/i386/i386.md (indirect_jump): Call convert_memory_address
12445 for -mindirect-branch-register.
12446 (tablejump): Likewise.
12447 (*sibcall_memory): Likewise.
12448 (*sibcall_value_memory): Likewise.
12449 Disallow peepholes of indirect call and jump via memory for
12450 -mindirect-branch-register.
12451 (*call_pop): Replace m with Bw.
12452 (*call_value_pop): Likewise.
12453 (*sibcall_pop_memory): Replace m with Bs.
12454 * config/i386/i386.opt (mindirect-branch-register): New option.
12455 * doc/invoke.texi: Document -mindirect-branch-register option.
12457 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12459 * config/i386/i386-protos.h (ix86_output_function_return): New.
12460 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
12461 set function_return_type.
12462 (indirect_thunk_name): Add ret_p to indicate thunk for function
12464 (output_indirect_thunk_function): Pass false to
12465 indirect_thunk_name.
12466 (ix86_output_indirect_branch_via_reg): Likewise.
12467 (ix86_output_indirect_branch_via_push): Likewise.
12468 (output_indirect_thunk_function): Create alias for function
12469 return thunk if regno < 0.
12470 (ix86_output_function_return): New function.
12471 (ix86_handle_fndecl_attribute): Handle function_return.
12472 (ix86_attribute_table): Add function_return.
12473 * config/i386/i386.h (machine_function): Add
12474 function_return_type.
12475 * config/i386/i386.md (simple_return_internal): Use
12476 ix86_output_function_return.
12477 (simple_return_internal_long): Likewise.
12478 * config/i386/i386.opt (mfunction-return=): New option.
12479 (indirect_branch): Mention -mfunction-return=.
12480 * doc/extend.texi: Document function_return function attribute.
12481 * doc/invoke.texi: Document -mfunction-return= option.
12483 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12485 * config/i386/i386-opts.h (indirect_branch): New.
12486 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
12487 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
12488 with local indirect jump when converting indirect call and jump.
12489 (ix86_set_indirect_branch_type): New.
12490 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
12491 (indirectlabelno): New.
12492 (indirect_thunk_needed): Likewise.
12493 (indirect_thunk_bnd_needed): Likewise.
12494 (indirect_thunks_used): Likewise.
12495 (indirect_thunks_bnd_used): Likewise.
12496 (INDIRECT_LABEL): Likewise.
12497 (indirect_thunk_name): Likewise.
12498 (output_indirect_thunk): Likewise.
12499 (output_indirect_thunk_function): Likewise.
12500 (ix86_output_indirect_branch_via_reg): Likewise.
12501 (ix86_output_indirect_branch_via_push): Likewise.
12502 (ix86_output_indirect_branch): Likewise.
12503 (ix86_output_indirect_jmp): Likewise.
12504 (ix86_code_end): Call output_indirect_thunk_function if needed.
12505 (ix86_output_call_insn): Call ix86_output_indirect_branch if
12507 (ix86_handle_fndecl_attribute): Handle indirect_branch.
12508 (ix86_attribute_table): Add indirect_branch.
12509 * config/i386/i386.h (machine_function): Add indirect_branch_type
12510 and has_local_indirect_jump.
12511 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
12513 (tablejump): Likewise.
12514 (*indirect_jump): Use ix86_output_indirect_jmp.
12515 (*tablejump_1): Likewise.
12516 (simple_return_indirect_internal): Likewise.
12517 * config/i386/i386.opt (mindirect-branch=): New option.
12518 (indirect_branch): New.
12521 (thunk-inline): Likewise.
12522 (thunk-extern): Likewise.
12523 * doc/extend.texi: Document indirect_branch function attribute.
12524 * doc/invoke.texi: Document -mindirect-branch= option.
12526 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
12529 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
12531 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
12533 * ipa-inline.c (want_inline_small_function_p): Return false if
12534 inlining has already failed with CIF_FINAL_ERROR.
12535 (update_caller_keys): Call want_inline_small_function_p before
12537 (update_callee_keys): Likewise.
12539 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
12541 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
12543 (rs6000_quadword_masked_address_p): Likewise.
12544 (quad_aligned_load_p): Likewise.
12545 (quad_aligned_store_p): Likewise.
12546 (const_load_sequence_p): Add comment to describe the outer-most loop.
12547 (mimic_memory_attributes_and_flags): New function.
12548 (rs6000_gen_stvx): Likewise.
12549 (replace_swapped_aligned_store): Likewise.
12550 (rs6000_gen_lvx): Likewise.
12551 (replace_swapped_aligned_load): Likewise.
12552 (replace_swapped_load_constant): Capitalize argument name in
12553 comment describing this function.
12554 (rs6000_analyze_swaps): Add a third pass to search for vector loads
12555 and stores that access quad-word aligned addresses and replace
12556 with stvx or lvx instructions when appropriate.
12557 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
12558 New function prototype.
12559 (rs6000_quadword_masked_address_p): Likewise.
12560 (rs6000_gen_lvx): Likewise.
12561 (rs6000_gen_stvx): Likewise.
12562 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
12563 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
12564 when memory address is aligned.
12565 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
12566 this split to select lvx instruction when memory address is aligned.
12567 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
12568 instruction when memory address is aligned.
12569 (*vsx_le_perm_load_v16qi): Likewise.
12570 (four unnamed splitters): Modify to select the stvx instruction
12571 when memory is aligned.
12573 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
12575 * predict.c (determine_unlikely_bbs): Handle correctly BBs
12576 which appears in the queue multiple times.
12578 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12579 Alan Hayward <alan.hayward@arm.com>
12580 David Sherwood <david.sherwood@arm.com>
12582 * tree-vectorizer.h (vec_lower_bound): New structure.
12583 (_loop_vec_info): Add check_nonzero and lower_bounds.
12584 (LOOP_VINFO_CHECK_NONZERO): New macro.
12585 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
12586 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
12587 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
12588 fields. Make seg_len the distance travelled, not including the
12590 (dr_direction_indicator): Declare.
12591 (dr_zero_step_indicator): Likewise.
12592 (dr_known_forward_stride_p): Likewise.
12593 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
12595 (runtime_alias_check_p): Allow runtime alias checks with
12597 (operator ==): Compare access_size and align.
12598 (prune_runtime_alias_test_list): Rework for new distinction between
12599 the access_size and seg_len.
12600 (create_intersect_range_checks_index): Likewise. Cope with polynomial
12602 (get_segment_min_max): New function.
12603 (create_intersect_range_checks): Use it.
12604 (dr_step_indicator): New function.
12605 (dr_direction_indicator): Likewise.
12606 (dr_zero_step_indicator): Likewise.
12607 (dr_known_forward_stride_p): Likewise.
12608 * tree-loop-distribution.c (data_ref_segment_size): Return
12609 DR_STEP * (niters - 1).
12610 (compute_alias_check_pairs): Update call to the dr_with_seg_len
12612 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
12613 (vect_preserves_scalar_order_p): New function, split out from...
12614 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
12615 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
12616 (vect_vfa_access_size): New function.
12617 (vect_vfa_align): Likewise.
12618 (vect_compile_time_alias): Take access_size_a and access_b arguments.
12619 (dump_lower_bound): New function.
12620 (vect_check_lower_bound): Likewise.
12621 (vect_small_gap_p): Likewise.
12622 (vectorizable_with_step_bound_p): Likewise.
12623 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
12624 depencies if the vectorization factor is 1. Convert the checks
12625 for nonzero steps into checks on the bounds of DR_STEP. Try using
12626 a bunds check for variable steps if the minimum required step is
12627 relatively small. Update calls to the dr_with_seg_len
12628 constructor and to vect_compile_time_alias.
12629 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
12631 (vect_loop_versioning): Call it.
12632 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
12634 (vect_estimate_min_profitable_iters): Account for any bounds checks.
12636 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12637 Alan Hayward <alan.hayward@arm.com>
12638 David Sherwood <david.sherwood@arm.com>
12640 * doc/sourcebuild.texi (vect_scatter_store): Document.
12641 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
12643 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
12645 * genopinit.c (main): Add supports_vec_scatter_store and
12646 supports_vec_scatter_store_cached to target_optabs.
12647 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
12648 IFN_MASK_SCATTER_STORE.
12649 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
12651 * internal-fn.h (internal_store_fn_p): Declare.
12652 (internal_fn_stored_value_index): Likewise.
12653 * internal-fn.c (scatter_store_direct): New macro.
12654 (expand_scatter_store_optab_fn): New function.
12655 (direct_scatter_store_optab_supported_p): New macro.
12656 (internal_store_fn_p): New function.
12657 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
12658 IFN_MASK_SCATTER_STORE.
12659 (internal_fn_mask_index): Likewise.
12660 (internal_fn_stored_value_index): New function.
12661 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
12662 for scatter stores.
12663 * optabs-query.h (supports_vec_scatter_store_p): Declare.
12664 * optabs-query.c (supports_vec_scatter_store_p): New function.
12665 * tree-vectorizer.h (vect_get_store_rhs): Declare.
12666 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
12667 true for scatter stores.
12668 (vect_gather_scatter_fn_p): Handle scatter stores too.
12669 (vect_check_gather_scatter): Consider using scatter stores if
12670 supports_vec_scatter_store_p.
12671 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
12672 scatter stores too.
12673 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
12674 internal_fn_stored_value_index.
12675 (check_load_store_masking): Handle scatter stores too.
12676 (vect_get_store_rhs): Make public.
12677 (vectorizable_call): Use internal_store_fn_p.
12678 (vectorizable_store): Handle scatter store internal functions.
12679 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
12680 when deciding whether the end of the group has been reached.
12681 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
12682 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
12683 (mask_scatter_store<mode>): New insns.
12685 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12686 Alan Hayward <alan.hayward@arm.com>
12687 David Sherwood <david.sherwood@arm.com>
12689 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
12690 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
12691 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
12693 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
12694 Use vect_truncate_gather_scatter_offset if we can't treat the
12695 operation as a normal gather load or scatter store.
12696 (get_group_load_store_type): Take the gather_scatter_info
12697 as argument. Try using a gather load or scatter store for
12698 single-element groups.
12699 (get_load_store_type): Update calls to get_group_load_store_type
12700 and vect_use_strided_gather_scatters_p.
12702 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12703 Alan Hayward <alan.hayward@arm.com>
12704 David Sherwood <david.sherwood@arm.com>
12706 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
12707 optional tree argument.
12708 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
12710 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
12711 but continue to use the current value as a fallback.
12712 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
12713 to compare the updates.
12714 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
12715 (get_load_store_type): Use it when handling a strided access.
12716 (vect_get_strided_load_store_ops): New function.
12717 (vect_get_data_ptr_increment): Likewise.
12718 (vectorizable_load): Handle strided gather loads. Always pass
12719 a step to vect_create_data_ref_ptr and bump_vector_ptr.
12721 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12722 Alan Hayward <alan.hayward@arm.com>
12723 David Sherwood <david.sherwood@arm.com>
12725 * doc/md.texi (gather_load@var{m}): Document.
12726 (mask_gather_load@var{m}): Likewise.
12727 * genopinit.c (main): Add supports_vec_gather_load and
12728 supports_vec_gather_load_cached to target_optabs.
12729 * optabs-tree.c (init_tree_optimization_optabs): Use
12730 ggc_cleared_alloc to allocate target_optabs.
12731 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
12732 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
12734 * internal-fn.h (internal_load_fn_p): Declare.
12735 (internal_gather_scatter_fn_p): Likewise.
12736 (internal_fn_mask_index): Likewise.
12737 (internal_gather_scatter_fn_supported_p): Likewise.
12738 * internal-fn.c (gather_load_direct): New macro.
12739 (expand_gather_load_optab_fn): New function.
12740 (direct_gather_load_optab_supported_p): New macro.
12741 (direct_internal_fn_optab): New function.
12742 (internal_load_fn_p): Likewise.
12743 (internal_gather_scatter_fn_p): Likewise.
12744 (internal_fn_mask_index): Likewise.
12745 (internal_gather_scatter_fn_supported_p): Likewise.
12746 * optabs-query.c (supports_at_least_one_mode_p): New function.
12747 (supports_vec_gather_load_p): Likewise.
12748 * optabs-query.h (supports_vec_gather_load_p): Declare.
12749 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
12750 and memory_type field.
12751 (NUM_PATTERNS): Bump to 15.
12752 * tree-vect-data-refs.c: Include internal-fn.h.
12753 (vect_gather_scatter_fn_p): New function.
12754 (vect_describe_gather_scatter_call): Likewise.
12755 (vect_check_gather_scatter): Try using internal functions for
12756 gather loads. Recognize existing calls to a gather load function.
12757 (vect_analyze_data_refs): Consider using gather loads if
12758 supports_vec_gather_load_p.
12759 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
12760 (vect_get_gather_scatter_offset_type): Likewise.
12761 (vect_convert_mask_for_vectype): Likewise.
12762 (vect_add_conversion_to_patterm): Likewise.
12763 (vect_try_gather_scatter_pattern): Likewise.
12764 (vect_recog_gather_scatter_pattern): New pattern recognizer.
12765 (vect_vect_recog_func_ptrs): Add it.
12766 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
12767 internal_fn_mask_index and internal_gather_scatter_fn_p.
12768 (check_load_store_masking): Take the gather_scatter_info as an
12769 argument and handle gather loads.
12770 (vect_get_gather_scatter_ops): New function.
12771 (vectorizable_call): Check internal_load_fn_p.
12772 (vectorizable_load): Likewise. Handle gather load internal
12774 (vectorizable_store): Update call to check_load_store_masking.
12775 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
12776 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
12777 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
12778 (aarch64_gather_scale_operand_d): New predicates.
12779 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
12780 (mask_gather_load<mode>): New insns.
12782 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12783 Alan Hayward <alan.hayward@arm.com>
12784 David Sherwood <david.sherwood@arm.com>
12786 * optabs.def (fold_left_plus_optab): New optab.
12787 * doc/md.texi (fold_left_plus_@var{m}): Document.
12788 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
12789 * internal-fn.c (fold_left_direct): Define.
12790 (expand_fold_left_optab_fn): Likewise.
12791 (direct_fold_left_optab_supported_p): Likewise.
12792 * fold-const-call.c (fold_const_fold_left): New function.
12793 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
12794 * tree-parloops.c (valid_reduction_p): New function.
12795 (gather_scalar_reductions): Use it.
12796 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
12797 (vect_finish_replace_stmt): Declare.
12798 * tree-vect-loop.c (fold_left_reduction_fn): New function.
12799 (needs_fold_left_reduction_p): New function, split out from...
12800 (vect_is_simple_reduction): ...here. Accept reductions that
12801 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
12802 (vect_force_simple_reduction): Also store the reduction type in
12803 the assignment's STMT_VINFO_REDUC_TYPE.
12804 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
12805 (merge_with_identity): New function.
12806 (vect_expand_fold_left): Likewise.
12807 (vectorize_fold_left_reduction): Likewise.
12808 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
12809 scalar phi in place for it. Check for target support and reject
12810 cases that would reassociate the operation. Defer the transform
12811 phase to vectorize_fold_left_reduction.
12812 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
12813 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
12814 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
12816 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12818 * tree-if-conv.c (predicate_mem_writes): Remove redundant
12819 call to ifc_temp_var.
12821 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12822 Alan Hayward <alan.hayward@arm.com>
12823 David Sherwood <david.sherwood@arm.com>
12825 * target.def (legitimize_address_displacement): Take the original
12826 offset as a poly_int.
12827 * targhooks.h (default_legitimize_address_displacement): Update
12829 * targhooks.c (default_legitimize_address_displacement): Likewise.
12830 * doc/tm.texi: Regenerate.
12831 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
12832 as an argument, moving assert of ad->disp == ad->disp_term to...
12833 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
12834 Try calling targetm.legitimize_address_displacement before expanding
12835 the address rather than afterwards, and adjust for the new interface.
12836 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
12837 Match the new hook interface. Handle SVE addresses.
12838 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
12839 new hook interface.
12841 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12843 * Makefile.in (OBJS): Add early-remat.o.
12844 * target.def (select_early_remat_modes): New hook.
12845 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
12846 * doc/tm.texi: Regenerate.
12847 * targhooks.h (default_select_early_remat_modes): Declare.
12848 * targhooks.c (default_select_early_remat_modes): New function.
12849 * timevar.def (TV_EARLY_REMAT): New timevar.
12850 * passes.def (pass_early_remat): New pass.
12851 * tree-pass.h (make_pass_early_remat): Declare.
12852 * early-remat.c: New file.
12853 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
12855 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
12857 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12858 Alan Hayward <alan.hayward@arm.com>
12859 David Sherwood <david.sherwood@arm.com>
12861 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
12862 vfm1 with a bound_epilog parameter.
12863 (vect_do_peeling): Update calls accordingly, and move the prologue
12864 call earlier in the function. Treat the base bound_epilog as 0 for
12865 fully-masked loops and retain vf - 1 for other loops. Add 1 to
12866 this base when peeling for gaps.
12867 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
12868 with fully-masked loops.
12869 (vect_estimate_min_profitable_iters): Handle the single peeled
12870 iteration in that case.
12872 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12873 Alan Hayward <alan.hayward@arm.com>
12874 David Sherwood <david.sherwood@arm.com>
12876 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
12877 single-element interleaving even if the size is not a power of 2.
12878 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
12879 accesses for single-element interleaving if the group size is
12882 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12883 Alan Hayward <alan.hayward@arm.com>
12884 David Sherwood <david.sherwood@arm.com>
12886 * doc/md.texi (fold_extract_last_@var{m}): Document.
12887 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
12888 * optabs.def (fold_extract_last_optab): New optab.
12889 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
12890 * internal-fn.c (fold_extract_direct): New macro.
12891 (expand_fold_extract_optab_fn): Likewise.
12892 (direct_fold_extract_optab_supported_p): Likewise.
12893 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
12894 * tree-vect-loop.c (vect_model_reduction_cost): Handle
12895 EXTRACT_LAST_REDUCTION.
12896 (get_initial_def_for_reduction): Do not create an initial vector
12897 for EXTRACT_LAST_REDUCTION reductions.
12898 (vectorizable_reduction): Leave the scalar phi in place for
12899 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
12900 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
12901 epilogue code for EXTRACT_LAST_REDUCTION and defer the
12902 transform phase to vectorizable_condition.
12903 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
12905 (vect_finish_stmt_generation): ...here.
12906 (vect_finish_replace_stmt): New function.
12907 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
12908 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
12910 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
12912 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12913 Alan Hayward <alan.hayward@arm.com>
12914 David Sherwood <david.sherwood@arm.com>
12916 * doc/md.texi (extract_last_@var{m}): Document.
12917 * optabs.def (extract_last_optab): New optab.
12918 * internal-fn.def (EXTRACT_LAST): New internal function.
12919 * internal-fn.c (cond_unary_direct): New macro.
12920 (expand_cond_unary_optab_fn): Likewise.
12921 (direct_cond_unary_optab_supported_p): Likewise.
12922 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
12923 loops using EXTRACT_LAST.
12924 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
12925 (extract_last_<mode>): ...this optab.
12926 (vec_extract<mode><Vel>): Update accordingly.
12928 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12929 Alan Hayward <alan.hayward@arm.com>
12930 David Sherwood <david.sherwood@arm.com>
12932 * target.def (empty_mask_is_expensive): New hook.
12933 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
12934 * doc/tm.texi: Regenerate.
12935 * targhooks.h (default_empty_mask_is_expensive): Declare.
12936 * targhooks.c (default_empty_mask_is_expensive): New function.
12937 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
12938 if the target says that empty masks are expensive.
12939 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
12941 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
12943 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12944 Alan Hayward <alan.hayward@arm.com>
12945 David Sherwood <david.sherwood@arm.com>
12947 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
12948 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
12949 (vect_use_loop_mask_for_alignment_p): New function.
12950 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
12951 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
12952 niters_skip argument. Make sure that the first niters_skip elements
12953 of the first iteration are inactive.
12954 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
12955 Update call to vect_set_loop_masks_directly.
12956 (get_misalign_in_elems): New function, split out from...
12957 (vect_gen_prolog_loop_niters): ...here.
12958 (vect_update_init_of_dr): Take a code argument that specifies whether
12959 the adjustment should be added or subtracted.
12960 (vect_update_init_of_drs): Likewise.
12961 (vect_prepare_for_masked_peels): New function.
12962 (vect_do_peeling): Skip prologue peeling if we're using a mask
12963 instead. Update call to vect_update_inits_of_drs.
12964 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
12966 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
12967 alignment. Do not include the number of peeled iterations in
12968 the minimum threshold in that case.
12969 (vectorizable_induction): Adjust the start value down by
12970 LOOP_VINFO_MASK_SKIP_NITERS iterations.
12971 (vect_transform_loop): Call vect_prepare_for_masked_peels.
12972 Take the number of skipped iterations into account when calculating
12974 * tree-vect-stmts.c (vect_gen_while_not): New function.
12976 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12977 Alan Hayward <alan.hayward@arm.com>
12978 David Sherwood <david.sherwood@arm.com>
12980 * doc/sourcebuild.texi (vect_fully_masked): Document.
12981 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
12982 default value to 0.
12983 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
12985 (vect_analyze_loop_2): ...here. Don't check the vectorization
12986 factor against the number of loop iterations if the loop is
12989 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12990 Alan Hayward <alan.hayward@arm.com>
12991 David Sherwood <david.sherwood@arm.com>
12993 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
12994 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
12995 (dump_groups): Update accordingly.
12996 (iv_use::mem_type): New member variable.
12997 (address_p): New function.
12998 (record_use): Add a mem_type argument and initialize the new
13000 (record_group_use): Add a mem_type argument. Use address_p.
13001 Remove obsolete null checks of base_object. Update call to record_use.
13002 (find_interesting_uses_op): Update call to record_group_use.
13003 (find_interesting_uses_cond): Likewise.
13004 (find_interesting_uses_address): Likewise.
13005 (get_mem_type_for_internal_fn): New function.
13006 (find_address_like_use): Likewise.
13007 (find_interesting_uses_stmt): Try find_address_like_use before
13008 calling find_interesting_uses_op.
13009 (addr_offset_valid_p): Use the iv mem_type field as the type
13010 of the addressed memory.
13011 (add_autoinc_candidates): Likewise.
13012 (get_address_cost): Likewise.
13013 (split_small_address_groups_p): Use address_p.
13014 (split_address_groups): Likewise.
13015 (add_iv_candidate_for_use): Likewise.
13016 (autoinc_possible_for_pair): Likewise.
13017 (rewrite_groups): Likewise.
13018 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
13019 (determine_group_iv_cost): Update after split of USE_ADDRESS.
13020 (get_alias_ptr_type_for_ptr_address): New function.
13021 (rewrite_use_address): Rewrite address uses in calls that were
13022 identified by find_address_like_use.
13024 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13025 Alan Hayward <alan.hayward@arm.com>
13026 David Sherwood <david.sherwood@arm.com>
13028 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
13030 * gimple-expr.h (is_gimple_addressable: Likewise.
13031 * gimple-expr.c (is_gimple_address): Likewise.
13032 * internal-fn.c (expand_call_mem_ref): New function.
13033 (expand_mask_load_optab_fn): Use it.
13034 (expand_mask_store_optab_fn): Likewise.
13036 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13037 Alan Hayward <alan.hayward@arm.com>
13038 David Sherwood <david.sherwood@arm.com>
13040 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
13041 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
13042 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
13043 (cond_umax@var{mode}): Document.
13044 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
13045 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
13046 (cond_umin_optab, cond_umax_optab): New optabs.
13047 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
13048 (COND_IOR, COND_XOR): New internal functions.
13049 * internal-fn.h (get_conditional_internal_fn): Declare.
13050 * internal-fn.c (cond_binary_direct): New macro.
13051 (expand_cond_binary_optab_fn): Likewise.
13052 (direct_cond_binary_optab_supported_p): Likewise.
13053 (get_conditional_internal_fn): New function.
13054 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
13055 Cope with reduction statements that are vectorized as calls rather
13057 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
13058 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
13059 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
13060 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
13061 (UNSPEC_COND_EOR): New unspecs.
13062 (optab): Add mappings for them.
13063 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
13064 (sve_int_op, sve_fp_op): New int attributes.
13066 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13067 Alan Hayward <alan.hayward@arm.com>
13068 David Sherwood <david.sherwood@arm.com>
13070 * optabs.def (while_ult_optab): New optab.
13071 * doc/md.texi (while_ult@var{m}@var{n}): Document.
13072 * internal-fn.def (WHILE_ULT): New internal function.
13073 * internal-fn.h (direct_internal_fn_supported_p): New override
13074 that takes two types as argument.
13075 * internal-fn.c (while_direct): New macro.
13076 (expand_while_optab_fn): New function.
13077 (convert_optab_supported_p): Likewise.
13078 (direct_while_optab_supported_p): New macro.
13079 * wide-int.h (wi::udiv_ceil): New function.
13080 * tree-vectorizer.h (rgroup_masks): New structure.
13081 (vec_loop_masks): New typedef.
13082 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
13083 and fully_masked_p.
13084 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
13085 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
13086 (vect_max_vf): New function.
13087 (slpeel_make_loop_iterate_ntimes): Delete.
13088 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
13089 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
13090 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
13091 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
13092 internal-fn.h, stor-layout.h and optabs-query.h.
13093 (vect_set_loop_mask): New function.
13094 (add_preheader_seq): Likewise.
13095 (add_header_seq): Likewise.
13096 (interleave_supported_p): Likewise.
13097 (vect_maybe_permute_loop_masks): Likewise.
13098 (vect_set_loop_masks_directly): Likewise.
13099 (vect_set_loop_condition_masked): Likewise.
13100 (vect_set_loop_condition_unmasked): New function, split out from
13101 slpeel_make_loop_iterate_ntimes.
13102 (slpeel_make_loop_iterate_ntimes): Rename to..
13103 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
13104 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
13105 (vect_do_peeling): Update call accordingly.
13106 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
13108 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13109 mask_compare_type, can_fully_mask_p and fully_masked_p.
13110 (release_vec_loop_masks): New function.
13111 (_loop_vec_info): Use it to free the loop masks.
13112 (can_produce_all_loop_masks_p): New function.
13113 (vect_get_max_nscalars_per_iter): Likewise.
13114 (vect_verify_full_masking): Likewise.
13115 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
13116 retries, and free the mask rgroups before retrying. Check loop-wide
13117 reasons for disallowing fully-masked loops. Make the final decision
13118 about whether use a fully-masked loop or not.
13119 (vect_estimate_min_profitable_iters): Do not assume that peeling
13120 for the number of iterations will be needed for fully-masked loops.
13121 (vectorizable_reduction): Disable fully-masked loops.
13122 (vectorizable_live_operation): Likewise.
13123 (vect_halve_mask_nunits): New function.
13124 (vect_double_mask_nunits): Likewise.
13125 (vect_record_loop_mask): Likewise.
13126 (vect_get_loop_mask): Likewise.
13127 (vect_transform_loop): Handle the case in which the final loop
13128 iteration might handle a partial vector. Call vect_set_loop_condition
13129 instead of slpeel_make_loop_iterate_ntimes.
13130 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
13131 (check_load_store_masking): New function.
13132 (prepare_load_store_mask): Likewise.
13133 (vectorizable_store): Handle fully-masked loops.
13134 (vectorizable_load): Likewise.
13135 (supportable_widening_operation): Use vect_halve_mask_nunits for
13137 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
13138 (vect_gen_while): New function.
13139 * config/aarch64/aarch64.md (umax<mode>3): New expander.
13140 (aarch64_uqdec<mode>): New insn.
13142 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13143 Alan Hayward <alan.hayward@arm.com>
13144 David Sherwood <david.sherwood@arm.com>
13146 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
13147 (reduc_xor_scal_optab): New optabs.
13148 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
13149 (reduc_xor_scal_@var{m}): Document.
13150 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
13151 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
13152 internal functions.
13153 * fold-const-call.c (fold_const_call): Handle them.
13154 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
13155 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
13156 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
13157 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
13158 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
13159 (UNSPEC_XORV): New unspecs.
13160 (optab): Add entries for them.
13161 (BITWISEV): New int iterator.
13162 (bit_reduc_op): New int attributes.
13164 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13165 Alan Hayward <alan.hayward@arm.com>
13166 David Sherwood <david.sherwood@arm.com>
13168 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
13169 * internal-fn.def (VEC_SHL_INSERT): New internal function.
13170 * optabs.def (vec_shl_insert_optab): New optab.
13171 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
13172 (duplicate_and_interleave): Likewise.
13173 * tree-vect-loop.c: Include internal-fn.h.
13174 (neutral_op_for_slp_reduction): New function, split out from
13175 get_initial_defs_for_reduction.
13176 (get_initial_def_for_reduction): Handle option 2 for variable-length
13177 vectors by loading the neutral value into a vector and then shifting
13178 the initial value into element 0.
13179 (get_initial_defs_for_reduction): Replace the code argument with
13180 the neutral value calculated by neutral_op_for_slp_reduction.
13181 Use gimple_build_vector for constant-length vectors.
13182 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
13183 but the first group_size elements have a neutral value.
13184 Use duplicate_and_interleave otherwise.
13185 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
13186 Update call to get_initial_defs_for_reduction. Handle SLP
13187 reductions for variable-length vectors by creating one vector
13188 result for each scalar result, with the elements associated
13189 with other scalar results stubbed out with the neutral value.
13190 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
13191 Require IFN_VEC_SHL_INSERT for double reductions on
13192 variable-length vectors, or SLP reductions that have
13193 a neutral value. Require can_duplicate_and_interleave_p
13194 support for variable-length unchained SLP reductions if there
13195 is no neutral value, such as for MIN/MAX reductions. Also require
13196 the number of vector elements to be a multiple of the number of
13197 SLP statements when doing variable-length unchained SLP reductions.
13198 Update call to vect_create_epilog_for_reduction.
13199 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
13200 and remove initial values.
13201 (duplicate_and_interleave): Make public.
13202 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
13203 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
13205 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13206 Alan Hayward <alan.hayward@arm.com>
13207 David Sherwood <david.sherwood@arm.com>
13209 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
13210 (can_duplicate_and_interleave_p): New function.
13211 (vect_get_and_check_slp_defs): Take the vector of statements
13212 rather than just the current one. Remove excess parentheses.
13213 Restriction rejectinon of vect_constant_def and vect_external_def
13214 for variable-length vectors to boolean types, or types for which
13215 can_duplicate_and_interleave_p is false.
13216 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
13217 (duplicate_and_interleave): New function.
13218 (vect_get_constant_vectors): Use gimple_build_vector for
13219 constant-length vectors and suitable variable-length constant
13220 vectors. Use duplicate_and_interleave for other variable-length
13221 vectors. Don't defer the update when inserting new statements.
13223 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13224 Alan Hayward <alan.hayward@arm.com>
13225 David Sherwood <david.sherwood@arm.com>
13227 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
13228 min_profitable_iters doesn't go negative.
13230 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13231 Alan Hayward <alan.hayward@arm.com>
13232 David Sherwood <david.sherwood@arm.com>
13234 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
13235 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
13236 * optabs.def (vec_mask_load_lanes_optab): New optab.
13237 (vec_mask_store_lanes_optab): Likewise.
13238 * internal-fn.def (MASK_LOAD_LANES): New internal function.
13239 (MASK_STORE_LANES): Likewise.
13240 * internal-fn.c (mask_load_lanes_direct): New macro.
13241 (mask_store_lanes_direct): Likewise.
13242 (expand_mask_load_optab_fn): Handle masked operations.
13243 (expand_mask_load_lanes_optab_fn): New macro.
13244 (expand_mask_store_optab_fn): Handle masked operations.
13245 (expand_mask_store_lanes_optab_fn): New macro.
13246 (direct_mask_load_lanes_optab_supported_p): Likewise.
13247 (direct_mask_store_lanes_optab_supported_p): Likewise.
13248 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
13250 (vect_load_lanes_supported): Likewise.
13251 * tree-vect-data-refs.c (strip_conversion): New function.
13252 (can_group_stmts_p): Likewise.
13253 (vect_analyze_data_ref_accesses): Use it instead of checking
13254 for a pair of assignments.
13255 (vect_store_lanes_supported): Take a masked_p parameter.
13256 (vect_load_lanes_supported): Likewise.
13257 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
13258 vect_store_lanes_supported and vect_load_lanes_supported.
13259 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
13260 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
13261 parameter. Don't allow gaps for masked accesses.
13262 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
13263 and vect_load_lanes_supported.
13264 (get_load_store_type): Take a masked_p parameter and update
13265 call to get_group_load_store_type.
13266 (vectorizable_store): Update call to get_load_store_type.
13267 Handle IFN_MASK_STORE_LANES.
13268 (vectorizable_load): Update call to get_load_store_type.
13269 Handle IFN_MASK_LOAD_LANES.
13271 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13272 Alan Hayward <alan.hayward@arm.com>
13273 David Sherwood <david.sherwood@arm.com>
13275 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
13277 * config/aarch64/aarch64-protos.h
13278 (aarch64_sve_struct_memory_operand_p): Declare.
13279 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
13280 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
13281 (VPRED, vpred): Handle SVE structure modes.
13282 * config/aarch64/constraints.md (Utx): New constraint.
13283 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
13284 (aarch64_sve_struct_nonimmediate_operand): New predicates.
13285 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
13286 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
13287 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
13288 structure modes. Split into pieces after RA.
13289 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
13290 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
13292 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
13293 SVE structure modes.
13294 (aarch64_classify_address): Likewise.
13295 (sizetochar): Move earlier in file.
13296 (aarch64_print_operand): Handle SVE register lists.
13297 (aarch64_array_mode): New function.
13298 (aarch64_sve_struct_memory_operand_p): Likewise.
13299 (TARGET_ARRAY_MODE): Redefine.
13301 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13302 Alan Hayward <alan.hayward@arm.com>
13303 David Sherwood <david.sherwood@arm.com>
13305 * target.def (array_mode): New target hook.
13306 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
13307 * doc/tm.texi: Regenerate.
13308 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
13309 * hooks.c (hook_optmode_mode_uhwi_none): New function.
13310 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
13311 targetm.array_mode.
13312 * stor-layout.c (mode_for_array): Likewise. Support polynomial
13315 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13316 Alan Hayward <alan.hayward@arm.com>
13317 David Sherwood <david.sherwood@arm.com>
13319 * fold-const.c (fold_binary_loc): Check the argument types
13320 rather than the result type when testing for a vector operation.
13322 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13324 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
13325 * doc/tm.texi: Regenerate.
13327 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13328 Alan Hayward <alan.hayward@arm.com>
13329 David Sherwood <david.sherwood@arm.com>
13331 * doc/invoke.texi (-msve-vector-bits=): Document new option.
13332 (sve): Document new AArch64 extension.
13333 * doc/md.texi (w): Extend the description of the AArch64
13334 constraint to include SVE vectors.
13335 (Upl, Upa): Document new AArch64 predicate constraints.
13336 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
13338 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
13339 (msve-vector-bits=): New option.
13340 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
13341 SVE when these are disabled.
13342 (sve): New extension.
13343 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
13344 modes. Adjust their number of units based on aarch64_sve_vg.
13345 (MAX_BITSIZE_MODE_ANY_MODE): Define.
13346 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
13347 aarch64_addr_query_type.
13348 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
13349 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
13350 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
13351 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
13352 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
13353 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
13354 (aarch64_simd_imm_zero_p): Delete.
13355 (aarch64_check_zero_based_sve_index_immediate): Declare.
13356 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
13357 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
13358 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
13359 (aarch64_sve_float_mul_immediate_p): Likewise.
13360 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
13361 rather than an rtx.
13362 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
13363 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
13364 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
13365 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
13366 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
13367 (aarch64_regmode_natural_size): Likewise.
13368 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
13369 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
13371 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
13372 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
13373 for VG and the SVE predicate registers.
13374 (V_ALIASES): Add a "z"-prefixed alias.
13375 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
13376 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
13377 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
13378 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
13379 (REG_CLASS_NAMES): Add entries for them.
13380 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
13381 and the predicate registers.
13382 (aarch64_sve_vg): Declare.
13383 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
13384 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
13385 (REGMODE_NATURAL_SIZE): Define.
13386 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
13388 * config/aarch64/aarch64.c: Include cfgrtl.h.
13389 (simd_immediate_info): Add a constructor for series vectors,
13390 and an associated step field.
13391 (aarch64_sve_vg): New variable.
13392 (aarch64_dbx_register_number): Handle VG and the predicate registers.
13393 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
13394 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
13395 (VEC_ANY_DATA, VEC_STRUCT): New constants.
13396 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
13397 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
13398 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
13399 (aarch64_get_mask_mode): New functions.
13400 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
13401 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
13402 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
13403 predicate modes and predicate registers. Explicitly restrict
13404 GPRs to modes of 16 bytes or smaller. Only allow FP registers
13405 to store a vector mode if it is recognized by
13406 aarch64_classify_vector_mode.
13407 (aarch64_regmode_natural_size): New function.
13408 (aarch64_hard_regno_caller_save_mode): Return the original mode
13410 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
13411 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
13412 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
13413 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
13415 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
13416 does not overlap dest if the function is frame-related. Handle
13418 (aarch64_split_add_offset): New function.
13419 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
13420 them aarch64_add_offset.
13421 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
13422 and update call to aarch64_sub_sp.
13423 (aarch64_add_cfa_expression): New function.
13424 (aarch64_expand_prologue): Pass extra temporary registers to the
13425 functions above. Handle the case in which we need to emit new
13426 DW_CFA_expressions for registers that were originally saved
13427 relative to the stack pointer, but now have to be expressed
13428 relative to the frame pointer.
13429 (aarch64_output_mi_thunk): Pass extra temporary registers to the
13431 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
13432 IP0 and IP1 values for SVE frames.
13433 (aarch64_expand_vec_series): New function.
13434 (aarch64_expand_sve_widened_duplicate): Likewise.
13435 (aarch64_expand_sve_const_vector): Likewise.
13436 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
13437 Handle SVE constants. Use emit_move_insn to move a force_const_mem
13438 into the register, rather than emitting a SET directly.
13439 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
13440 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
13441 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
13442 (offset_9bit_signed_scaled_p): New functions.
13443 (aarch64_replicate_bitmask_imm): New function.
13444 (aarch64_bitmask_imm): Use it.
13445 (aarch64_cannot_force_const_mem): Reject expressions involving
13446 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
13447 (aarch64_classify_index): Handle SVE indices, by requiring
13448 a plain register index with a scale that matches the element size.
13449 (aarch64_classify_address): Handle SVE addresses. Assert that
13450 the mode of the address is VOIDmode or an integer mode.
13451 Update call to aarch64_classify_symbol.
13452 (aarch64_classify_symbolic_expression): Update call to
13453 aarch64_classify_symbol.
13454 (aarch64_const_vec_all_in_range_p): New function.
13455 (aarch64_print_vector_float_operand): Likewise.
13456 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
13457 "vN" for FP registers with SVE modes. Handle (const ...) vectors
13458 and the FP immediates 1.0 and 0.5.
13459 (aarch64_print_address_internal): Handle SVE addresses.
13460 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
13461 (aarch64_regno_regclass): Handle predicate registers.
13462 (aarch64_secondary_reload): Handle big-endian reloads of SVE
13464 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
13465 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
13466 (aarch64_convert_sve_vector_bits): New function.
13467 (aarch64_override_options): Use it to handle -msve-vector-bits=.
13468 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
13469 rather than an rtx.
13470 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
13471 Handle SVE vector and predicate modes. Accept VL-based constants
13472 that need only one temporary register, and VL offsets that require
13473 no temporary registers.
13474 (aarch64_conditional_register_usage): Mark the predicate registers
13475 as fixed if SVE isn't available.
13476 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
13477 Return true for SVE vector and predicate modes.
13478 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
13479 rather than an unsigned int. Handle SVE modes.
13480 (aarch64_preferred_simd_mode): Update call accordingly. Handle
13482 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
13484 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
13485 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
13486 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
13487 (aarch64_sve_float_mul_immediate_p): New functions.
13488 (aarch64_sve_valid_immediate): New function.
13489 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
13490 Explicitly reject structure modes. Check for INDEX constants.
13491 Handle PTRUE and PFALSE constants.
13492 (aarch64_check_zero_based_sve_index_immediate): New function.
13493 (aarch64_simd_imm_zero_p): Delete.
13494 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
13495 vector modes. Accept constants in the range of CNT[BHWD].
13496 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
13497 ask for an Advanced SIMD mode.
13498 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
13499 (aarch64_simd_vector_alignment): Handle SVE predicates.
13500 (aarch64_vectorize_preferred_vector_alignment): New function.
13501 (aarch64_simd_vector_alignment_reachable): Use it instead of
13503 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
13504 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
13506 (MAX_VECT_LEN): Delete.
13507 (expand_vec_perm_d): Add a vec_flags field.
13508 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
13509 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
13510 (aarch64_evpc_ext): Don't apply a big-endian lane correction
13512 (aarch64_evpc_rev): Rename to...
13513 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
13514 (aarch64_evpc_rev_global): New function.
13515 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
13516 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
13518 (aarch64_evpc_sve_tbl): New function.
13519 (aarch64_expand_vec_perm_const_1): Update after rename of
13520 aarch64_evpc_rev. Handle SVE permutes too, trying
13521 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
13522 than aarch64_evpc_tbl.
13523 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
13524 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
13525 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
13526 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
13527 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
13528 (aarch64_expand_sve_vcond): New functions.
13529 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
13530 of aarch64_vector_mode_p.
13531 (aarch64_dwarf_poly_indeterminate_value): New function.
13532 (aarch64_compute_pressure_classes): Likewise.
13533 (aarch64_can_change_mode_class): Likewise.
13534 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
13535 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
13536 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
13537 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
13538 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
13539 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
13540 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
13541 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
13543 (Dn, Dl, Dr): Accept const as well as const_vector.
13544 (Dz): Likewise. Compare against CONST0_RTX.
13545 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
13546 of "vector" where appropriate.
13547 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
13548 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
13549 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
13550 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
13551 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
13552 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
13553 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
13554 (v_int_equiv): Extend to SVE modes.
13555 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
13557 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
13558 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
13559 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
13560 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
13561 (SVE_COND_FP_CMP): New int iterators.
13562 (perm_hilo): Handle the new unpack unspecs.
13563 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
13565 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
13566 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
13567 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
13568 (aarch64_equality_operator, aarch64_constant_vector_operand)
13569 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
13570 (aarch64_sve_nonimmediate_operand): Likewise.
13571 (aarch64_sve_general_operand): Likewise.
13572 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
13573 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
13574 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
13575 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
13576 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
13577 (aarch64_sve_float_arith_immediate): Likewise.
13578 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
13579 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
13580 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
13581 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
13582 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
13583 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
13584 (aarch64_sve_float_arith_operand): Likewise.
13585 (aarch64_sve_float_arith_with_sub_operand): Likewise.
13586 (aarch64_sve_float_mul_operand): Likewise.
13587 (aarch64_sve_vec_perm_operand): Likewise.
13588 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
13589 (aarch64_mov_operand): Accept const_poly_int and const_vector.
13590 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
13591 as well as const_vector.
13592 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
13593 in file. Use CONST0_RTX and CONSTM1_RTX.
13594 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
13595 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
13596 Use aarch64_simd_imm_zero.
13597 * config/aarch64/aarch64-sve.md: New file.
13598 * config/aarch64/aarch64.md: Include it.
13599 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
13600 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
13601 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
13602 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
13603 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
13604 (sve): New attribute.
13605 (enabled): Disable instructions with the sve attribute unless
13607 (movqi, movhi): Pass CONST_POLY_INT operaneds through
13608 aarch64_expand_mov_immediate.
13609 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
13610 CNT[BHSD] immediates.
13611 (movti): Split CONST_POLY_INT moves into two halves.
13612 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
13613 Split additions that need a temporary here if the destination
13614 is the stack pointer.
13615 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
13616 (*add<mode>3_poly_1): New instruction.
13617 (set_clobber_cc): New expander.
13619 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13621 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
13622 parameter and use it instead of GET_MODE_SIZE (innermode). Use
13623 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
13624 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
13625 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
13626 Change innermode from fixed_mode_size to machine_mode.
13627 (simplify_subreg): Update call accordingly. Handle a constant-sized
13628 subreg of a variable-length CONST_VECTOR.
13630 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13631 Alan Hayward <alan.hayward@arm.com>
13632 David Sherwood <david.sherwood@arm.com>
13634 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
13635 (add_offset_to_base): New function, split out from...
13636 (create_mem_ref): ...here. When handling a scale other than 1,
13637 check first whether the address is valid without the offset.
13638 Add it into the base if so, leaving the index and scale as-is.
13640 2018-01-12 Jakub Jelinek <jakub@redhat.com>
13643 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
13644 fold_for_warn before checking if arg2 is INTEGER_CST.
13646 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
13648 * config/rs6000/predicates.md (load_multiple_operation): Delete.
13649 (store_multiple_operation): Delete.
13650 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
13651 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
13652 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
13653 guarded by TARGET_STRING.
13654 (rs6000_output_load_multiple): Delete.
13655 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
13656 OPTION_MASK_STRING / TARGET_STRING handling.
13657 (print_operand) <'N', 'O'>: Add comment that these are unused now.
13658 (const rs6000_opt_masks) <"string">: Change mask to 0.
13659 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
13660 (MASK_STRING): Delete.
13661 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
13663 (load_multiple): Delete.
13670 (store_multiple): Delete.
13677 (movmemsi_8reg): Delete.
13678 (corresponding unnamed define_insn): Delete.
13679 (movmemsi_6reg): Delete.
13680 (corresponding unnamed define_insn): Delete.
13681 (movmemsi_4reg): Delete.
13682 (corresponding unnamed define_insn): Delete.
13683 (movmemsi_2reg): Delete.
13684 (corresponding unnamed define_insn): Delete.
13685 (movmemsi_1reg): Delete.
13686 (corresponding unnamed define_insn): Delete.
13687 * config/rs6000/rs6000.opt (mno-string): New.
13688 (mstring): Replace by deprecation warning stub.
13689 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
13691 2018-01-12 Jakub Jelinek <jakub@redhat.com>
13693 * regrename.c (regrename_do_replace): If replacing the same
13694 reg multiple times, try to reuse last created gen_raw_REG.
13697 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
13698 main to workaround a bug in GDB.
13700 2018-01-12 Tom de Vries <tom@codesourcery.com>
13703 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
13705 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
13707 PR rtl-optimization/80481
13708 * ira-color.c (get_cap_member): New function.
13709 (allocnos_conflict_by_live_ranges_p): Use it.
13710 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
13711 (setup_slot_coalesced_allocno_live_ranges): Ditto.
13713 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
13716 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
13717 (*saddl_se_1): Ditto.
13718 (*ssubsi_1): Ditto.
13719 (*ssubl_se_1): Ditto.
13721 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13723 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
13724 rather than wi::to_widest for DR_INITs.
13725 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
13726 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
13727 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
13729 (vect_analyze_group_access_1): Note that here.
13731 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13733 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
13734 polynomial type sizes.
13736 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13738 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
13739 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
13740 (gimple_add_tmp_var): Likewise.
13742 2018-01-12 Martin Liska <mliska@suse.cz>
13744 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
13745 (gimple_alloc_sizes): Likewise.
13746 (dump_gimple_statistics): Use PRIu64 in printf format.
13747 * gimple.h: Change uint64_t to int.
13749 2018-01-12 Martin Liska <mliska@suse.cz>
13751 * tree-core.h: Use uint64_t instead of int.
13752 * tree.c (tree_node_counts): Likewise.
13753 (tree_node_sizes): Likewise.
13754 (dump_tree_statistics): Use PRIu64 in printf format.
13756 2018-01-12 Martin Liska <mliska@suse.cz>
13758 * Makefile.in: As qsort_chk is implemented in vec.c, add
13759 vec.o to linkage of gencfn-macros.
13760 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
13761 passing the info to record_node_allocation_statistics.
13762 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
13764 * ggc-common.c (struct ggc_usage): Add operator== and use
13765 it in operator< and compare function.
13766 * mem-stats.h (struct mem_usage): Likewise.
13767 * vec.c (struct vec_usage): Remove operator< and compare
13768 function. Can be simply inherited.
13770 2018-01-12 Martin Jambor <mjambor@suse.cz>
13773 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
13774 * tree-ssa-math-opts.c: Include domwalk.h.
13775 (convert_mult_to_fma_1): New function.
13776 (fma_transformation_info): New type.
13777 (fma_deferring_state): Likewise.
13778 (cancel_fma_deferring): New function.
13779 (result_of_phi): Likewise.
13780 (last_fma_candidate_feeds_initial_phi): Likewise.
13781 (convert_mult_to_fma): Added deferring logic, split actual
13782 transformation to convert_mult_to_fma_1.
13783 (math_opts_dom_walker): New type.
13784 (math_opts_dom_walker::after_dom_children): New method, body moved
13785 here from pass_optimize_widening_mul::execute, added deferring logic
13787 (pass_optimize_widening_mul::execute): Moved most of code to
13788 math_opts_dom_walker::after_dom_children.
13789 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
13790 * config/i386/i386.c (ix86_option_override_internal): Added
13791 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
13793 2018-01-12 Richard Biener <rguenther@suse.de>
13796 * dwarf2out.c (gen_variable_die): Do not reset old_die for
13797 inline instance vars.
13799 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
13802 * config/rx/rx.c (rx_is_restricted_memory_address):
13803 Handle SUBREG case.
13805 2018-01-12 Richard Biener <rguenther@suse.de>
13807 PR tree-optimization/80846
13808 * target.def (split_reduction): New target hook.
13809 * targhooks.c (default_split_reduction): New function.
13810 * targhooks.h (default_split_reduction): Declare.
13811 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
13812 target requests first reduce vectors by combining low and high
13814 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
13815 (get_vectype_for_scalar_type_and_size): Export.
13816 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
13817 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
13818 * doc/tm.texi: Regenerate.
13819 * config/i386/i386.c (ix86_split_reduction): Implement
13820 TARGET_VECTORIZE_SPLIT_REDUCTION.
13822 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
13825 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
13826 in PIC mode except for TARGET_VXWORKS_RTP.
13827 * config/sparc/sparc.c: Include cfgrtl.h.
13828 (TARGET_INIT_PIC_REG): Define.
13829 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
13830 (sparc_pic_register_p): New predicate.
13831 (sparc_legitimate_address_p): Use it.
13832 (sparc_legitimize_pic_address): Likewise.
13833 (sparc_delegitimize_address): Likewise.
13834 (sparc_mode_dependent_address_p): Likewise.
13835 (gen_load_pcrel_sym): Remove 4th parameter.
13836 (load_got_register): Adjust call to above. Remove obsolete stuff.
13837 (sparc_expand_prologue): Do not call load_got_register here.
13838 (sparc_flat_expand_prologue): Likewise.
13839 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
13840 (sparc_use_pseudo_pic_reg): New function.
13841 (sparc_init_pic_reg): Likewise.
13842 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
13843 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
13845 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
13847 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
13848 Add item for branch_cost.
13850 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
13852 PR rtl-optimization/83565
13853 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
13854 not extend the result to a larger mode for rotate operations.
13855 (num_sign_bit_copies1): Likewise.
13857 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13860 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
13862 Use values-Xc.o for -pedantic.
13863 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
13865 2018-01-12 Martin Liska <mliska@suse.cz>
13868 * ipa-devirt.c (final_warning_record::grow_type_warnings):
13870 (possible_polymorphic_call_targets): Use it.
13871 (ipa_devirt): Likewise.
13873 2018-01-12 Martin Liska <mliska@suse.cz>
13875 * profile-count.h (enum profile_quality): Use 0 as invalid
13876 enum value of profile_quality.
13878 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
13880 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
13881 -mext-string options.
13883 2018-01-12 Richard Biener <rguenther@suse.de>
13885 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
13886 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
13887 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
13889 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
13891 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
13893 * configure.ac (--with-long-double-format): Add support for the
13894 configuration option to change the default long double format on
13896 * config.gcc (powerpc*-linux*-*): Likewise.
13897 * configure: Regenerate.
13898 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
13899 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
13900 used without modification.
13902 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13904 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
13905 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
13906 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
13907 MISC_BUILTIN_SPEC_BARRIER.
13908 (rs6000_init_builtins): Likewise.
13909 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
13911 (speculation_barrier): New define_insn.
13912 * doc/extend.texi: Document __builtin_speculation_barrier.
13914 2018-01-11 Jakub Jelinek <jakub@redhat.com>
13917 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
13918 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
13919 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
13921 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
13922 integral modes instead of "ss" and "sd".
13923 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
13924 vectors with 32-bit and 64-bit elements.
13925 (vecdupssescalarmodesuffix): New mode attribute.
13926 (vec_dup<mode>): Use it.
13928 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
13931 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
13932 frame if argument is passed on stack.
13934 2018-01-11 Jakub Jelinek <jakub@redhat.com>
13937 * ree.c (combine_reaching_defs): Optimize also
13938 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
13939 reg2=any_extend(exp); reg1=reg2;, formatting fix.
13941 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
13943 PR middle-end/83189
13944 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
13946 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
13948 PR middle-end/83718
13949 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
13950 after they are computed.
13952 2018-01-11 Bin Cheng <bin.cheng@arm.com>
13954 PR tree-optimization/83695
13955 * gimple-loop-linterchange.cc
13956 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
13957 reset cached scev information after interchange.
13958 (pass_linterchange::execute): Remove call to scev_reset_htab.
13960 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13962 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
13963 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
13964 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
13965 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
13966 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
13967 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
13968 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
13969 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
13970 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
13971 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
13972 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
13973 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
13974 (V_lane_reg): Likewise.
13975 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
13977 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
13978 (vfmal_lane_low<mode>_intrinsic,
13979 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
13980 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
13981 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
13982 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
13983 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
13984 vfmsl_lane_high<mode>_intrinsic): New define_insns.
13986 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13988 * config/arm/arm-cpus.in (fp16fml): New feature.
13989 (ALL_SIMD): Add fp16fml.
13990 (armv8.2-a): Add fp16fml as an option.
13991 (armv8.3-a): Likewise.
13992 (armv8.4-a): Add fp16fml as part of fp16.
13993 * config/arm/arm.h (TARGET_FP16FML): Define.
13994 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
13996 * config/arm/arm-modes.def (V2HF): Define.
13997 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
13998 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
13999 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
14000 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
14001 vfmsl_low, vfmsl_high): New set of builtins.
14002 * config/arm/iterators.md (PLUSMINUS): New code iterator.
14003 (vfml_op): New code attribute.
14004 (VFMLHALVES): New int iterator.
14005 (VFML, VFMLSEL): New mode attributes.
14006 (V_reg): Define mapping for V2HF.
14007 (V_hi, V_lo): New mode attributes.
14008 (VF_constraint): Likewise.
14009 (vfml_half, vfml_half_selector): New int attributes.
14010 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
14012 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
14013 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
14015 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
14016 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
14017 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
14018 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
14020 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
14021 Document new effective target and option set.
14023 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14025 * config/arm/arm-cpus.in (armv8_4): New feature.
14026 (ARMv8_4a): New fgroup.
14027 (armv8.4-a): New arch.
14028 * config/arm/arm-tables.opt: Regenerate.
14029 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
14030 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
14031 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
14032 Add matching rules for -march=armv8.4-a and extensions.
14033 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
14035 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
14038 * config/rx/rx.md (BW): New mode attribute.
14039 (sync_lock_test_and_setsi): Add mode suffix to insn output.
14041 2018-01-11 Richard Biener <rguenther@suse.de>
14043 PR tree-optimization/83435
14044 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
14045 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
14046 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
14048 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14049 Alan Hayward <alan.hayward@arm.com>
14050 David Sherwood <david.sherwood@arm.com>
14052 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
14054 (aarch64_classify_address): Initialize it. Track polynomial offsets.
14055 (aarch64_print_address_internal): Use it to check for a zero offset.
14057 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14058 Alan Hayward <alan.hayward@arm.com>
14059 David Sherwood <david.sherwood@arm.com>
14061 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
14062 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
14063 Return a poly_int64 rather than a HOST_WIDE_INT.
14064 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
14065 rather than a HOST_WIDE_INT.
14066 * config/aarch64/aarch64.h (aarch64_frame): Protect with
14067 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
14068 hard_fp_offset, frame_size, initial_adjust, callee_offset and
14069 final_offset from HOST_WIDE_INT to poly_int64.
14070 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
14071 to_constant when getting the number of units in an Advanced SIMD
14073 (aarch64_builtin_vectorized_function): Check for a constant number
14075 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
14077 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
14078 attribute instead of GET_MODE_NUNITS.
14079 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
14080 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
14081 GET_MODE_SIZE for fixed-size registers.
14082 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
14083 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
14084 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
14085 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
14086 (aarch64_print_operand, aarch64_print_address_internal)
14087 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
14088 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
14089 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
14090 Handle polynomial GET_MODE_SIZE.
14091 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
14092 wider than SImode without modification.
14093 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
14094 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
14095 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
14096 passing and returning SVE modes.
14097 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
14098 rather than GEN_INT.
14099 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
14100 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
14101 (aarch64_allocate_and_probe_stack_space): Likewise.
14102 (aarch64_layout_frame): Cope with polynomial offsets.
14103 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
14104 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
14105 polynomial offsets.
14106 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
14107 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
14108 poly_int64 rather than a HOST_WIDE_INT.
14109 (aarch64_get_separate_components, aarch64_process_components)
14110 (aarch64_expand_prologue, aarch64_expand_epilogue)
14111 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
14112 (aarch64_anchor_offset): New function, split out from...
14113 (aarch64_legitimize_address): ...here.
14114 (aarch64_builtin_vectorization_cost): Handle polynomial
14115 TYPE_VECTOR_SUBPARTS.
14116 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
14118 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
14119 number of elements from the PARALLEL rather than the mode.
14120 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
14121 rather than GET_MODE_BITSIZE.
14122 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
14123 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
14124 (aarch64_expand_vec_perm_const_1): Handle polynomial
14125 d->perm.length () and d->perm elements.
14126 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
14127 Apply to_constant to d->perm elements.
14128 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
14129 polynomial CONST_VECTOR_NUNITS.
14130 (aarch64_move_pointer): Take amount as a poly_int64 rather
14132 (aarch64_progress_pointer): Avoid temporary variable.
14133 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
14134 the mode attribute instead of GET_MODE.
14136 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14137 Alan Hayward <alan.hayward@arm.com>
14138 David Sherwood <david.sherwood@arm.com>
14140 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
14141 x exists before using it.
14142 (aarch64_add_constant_internal): Rename to...
14143 (aarch64_add_offset_1): ...this. Replace regnum with separate
14144 src and dest rtxes. Handle the case in which they're different,
14145 including when the offset is zero. Replace scratchreg with an rtx.
14146 Use 2 additions if there is no spare register into which we can
14147 move a 16-bit constant.
14148 (aarch64_add_constant): Delete.
14149 (aarch64_add_offset): Replace reg with separate src and dest
14150 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
14151 Use aarch64_add_offset_1.
14152 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
14153 an rtx rather than an int. Take the delta as a poly_int64
14154 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
14155 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
14156 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
14157 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
14158 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
14159 and aarch64_add_sp.
14160 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
14161 aarch64_add_constant.
14163 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14165 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
14166 Use scalar_float_mode.
14168 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14170 * config/aarch64/aarch64-simd.md
14171 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
14172 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
14173 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
14174 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
14175 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
14176 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
14177 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
14178 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
14179 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
14180 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
14182 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14185 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
14186 targ_options->x_arm_arch_string is non NULL.
14188 2018-01-11 Tamar Christina <tamar.christina@arm.com>
14190 * config/aarch64/aarch64.h
14191 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
14193 2018-01-11 Sudakshina Das <sudi.das@arm.com>
14196 * expmed.c (emit_store_flag_force): Swap if const op0
14197 and change VOIDmode to mode of op0.
14199 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14201 PR rtl-optimization/83761
14202 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
14203 than bytes to mode_for_size.
14205 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14207 PR middle-end/83189
14208 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
14209 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
14212 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14214 PR middle-end/83575
14215 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
14216 when in layout mode.
14217 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
14218 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
14221 2018-01-10 Michael Collison <michael.collison@arm.com>
14223 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
14224 * config/aarch64/aarch64-option-extension.def: Add
14225 AARCH64_OPT_EXTENSION of 'fp16fml'.
14226 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14227 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
14228 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
14229 * config/aarch64/constraints.md (Ui7): New constraint.
14230 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
14231 (VFMLA_SEL_W): Ditto.
14234 (VFMLA16_LOW): New int iterator.
14235 (VFMLA16_HIGH): Ditto.
14236 (UNSPEC_FMLAL): New unspec.
14237 (UNSPEC_FMLSL): Ditto.
14238 (UNSPEC_FMLAL2): Ditto.
14239 (UNSPEC_FMLSL2): Ditto.
14240 (f16mac): New code attribute.
14241 * config/aarch64/aarch64-simd-builtins.def
14242 (aarch64_fmlal_lowv2sf): Ditto.
14243 (aarch64_fmlsl_lowv2sf): Ditto.
14244 (aarch64_fmlalq_lowv4sf): Ditto.
14245 (aarch64_fmlslq_lowv4sf): Ditto.
14246 (aarch64_fmlal_highv2sf): Ditto.
14247 (aarch64_fmlsl_highv2sf): Ditto.
14248 (aarch64_fmlalq_highv4sf): Ditto.
14249 (aarch64_fmlslq_highv4sf): Ditto.
14250 (aarch64_fmlal_lane_lowv2sf): Ditto.
14251 (aarch64_fmlsl_lane_lowv2sf): Ditto.
14252 (aarch64_fmlal_laneq_lowv2sf): Ditto.
14253 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
14254 (aarch64_fmlalq_lane_lowv4sf): Ditto.
14255 (aarch64_fmlsl_lane_lowv4sf): Ditto.
14256 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
14257 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
14258 (aarch64_fmlal_lane_highv2sf): Ditto.
14259 (aarch64_fmlsl_lane_highv2sf): Ditto.
14260 (aarch64_fmlal_laneq_highv2sf): Ditto.
14261 (aarch64_fmlsl_laneq_highv2sf): Ditto.
14262 (aarch64_fmlalq_lane_highv4sf): Ditto.
14263 (aarch64_fmlsl_lane_highv4sf): Ditto.
14264 (aarch64_fmlalq_laneq_highv4sf): Ditto.
14265 (aarch64_fmlsl_laneq_highv4sf): Ditto.
14266 * config/aarch64/aarch64-simd.md:
14267 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
14268 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
14269 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
14270 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
14271 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
14272 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
14273 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
14274 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
14275 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
14276 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
14277 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
14278 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
14279 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
14280 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
14281 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
14282 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
14283 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
14284 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
14285 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
14286 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
14287 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
14288 (vfmlsl_low_u32): Ditto.
14289 (vfmlalq_low_u32): Ditto.
14290 (vfmlslq_low_u32): Ditto.
14291 (vfmlal_high_u32): Ditto.
14292 (vfmlsl_high_u32): Ditto.
14293 (vfmlalq_high_u32): Ditto.
14294 (vfmlslq_high_u32): Ditto.
14295 (vfmlal_lane_low_u32): Ditto.
14296 (vfmlsl_lane_low_u32): Ditto.
14297 (vfmlal_laneq_low_u32): Ditto.
14298 (vfmlsl_laneq_low_u32): Ditto.
14299 (vfmlalq_lane_low_u32): Ditto.
14300 (vfmlslq_lane_low_u32): Ditto.
14301 (vfmlalq_laneq_low_u32): Ditto.
14302 (vfmlslq_laneq_low_u32): Ditto.
14303 (vfmlal_lane_high_u32): Ditto.
14304 (vfmlsl_lane_high_u32): Ditto.
14305 (vfmlal_laneq_high_u32): Ditto.
14306 (vfmlsl_laneq_high_u32): Ditto.
14307 (vfmlalq_lane_high_u32): Ditto.
14308 (vfmlslq_lane_high_u32): Ditto.
14309 (vfmlalq_laneq_high_u32): Ditto.
14310 (vfmlslq_laneq_high_u32): Ditto.
14311 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
14312 (AARCH64_FL_FOR_ARCH8_4): New.
14313 (AARCH64_ISA_F16FML): New ISA flag.
14314 (TARGET_F16FML): New feature flag for fp16fml.
14315 (doc/invoke.texi): Document new fp16fml option.
14317 2018-01-10 Michael Collison <michael.collison@arm.com>
14319 * config/aarch64/aarch64-builtins.c:
14320 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
14321 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14322 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
14323 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
14324 (AARCH64_ISA_SHA3): New ISA flag.
14325 (TARGET_SHA3): New feature flag for sha3.
14326 * config/aarch64/iterators.md (sha512_op): New int attribute.
14327 (CRYPTO_SHA512): New int iterator.
14328 (UNSPEC_SHA512H): New unspec.
14329 (UNSPEC_SHA512H2): Ditto.
14330 (UNSPEC_SHA512SU0): Ditto.
14331 (UNSPEC_SHA512SU1): Ditto.
14332 * config/aarch64/aarch64-simd-builtins.def
14333 (aarch64_crypto_sha512hqv2di): New builtin.
14334 (aarch64_crypto_sha512h2qv2di): Ditto.
14335 (aarch64_crypto_sha512su0qv2di): Ditto.
14336 (aarch64_crypto_sha512su1qv2di): Ditto.
14337 (aarch64_eor3qv8hi): Ditto.
14338 (aarch64_rax1qv2di): Ditto.
14339 (aarch64_xarqv2di): Ditto.
14340 (aarch64_bcaxqv8hi): Ditto.
14341 * config/aarch64/aarch64-simd.md:
14342 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
14343 (aarch64_crypto_sha512su0qv2di): Ditto.
14344 (aarch64_crypto_sha512su1qv2di): Ditto.
14345 (aarch64_eor3qv8hi): Ditto.
14346 (aarch64_rax1qv2di): Ditto.
14347 (aarch64_xarqv2di): Ditto.
14348 (aarch64_bcaxqv8hi): Ditto.
14349 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
14350 (vsha512h2q_u64): Ditto.
14351 (vsha512su0q_u64): Ditto.
14352 (vsha512su1q_u64): Ditto.
14353 (veor3q_u16): Ditto.
14354 (vrax1q_u64): Ditto.
14355 (vxarq_u64): Ditto.
14356 (vbcaxq_u16): Ditto.
14357 * config/arm/types.md (crypto_sha512): New type attribute.
14358 (crypto_sha3): Ditto.
14359 (doc/invoke.texi): Document new sha3 option.
14361 2018-01-10 Michael Collison <michael.collison@arm.com>
14363 * config/aarch64/aarch64-builtins.c:
14364 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
14365 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14366 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
14367 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
14368 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
14369 (AARCH64_ISA_SM4): New ISA flag.
14370 (TARGET_SM4): New feature flag for sm4.
14371 * config/aarch64/aarch64-simd-builtins.def
14372 (aarch64_sm3ss1qv4si): Ditto.
14373 (aarch64_sm3tt1aq4si): Ditto.
14374 (aarch64_sm3tt1bq4si): Ditto.
14375 (aarch64_sm3tt2aq4si): Ditto.
14376 (aarch64_sm3tt2bq4si): Ditto.
14377 (aarch64_sm3partw1qv4si): Ditto.
14378 (aarch64_sm3partw2qv4si): Ditto.
14379 (aarch64_sm4eqv4si): Ditto.
14380 (aarch64_sm4ekeyqv4si): Ditto.
14381 * config/aarch64/aarch64-simd.md:
14382 (aarch64_sm3ss1qv4si): Ditto.
14383 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
14384 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
14385 (aarch64_sm4eqv4si): Ditto.
14386 (aarch64_sm4ekeyqv4si): Ditto.
14387 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
14388 (sm3part_op): Ditto.
14389 (CRYPTO_SM3TT): Ditto.
14390 (CRYPTO_SM3PART): Ditto.
14391 (UNSPEC_SM3SS1): New unspec.
14392 (UNSPEC_SM3TT1A): Ditto.
14393 (UNSPEC_SM3TT1B): Ditto.
14394 (UNSPEC_SM3TT2A): Ditto.
14395 (UNSPEC_SM3TT2B): Ditto.
14396 (UNSPEC_SM3PARTW1): Ditto.
14397 (UNSPEC_SM3PARTW2): Ditto.
14398 (UNSPEC_SM4E): Ditto.
14399 (UNSPEC_SM4EKEY): Ditto.
14400 * config/aarch64/constraints.md (Ui2): New constraint.
14401 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
14402 * config/arm/types.md (crypto_sm3): New type attribute.
14403 (crypto_sm4): Ditto.
14404 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
14405 (vsm3tt1aq_u32): Ditto.
14406 (vsm3tt1bq_u32): Ditto.
14407 (vsm3tt2aq_u32): Ditto.
14408 (vsm3tt2bq_u32): Ditto.
14409 (vsm3partw1q_u32): Ditto.
14410 (vsm3partw2q_u32): Ditto.
14411 (vsm4eq_u32): Ditto.
14412 (vsm4ekeyq_u32): Ditto.
14413 (doc/invoke.texi): Document new sm4 option.
14415 2018-01-10 Michael Collison <michael.collison@arm.com>
14417 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
14418 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
14419 (AARCH64_FL_FOR_ARCH8_4): New.
14420 (AARCH64_FL_V8_4): New flag.
14421 (doc/invoke.texi): Document new armv8.4-a option.
14423 2018-01-10 Michael Collison <michael.collison@arm.com>
14425 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14426 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
14427 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
14428 * config/aarch64/aarch64-option-extension.def: Add
14429 AARCH64_OPT_EXTENSION of 'sha2'.
14430 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
14431 (crypto): Disable sha2 and aes if crypto disabled.
14432 (crypto): Enable aes and sha2 if enabled.
14433 (simd): Disable sha2 and aes if simd disabled.
14434 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
14436 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
14437 (TARGET_SHA2): New feature flag for sha2.
14438 (TARGET_AES): New feature flag for aes.
14439 * config/aarch64/aarch64-simd.md:
14440 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
14441 conditional on TARGET_AES.
14442 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
14443 (aarch64_crypto_sha1hsi): Make pattern conditional
14445 (aarch64_crypto_sha1hv4si): Ditto.
14446 (aarch64_be_crypto_sha1hv4si): Ditto.
14447 (aarch64_crypto_sha1su1v4si): Ditto.
14448 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
14449 (aarch64_crypto_sha1su0v4si): Ditto.
14450 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
14451 (aarch64_crypto_sha256su0v4si): Ditto.
14452 (aarch64_crypto_sha256su1v4si): Ditto.
14453 (doc/invoke.texi): Document new aes and sha2 options.
14455 2018-01-10 Martin Sebor <msebor@redhat.com>
14457 PR tree-optimization/83781
14458 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
14461 2018-01-11 Martin Sebor <msebor@gmail.com>
14462 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14464 PR tree-optimization/83501
14465 PR tree-optimization/81703
14467 * tree-ssa-strlen.c (get_string_cst): Rename...
14468 (get_string_len): ...to this. Handle global constants.
14469 (handle_char_store): Adjust.
14471 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
14472 Jim Wilson <jimw@sifive.com>
14474 * config/riscv/riscv-protos.h (riscv_output_return): New.
14475 * config/riscv/riscv.c (struct machine_function): New naked_p field.
14476 (riscv_attribute_table, riscv_output_return),
14477 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
14478 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
14479 (riscv_compute_frame_info): Only compute frame->mask if not a naked
14481 (riscv_expand_prologue): Add early return for naked function.
14482 (riscv_expand_epilogue): Likewise.
14483 (riscv_function_ok_for_sibcall): Return false for naked function.
14484 (riscv_set_current_function): New.
14485 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
14486 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
14487 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
14488 * doc/extend.texi (RISC-V Function Attributes): New.
14490 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
14492 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
14493 check for 128-bit long double before checking TCmode.
14494 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
14495 128-bit long doubles before checking TFmode or TCmode.
14496 (FLOAT128_IBM_P): Likewise.
14498 2018-01-10 Martin Sebor <msebor@redhat.com>
14500 PR tree-optimization/83671
14501 * builtins.c (c_strlen): Unconditionally return zero for the empty
14503 Use -Warray-bounds for warnings.
14504 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
14505 for non-constant array indices with COMPONENT_REF, arrays of
14506 arrays, and pointers to arrays.
14507 (gimple_fold_builtin_strlen): Determine and set length range for
14508 non-constant character arrays.
14510 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
14512 PR middle-end/81897
14513 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
14516 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
14518 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
14520 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
14523 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
14524 VECTOR_MEM_ALTIVEC_OR_VSX_P.
14525 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
14526 indexed_or_indirect_operand predicate.
14527 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
14528 (*vsx_le_perm_load_v8hi): Likewise.
14529 (*vsx_le_perm_load_v16qi): Likewise.
14530 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
14531 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
14532 (*vsx_le_perm_store_v8hi): Likewise.
14533 (*vsx_le_perm_store_v16qi): Likewise.
14534 (eight unnamed splitters): Likewise.
14536 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
14538 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
14539 * config/rs6000/emmintrin.h: Likewise.
14540 * config/rs6000/mmintrin.h: Likewise.
14541 * config/rs6000/xmmintrin.h: Likewise.
14543 2018-01-10 David Malcolm <dmalcolm@redhat.com>
14546 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
14548 * tree.c (tree_nop_conversion): Return true for location wrapper
14550 (maybe_wrap_with_location): New function.
14551 (selftest::check_strip_nops): New function.
14552 (selftest::test_location_wrappers): New function.
14553 (selftest::tree_c_tests): Call it.
14554 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
14555 (maybe_wrap_with_location): New decl.
14556 (EXPR_LOCATION_WRAPPER_P): New macro.
14557 (location_wrapper_p): New inline function.
14558 (tree_strip_any_location_wrapper): New inline function.
14560 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
14563 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
14564 stack_realign_offset for the largest alignment of stack slot
14566 (ix86_find_max_used_stack_alignment): New function.
14567 (ix86_finalize_stack_frame_flags): Use it. Set
14568 max_used_stack_alignment if we don't realign stack.
14569 * config/i386/i386.h (machine_function): Add
14570 max_used_stack_alignment.
14572 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
14574 * config/arm/arm.opt (-mbranch-cost): New option.
14575 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
14578 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
14581 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
14582 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
14584 2018-01-10 Richard Biener <rguenther@suse.de>
14587 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
14588 early out so it also covers the case where we have a non-NULL
14591 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
14593 PR tree-optimization/83753
14594 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
14595 for non-strided grouped accesses if the number of elements is 1.
14597 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14600 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
14601 * i386.h (TARGET_USE_GATHER): Define.
14602 * x86-tune.def (X86_TUNE_USE_GATHER): New.
14604 2018-01-10 Martin Liska <mliska@suse.cz>
14607 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
14608 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
14610 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
14611 CLEANUP_NO_PARTITIONING is not set.
14613 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
14615 * doc/rtl.texi: Remove documentation of (const ...) wrappers
14616 for vectors, as a partial revert of r254296.
14617 * rtl.h (const_vec_p): Delete.
14618 (const_vec_duplicate_p): Don't test for vector CONSTs.
14619 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
14620 * expmed.c (make_tree): Likewise.
14623 * common.md (E, F): Use CONSTANT_P instead of checking for
14625 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
14626 checking for CONST_VECTOR.
14628 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14630 PR middle-end/83575
14631 * predict.c (force_edge_cold): Handle in more sane way edges
14632 with no prediction.
14634 2018-01-09 Carl Love <cel@us.ibm.com>
14636 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
14638 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
14639 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
14640 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
14641 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
14642 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
14643 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
14644 * config/rs6000/rs6000-protos.h: Add extern defition for
14645 rs6000_generate_float2_double_code.
14646 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
14648 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
14649 (float2_v2df): Add define_expand.
14651 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
14654 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
14655 op_mode in the force_to_mode call.
14657 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
14659 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
14660 instead of checking each element individually.
14661 (aarch64_evpc_uzp): Likewise.
14662 (aarch64_evpc_zip): Likewise.
14663 (aarch64_evpc_ext): Likewise.
14664 (aarch64_evpc_rev): Likewise.
14665 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
14666 instead of checking each element individually. Return true without
14668 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
14669 whether all selected elements come from the same input, instead of
14670 checking each element individually. Remove calls to gen_rtx_REG,
14671 start_sequence and end_sequence and instead assert that no rtl is
14674 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
14676 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
14677 order of HIGH and CONST checks.
14679 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
14681 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
14682 if the destination isn't an SSA_NAME.
14684 2018-01-09 Richard Biener <rguenther@suse.de>
14686 PR tree-optimization/83668
14687 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
14689 (canonicalize_loop_form): ... here, renamed from ...
14690 (canonicalize_loop_closed_ssa_form): ... this and amended to
14691 swap successor edges for loop exit blocks to make us use
14692 the RPO order we need for initial schedule generation.
14694 2018-01-09 Joseph Myers <joseph@codesourcery.com>
14696 PR tree-optimization/64811
14697 * match.pd: When optimizing comparisons with Inf, avoid
14698 introducing or losing exceptions from comparisons with NaN.
14700 2018-01-09 Martin Liska <mliska@suse.cz>
14703 * asan.c (shadow_mem_size): Add gcc_assert.
14705 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
14707 Don't save registers in main().
14710 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
14711 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
14712 * config/avr/avr.c (avr_set_current_function): Don't error if
14713 naked, OS_task or OS_main are specified at the same time.
14714 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
14716 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
14718 * common/config/avr/avr-common.c (avr_option_optimization_table):
14719 Switch on -mmain-is-OS_task for optimizing compilations.
14721 2018-01-09 Richard Biener <rguenther@suse.de>
14723 PR tree-optimization/83572
14724 * graphite.c: Include cfganal.h.
14725 (graphite_transform_loops): Connect infinite loops to exit
14726 and remove fake edges at the end.
14728 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14730 * ipa-inline.c (edge_badness): Revert accidental checkin.
14732 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14735 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
14736 symbols; not inline clones.
14738 2018-01-09 Jakub Jelinek <jakub@redhat.com>
14741 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
14742 hard registers. Formatting fixes.
14744 PR preprocessor/83722
14745 * gcc.c (try_generate_repro): Pass
14746 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
14747 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
14750 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
14751 Kito Cheng <kito.cheng@gmail.com>
14753 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
14754 (riscv_leaf_function_p): Delete.
14755 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
14757 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14759 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
14761 (do_ifelse): New function.
14762 (do_isel): New function.
14763 (do_sub3): New function.
14764 (do_add3): New function.
14765 (do_load_mask_compare): New function.
14766 (do_overlap_load_compare): New function.
14767 (expand_compare_loop): New function.
14768 (expand_block_compare): Call expand_compare_loop() when appropriate.
14769 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
14770 option description.
14771 (-mblock-compare-inline-loop-limit): New option.
14773 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14776 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
14777 Reverse order of second and third operands in first alternative.
14778 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
14779 of first and second elements in UNSPEC_VPERMR vector.
14780 (altivec_expand_vec_perm_le): Likewise.
14782 2018-01-08 Jeff Law <law@redhat.com>
14784 PR rtl-optimizatin/81308
14785 * tree-switch-conversion.c (cfg_altered): New file scoped static.
14786 (process_switch): If group_case_labels makes a change, then set
14788 (pass_convert_switch::execute): If a switch is converted, then
14789 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
14791 PR rtl-optimization/81308
14792 * recog.c (split_all_insns): Conditionally cleanup the CFG after
14795 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
14797 PR target/83663 - Revert r255946
14798 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
14799 generation for cases where splatting a value is not useful.
14800 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
14801 across a vec_duplicate and a paradoxical subreg forming a vector
14802 mode to a vec_concat.
14804 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14806 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
14807 -march=armv8.3-a variants.
14808 * config/arm/t-multilib: Likewise.
14809 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
14811 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14813 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
14815 (cceq_ior_compare_complement): Give it a name so I can use it, and
14816 change boolean_or_operator predicate to boolean_operator so it can
14817 be used to generate a crand.
14818 (eqne): New code iterator.
14819 (bd/bd_neg): New code_attrs.
14820 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
14821 a single define_insn.
14822 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
14823 decrement (bdnzt/bdnzf/bdzt/bdzf).
14824 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
14825 with the new names of the branch decrement patterns, and added the
14826 names of the branch decrement conditional patterns.
14828 2018-01-08 Richard Biener <rguenther@suse.de>
14830 PR tree-optimization/83563
14831 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
14834 2018-01-08 Richard Biener <rguenther@suse.de>
14836 PR middle-end/83713
14837 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
14839 2018-01-08 Richard Biener <rguenther@suse.de>
14841 PR tree-optimization/83685
14842 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
14843 references to abnormals.
14845 2018-01-08 Richard Biener <rguenther@suse.de>
14848 * dwarf2out.c (output_indirect_strings): Handle empty
14849 skeleton_debug_str_hash.
14850 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
14852 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
14854 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
14855 (emit_store_direct): Likewise.
14856 (arc_trampoline_adjust_address): Likewise.
14857 (arc_asm_trampoline_template): New function.
14858 (arc_initialize_trampoline): Use asm_trampoline_template.
14859 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
14860 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
14861 * config/arc/arc.md (flush_icache): Delete pattern.
14863 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
14865 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
14866 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
14869 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
14872 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
14873 by not USED_FOR_TARGET.
14874 (make_pass_resolve_sw_modes): Likewise.
14876 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
14878 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
14881 2018-01-08 Richard Biener <rguenther@suse.de>
14883 PR middle-end/83580
14884 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
14886 2018-01-08 Richard Biener <rguenther@suse.de>
14888 PR middle-end/83517
14889 * match.pd ((t * 2) / 2) -> t): Add missing :c.
14891 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
14893 PR middle-end/81897
14894 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
14895 basic blocks with a small number of successors.
14896 (convert_control_dep_chain_into_preds): Improve handling of
14898 (dump_predicates): Split apart into...
14899 (dump_pred_chain): ...here...
14900 (dump_pred_info): ...and here.
14901 (can_one_predicate_be_invalidated_p): Add debugging printfs.
14902 (can_chain_union_be_invalidated_p): Improve check for invalidation
14904 (uninit_uses_cannot_happen): Avoid unnecessary if
14905 convert_control_dep_chain_into_preds yielded nothing.
14907 2018-01-06 Martin Sebor <msebor@redhat.com>
14909 PR tree-optimization/83640
14910 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
14911 subtracting negative offset from size.
14912 (builtin_access::overlap): Adjust offset bounds of the access to fall
14913 within the size of the object if possible.
14915 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
14917 PR rtl-optimization/83699
14918 * expmed.c (extract_bit_field_1): Restrict the vector usage of
14919 extract_bit_field_as_subreg to cases in which the extracted
14920 value is also a vector.
14922 * lra-constraints.c (process_alt_operands): Test for the equivalence
14923 substitutions when detecting a possible reload cycle.
14925 2018-01-06 Jakub Jelinek <jakub@redhat.com>
14928 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
14929 by default if flag_selective_schedling{,2}. Formatting fixes.
14931 PR rtl-optimization/83682
14932 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
14933 if it has non-VECTOR_MODE element mode.
14934 (vec_duplicate_p): Likewise.
14936 PR middle-end/83694
14937 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
14938 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
14940 2018-01-05 Jakub Jelinek <jakub@redhat.com>
14943 * config/i386/i386-builtin.def
14944 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
14945 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
14946 Require also OPTION_MASK_ISA_AVX512F in addition to
14947 OPTION_MASK_ISA_GFNI.
14948 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
14949 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
14950 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
14951 to OPTION_MASK_ISA_GFNI.
14952 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
14953 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
14954 OPTION_MASK_ISA_AVX512BW.
14955 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
14956 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
14957 addition to OPTION_MASK_ISA_GFNI.
14958 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
14959 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
14960 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
14961 to OPTION_MASK_ISA_GFNI.
14962 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
14963 a requirement for all ISAs rather than any of them with a few
14965 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
14967 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
14968 bitmasks to be enabled with 3 exceptions, instead of requiring any
14969 enabled ISA with lots of exceptions.
14970 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
14971 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
14972 Change avx512bw in isa attribute to avx512f.
14973 * config/i386/sgxintrin.h: Add license boilerplate.
14974 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
14975 to __AVX512F__ and __AVX512VL to __AVX512VL__.
14976 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
14977 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
14979 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
14980 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
14981 temporarily sse2 rather than sse if not enabled already.
14984 * config/i386/sse.md (VI248_VLBW): Rename to ...
14985 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
14986 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
14987 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
14988 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
14989 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
14990 mode iterator instead of VI248_VLBW.
14992 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
14994 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
14995 (record_modified): Skip clobbers; add debug output.
14996 (param_change_prob): Use sreal frequencies.
14998 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15000 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
15001 punt for user-aligned variables.
15003 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15005 * tree-chrec.c (chrec_contains_symbols): Return true for
15008 2018-01-05 Sudakshina Das <sudi.das@arm.com>
15011 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
15012 of (x|y) == x for BICS pattern.
15014 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15016 PR tree-optimization/83605
15017 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
15018 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
15021 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
15023 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
15024 * config/epiphany/rtems.h: New file.
15026 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15027 Uros Bizjak <ubizjak@gmail.com>
15030 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
15031 QIreg_operand instead of register_operand predicate.
15032 * config/i386/i386.c (ix86_rop_should_change_byte_p,
15033 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
15034 comments instead of -fmitigate[-_]rop.
15036 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15039 * cgraphunit.c (symbol_table::compile): Switch to text_section
15040 before calling assembly_start debug hook.
15041 * run-rtl-passes.c (run_rtl_passes): Likewise.
15044 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15046 * tree-vrp.c (extract_range_from_binary_expr_1): Check
15047 range_int_cst_p rather than !symbolic_range_p before calling
15048 extract_range_from_multiplicative_op_1.
15050 2018-01-04 Jeff Law <law@redhat.com>
15052 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
15053 redundant test in assertion.
15055 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15057 * doc/rtl.texi: Document machine_mode wrapper classes.
15059 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15061 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
15062 using tree_to_uhwi.
15064 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15066 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
15067 the VEC_PERM_EXPR fold to fail.
15069 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15072 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
15073 to switched_sections.
15075 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15078 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
15079 test for d.testing.
15081 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
15084 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
15085 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
15087 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15090 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
15091 is BLKmode and bitpos not zero or mode change is needed.
15093 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15096 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
15099 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
15102 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
15103 instead of MULT rtx. Update all corresponding splitters.
15104 (*saddl_se): Ditto.
15105 (*ssub<modesuffix>): Ditto.
15106 (*ssubl_se): Ditto.
15107 (*cmp_sadd_di): Update split patterns.
15108 (*cmp_sadd_si): Ditto.
15109 (*cmp_sadd_sidi): Ditto.
15110 (*cmp_ssub_di): Ditto.
15111 (*cmp_ssub_si): Ditto.
15112 (*cmp_ssub_sidi): Ditto.
15113 * config/alpha/predicates.md (const23_operand): New predicate.
15114 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
15115 Look for ASHIFT, not MULT inner operand.
15116 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
15118 2018-01-04 Martin Liska <mliska@suse.cz>
15120 PR gcov-profile/83669
15121 * gcov.c (output_intermediate_file): Add version to intermediate
15123 * doc/gcov.texi: Document new field 'version' in intermediate
15124 file format. Fix location of '-k' option of gcov command.
15126 2018-01-04 Martin Liska <mliska@suse.cz>
15129 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
15131 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15133 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
15135 2018-01-03 Martin Sebor <msebor@redhat.com>
15137 PR tree-optimization/83655
15138 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
15139 checking calls with invalid arguments.
15141 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15143 * tree-vect-stmts.c (vect_get_store_rhs): New function.
15144 (vectorizable_mask_load_store): Delete.
15145 (vectorizable_call): Return false for masked loads and stores.
15146 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
15147 instead of gimple_assign_rhs1.
15148 (vectorizable_load): Handle IFN_MASK_LOAD.
15149 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
15151 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15153 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
15155 (vectorizable_mask_load_store): ...here.
15156 (vectorizable_load): ...and here.
15158 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15160 * tree-vect-stmts.c (vect_build_all_ones_mask)
15161 (vect_build_zero_merge_argument): New functions, split out from...
15162 (vectorizable_load): ...here.
15164 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15166 * tree-vect-stmts.c (vect_check_store_rhs): New function,
15168 (vectorizable_mask_load_store): ...here.
15169 (vectorizable_store): ...and here.
15171 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15173 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
15175 (vectorizable_mask_load_store): ...here.
15177 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15179 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
15180 (vect_model_store_cost): Take a vec_load_store_type instead of a
15182 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
15183 (vect_model_store_cost): Take a vec_load_store_type instead of a
15185 (vectorizable_mask_load_store): Update accordingly.
15186 (vectorizable_store): Likewise.
15187 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
15189 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15191 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
15192 IFN_MASK_LOAD calls here rather than...
15193 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
15195 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15196 Alan Hayward <alan.hayward@arm.com>
15197 David Sherwood <david.sherwood@arm.com>
15199 * expmed.c (extract_bit_field_1): For vector extracts,
15200 fall back to extract_bit_field_as_subreg if vec_extract
15203 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15204 Alan Hayward <alan.hayward@arm.com>
15205 David Sherwood <david.sherwood@arm.com>
15207 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
15208 they are variable or constant sized.
15209 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
15210 slots for constant-sized data.
15212 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15213 Alan Hayward <alan.hayward@arm.com>
15214 David Sherwood <david.sherwood@arm.com>
15216 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
15217 handling COND_EXPRs with boolean comparisons, try to find a better
15218 basis for the mask type than the boolean itself.
15220 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15222 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
15223 is calculated and how it can be overridden.
15224 * genmodes.c (max_bitsize_mode_any_mode): New variable.
15225 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
15227 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
15230 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15231 Alan Hayward <alan.hayward@arm.com>
15232 David Sherwood <david.sherwood@arm.com>
15234 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
15235 Remove the mode argument.
15236 (aarch64_simd_valid_immediate): Remove the mode and inverse
15238 * config/aarch64/iterators.md (bitsize): New iterator.
15239 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
15240 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
15241 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
15242 aarch64_simd_valid_immediate.
15243 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
15244 (aarch64_reg_or_bic_imm): Likewise.
15245 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
15246 with an insn_type enum and msl with a modifier_type enum.
15247 Replace element_width with a scalar_mode. Change the shift
15248 to unsigned int. Add constructors for scalar_float_mode and
15249 scalar_int_mode elements.
15250 (aarch64_vect_float_const_representable_p): Delete.
15251 (aarch64_can_const_movi_rtx_p)
15252 (aarch64_simd_scalar_immediate_valid_for_move)
15253 (aarch64_simd_make_constant): Update call to
15254 aarch64_simd_valid_immediate.
15255 (aarch64_advsimd_valid_immediate_hs): New function.
15256 (aarch64_advsimd_valid_immediate): Likewise.
15257 (aarch64_simd_valid_immediate): Remove mode and inverse
15258 arguments. Rewrite to use the above. Use const_vec_duplicate_p
15259 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
15260 and aarch64_float_const_representable_p on the result.
15261 (aarch64_output_simd_mov_immediate): Remove mode argument.
15262 Update call to aarch64_simd_valid_immediate and use of
15263 simd_immediate_info.
15264 (aarch64_output_scalar_simd_mov_immediate): Update call
15267 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15268 Alan Hayward <alan.hayward@arm.com>
15269 David Sherwood <david.sherwood@arm.com>
15271 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
15272 (mode_nunits): Likewise CONST_MODE_NUNITS.
15273 * machmode.def (ADJUST_NUNITS): Document.
15274 * genmodes.c (mode_data::need_nunits_adj): New field.
15275 (blank_mode): Update accordingly.
15276 (adj_nunits): New variable.
15277 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
15279 (emit_mode_size_inline): Set need_bytesize_adj for all modes
15280 listed in adj_nunits.
15281 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
15282 listed in adj_nunits. Don't emit case statements for such modes.
15283 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
15284 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
15285 nothing if adj_nunits is nonnull.
15286 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
15287 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
15288 (emit_mode_fbit): Update use of print_maybe_const_decl.
15289 (emit_move_size): Likewise. Treat the array as non-const
15291 (emit_mode_adjustments): Handle adj_nunits.
15293 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15295 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
15296 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
15297 (VECTOR_MODES): Use it.
15298 (make_vector_modes): Take the prefix as an argument.
15300 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15301 Alan Hayward <alan.hayward@arm.com>
15302 David Sherwood <david.sherwood@arm.com>
15304 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
15305 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
15306 for MODE_VECTOR_BOOL.
15307 * machmode.def (VECTOR_BOOL_MODE): Document.
15308 * genmodes.c (VECTOR_BOOL_MODE): New macro.
15309 (make_vector_bool_mode): New function.
15310 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
15312 * lto-streamer-in.c (lto_input_mode_table): Likewise.
15313 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
15315 * stor-layout.c (int_mode_for_mode): Likewise.
15316 * tree.c (build_vector_type_for_mode): Likewise.
15317 * varasm.c (output_constant_pool_2): Likewise.
15318 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
15319 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
15320 for MODE_VECTOR_BOOL.
15321 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
15322 of mode class checks.
15323 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
15324 instead of a list of mode class checks.
15325 (expand_vector_scalar_condition): Likewise.
15326 (type_for_widest_vector_mode): Handle BImode as an inner mode.
15328 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15329 Alan Hayward <alan.hayward@arm.com>
15330 David Sherwood <david.sherwood@arm.com>
15332 * machmode.h (mode_size): Change from unsigned short to
15334 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
15335 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
15336 or if measurement_type is not polynomial.
15337 (fixed_size_mode::includes_p): Check for constant-sized modes.
15338 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
15339 return a poly_uint16 rather than an unsigned short.
15340 (emit_mode_size): Change the type of mode_size from unsigned short
15341 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
15342 (emit_mode_adjustments): Cope with polynomial vector sizes.
15343 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15345 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15347 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
15348 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
15349 * caller-save.c (setup_save_areas): Likewise.
15350 (replace_reg_with_saved_mem): Likewise.
15351 * calls.c (emit_library_call_value_1): Likewise.
15352 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
15353 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
15354 (gen_lowpart_for_combine): Likewise.
15355 * convert.c (convert_to_integer_1): Likewise.
15356 * cse.c (equiv_constant, cse_insn): Likewise.
15357 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
15358 (cselib_subst_to_values): Likewise.
15359 * dce.c (word_dce_process_block): Likewise.
15360 * df-problems.c (df_word_lr_mark_ref): Likewise.
15361 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
15362 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
15363 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
15364 (rtl_for_decl_location): Likewise.
15365 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
15366 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
15367 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
15368 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
15369 (expand_expr_real_1): Likewise.
15370 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
15371 (pad_below): Likewise.
15372 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
15373 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
15374 * ira.c (get_subreg_tracking_sizes): Likewise.
15375 * ira-build.c (ira_create_allocno_objects): Likewise.
15376 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
15377 (ira_sort_regnos_for_alter_reg): Likewise.
15378 * ira-costs.c (record_operand_costs): Likewise.
15379 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
15380 (resolve_simple_move): Likewise.
15381 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
15382 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
15383 (lra_constraints): Likewise.
15384 (CONST_POOL_OK_P): Reject variable-sized modes.
15385 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
15386 (add_pseudo_to_slot, lra_spill): Likewise.
15387 * omp-low.c (omp_clause_aligned_alignment): Likewise.
15388 * optabs-query.c (get_best_extraction_insn): Likewise.
15389 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
15390 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
15391 (expand_mult_highpart, valid_multiword_target_p): Likewise.
15392 * recog.c (offsettable_address_addr_space_p): Likewise.
15393 * regcprop.c (maybe_mode_change): Likewise.
15394 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
15395 * regrename.c (build_def_use): Likewise.
15396 * regstat.c (dump_reg_info): Likewise.
15397 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
15398 (find_reloads, find_reloads_subreg_address): Likewise.
15399 * reload1.c (eliminate_regs_1): Likewise.
15400 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
15401 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
15402 (simplify_binary_operation_1, simplify_subreg): Likewise.
15403 * targhooks.c (default_function_arg_padding): Likewise.
15404 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
15405 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
15406 (verify_gimple_assign_ternary): Likewise.
15407 * tree-inline.c (estimate_move_cost): Likewise.
15408 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15409 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
15410 (get_address_cost_ainc): Likewise.
15411 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
15412 (vect_supportable_dr_alignment): Likewise.
15413 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15414 (vectorizable_reduction): Likewise.
15415 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
15416 (vectorizable_operation, vectorizable_load): Likewise.
15417 * tree.c (build_same_sized_truth_vector_type): Likewise.
15418 * valtrack.c (cleanup_auto_inc_dec): Likewise.
15419 * var-tracking.c (emit_note_insn_var_location): Likewise.
15420 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
15421 (ADDR_VEC_ALIGN): Likewise.
15423 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15424 Alan Hayward <alan.hayward@arm.com>
15425 David Sherwood <david.sherwood@arm.com>
15427 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
15429 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
15430 or if measurement_type is polynomial.
15431 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
15432 * combine.c (make_extraction): Likewise.
15433 * dse.c (find_shift_sequence): Likewise.
15434 * dwarf2out.c (mem_loc_descriptor): Likewise.
15435 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
15436 (extract_bit_field, extract_low_bits): Likewise.
15437 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
15438 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
15439 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
15440 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
15441 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
15442 * reload.c (find_reloads): Likewise.
15443 * reload1.c (alter_reg): Likewise.
15444 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
15445 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
15446 * tree-if-conv.c (predicate_mem_writes): Likewise.
15447 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
15448 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
15449 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
15450 * valtrack.c (dead_debug_insert_temp): Likewise.
15451 * varasm.c (mergeable_constant_section): Likewise.
15452 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
15454 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15455 Alan Hayward <alan.hayward@arm.com>
15456 David Sherwood <david.sherwood@arm.com>
15458 * expr.c (expand_assignment): Cope with polynomial mode sizes
15459 when assigning to a CONCAT.
15461 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15462 Alan Hayward <alan.hayward@arm.com>
15463 David Sherwood <david.sherwood@arm.com>
15465 * machmode.h (mode_precision): Change from unsigned short to
15467 (mode_to_precision): Return a poly_uint16 rather than an unsigned
15469 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
15470 or if measurement_type is not polynomial.
15471 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
15472 in which the mode is already known to be a scalar_int_mode.
15473 * genmodes.c (emit_mode_precision): Change the type of mode_precision
15474 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
15476 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15477 for GET_MODE_PRECISION.
15478 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15479 for GET_MODE_PRECISION.
15480 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
15482 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
15483 (expand_field_assignment, make_extraction): Likewise.
15484 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
15485 (get_last_value): Likewise.
15486 * convert.c (convert_to_integer_1): Likewise.
15487 * cse.c (cse_insn): Likewise.
15488 * expr.c (expand_expr_real_1): Likewise.
15489 * lra-constraints.c (simplify_operand_subreg): Likewise.
15490 * optabs-query.c (can_atomic_load_p): Likewise.
15491 * optabs.c (expand_atomic_load): Likewise.
15492 (expand_atomic_store): Likewise.
15493 * ree.c (combine_reaching_defs): Likewise.
15494 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
15495 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
15496 * tree.h (type_has_mode_precision_p): Likewise.
15497 * ubsan.c (instrument_si_overflow): Likewise.
15499 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15500 Alan Hayward <alan.hayward@arm.com>
15501 David Sherwood <david.sherwood@arm.com>
15503 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
15504 polynomial numbers of units.
15505 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
15506 (valid_vector_subparts_p): New function.
15507 (build_vector_type): Remove temporary shim and take the number
15508 of units as a poly_uint64 rather than an int.
15509 (build_opaque_vector_type): Take the number of units as a
15510 poly_uint64 rather than an int.
15511 * tree.c (build_vector_from_ctor): Handle polynomial
15512 TYPE_VECTOR_SUBPARTS.
15513 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
15514 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
15515 (build_vector_from_val): If the number of units is variable,
15516 use build_vec_duplicate_cst for constant operands and
15517 VEC_DUPLICATE_EXPR otherwise.
15518 (make_vector_type): Remove temporary is_constant ().
15519 (build_vector_type, build_opaque_vector_type): Take the number of
15520 units as a poly_uint64 rather than an int.
15521 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
15523 * cfgexpand.c (expand_debug_expr): Likewise.
15524 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
15525 (store_constructor, expand_expr_real_1): Likewise.
15526 (const_scalar_mask_from_tree): Likewise.
15527 * fold-const-call.c (fold_const_reduction): Likewise.
15528 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
15529 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
15530 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
15531 (fold_relational_const): Likewise.
15532 (native_interpret_vector): Likewise. Change the size from an
15533 int to an unsigned int.
15534 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
15535 TYPE_VECTOR_SUBPARTS.
15536 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
15537 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
15538 duplicating a non-constant operand into a variable-length vector.
15539 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
15540 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
15541 * ipa-icf.c (sem_variable::equals): Likewise.
15542 * match.pd: Likewise.
15543 * omp-simd-clone.c (simd_clone_subparts): Likewise.
15544 * print-tree.c (print_node): Likewise.
15545 * stor-layout.c (layout_type): Likewise.
15546 * targhooks.c (default_builtin_vectorization_cost): Likewise.
15547 * tree-cfg.c (verify_gimple_comparison): Likewise.
15548 (verify_gimple_assign_binary): Likewise.
15549 (verify_gimple_assign_ternary): Likewise.
15550 (verify_gimple_assign_single): Likewise.
15551 * tree-pretty-print.c (dump_generic_node): Likewise.
15552 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15553 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
15554 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
15555 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
15556 (vect_shift_permute_load_chain): Likewise.
15557 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
15558 (expand_vector_condition, optimize_vector_constructor): Likewise.
15559 (lower_vec_perm, get_compute_type): Likewise.
15560 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15561 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
15562 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
15563 (vect_recog_mask_conversion_pattern): Likewise.
15564 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
15565 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
15566 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15567 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
15568 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
15569 (vectorizable_shift, vectorizable_operation, vectorizable_store)
15570 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
15571 (supportable_widening_operation): Likewise.
15572 (supportable_narrowing_operation): Likewise.
15573 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
15575 * varasm.c (output_constant): Likewise.
15577 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15578 Alan Hayward <alan.hayward@arm.com>
15579 David Sherwood <david.sherwood@arm.com>
15581 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
15582 so that both the length == 3 and length != 3 cases set up their
15583 own permute vectors. Add comments explaining why we know the
15584 number of elements is constant.
15585 (vect_permute_load_chain): Likewise.
15587 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15588 Alan Hayward <alan.hayward@arm.com>
15589 David Sherwood <david.sherwood@arm.com>
15591 * machmode.h (mode_nunits): Change from unsigned char to
15593 (ONLY_FIXED_SIZE_MODES): New macro.
15594 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
15595 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
15596 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
15598 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
15599 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
15600 or if measurement_type is not polynomial.
15601 * genmodes.c (ZERO_COEFFS): New macro.
15602 (emit_mode_nunits_inline): Make mode_nunits_inline return a
15604 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
15605 Use ZERO_COEFFS when emitting initializers.
15606 * data-streamer.h (bp_pack_poly_value): New function.
15607 (bp_unpack_poly_value): Likewise.
15608 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15609 for GET_MODE_NUNITS.
15610 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15611 for GET_MODE_NUNITS.
15612 * tree.c (make_vector_type): Remove temporary shim and make
15613 the real function take the number of units as a poly_uint64
15614 rather than an int.
15615 (build_vector_type_for_mode): Handle polynomial nunits.
15616 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
15617 * emit-rtl.c (const_vec_series_p_1): Likewise.
15618 (gen_rtx_CONST_VECTOR): Likewise.
15619 * fold-const.c (test_vec_duplicate_folding): Likewise.
15620 * genrecog.c (validate_pattern): Likewise.
15621 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
15622 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
15623 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
15624 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
15625 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
15626 * rtlanal.c (subreg_get_info): Likewise.
15627 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15628 (vect_grouped_load_supported): Likewise.
15629 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
15630 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
15631 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
15632 (simplify_const_unary_operation, simplify_binary_operation_1)
15633 (simplify_const_binary_operation, simplify_ternary_operation)
15634 (test_vector_ops_duplicate, test_vector_ops): Likewise.
15635 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
15636 instead of CONST_VECTOR_NUNITS.
15637 * varasm.c (output_constant_pool_2): Likewise.
15638 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
15639 explicit-encoded elements in the XVEC for variable-length vectors.
15641 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15643 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
15645 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15646 Alan Hayward <alan.hayward@arm.com>
15647 David Sherwood <david.sherwood@arm.com>
15649 * coretypes.h (fixed_size_mode): Declare.
15650 (fixed_size_mode_pod): New typedef.
15651 * builtins.h (target_builtins::x_apply_args_mode)
15652 (target_builtins::x_apply_result_mode): Change type to
15653 fixed_size_mode_pod.
15654 * builtins.c (apply_args_size, apply_result_size, result_vector)
15655 (expand_builtin_apply_args_1, expand_builtin_apply)
15656 (expand_builtin_return): Update accordingly.
15658 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15660 * cse.c (hash_rtx_cb): Hash only the encoded elements.
15661 * cselib.c (cselib_hash_rtx): Likewise.
15662 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
15663 CONST_VECTOR encoding.
15665 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15666 Jeff Law <law@redhat.com>
15669 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
15670 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
15671 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
15672 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
15675 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
15676 explicitly probe *sp in a noreturn function if there were any callee
15677 register saves or frame pointer is needed.
15679 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15682 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
15683 BLKmode for ternary, binary or unary expressions.
15686 * var-tracking.c (delete_vta_debug_insn): New inline function.
15687 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
15688 insns from get_insns () to NULL instead of each bb separately.
15689 Use delete_vta_debug_insn. No longer static.
15690 (vt_debug_insns_local, variable_tracking_main_1): Adjust
15691 delete_vta_debug_insns callers.
15692 * rtl.h (delete_vta_debug_insns): Declare.
15693 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
15694 instead of variable_tracking_main.
15696 2018-01-03 Martin Sebor <msebor@redhat.com>
15698 PR tree-optimization/83603
15699 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
15700 arguments past the endof the argument list in functions declared
15701 without a prototype.
15702 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
15703 Avoid checking when arguments are null.
15705 2018-01-03 Martin Sebor <msebor@redhat.com>
15708 * doc/extend.texi (attribute const): Fix a typo.
15709 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
15710 issuing -Wsuggest-attribute for void functions.
15712 2018-01-03 Martin Sebor <msebor@redhat.com>
15714 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
15715 offset_int::from instead of wide_int::to_shwi.
15716 (maybe_diag_overlap): Remove assertion.
15717 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
15718 * gimple-ssa-sprintf.c (format_directive): Same.
15719 (parse_directive): Same.
15720 (sprintf_dom_walker::compute_format_length): Same.
15721 (try_substitute_return_value): Same.
15723 2018-01-03 Jeff Law <law@redhat.com>
15725 PR middle-end/83654
15726 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
15727 non-constant residual for zero at runtime and avoid probing in
15728 that case. Reorganize code for trailing problem to mirror handling
15731 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15733 PR tree-optimization/83501
15734 * tree-ssa-strlen.c (get_string_cst): New.
15735 (handle_char_store): Call get_string_cst.
15737 2018-01-03 Martin Liska <mliska@suse.cz>
15739 PR tree-optimization/83593
15740 * tree-ssa-strlen.c: Include tree-cfg.h.
15741 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
15742 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
15743 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
15745 (strlen_dom_walker::before_dom_children): Call
15746 gimple_purge_dead_eh_edges. Dump tranformation with details
15748 (strlen_dom_walker::before_dom_children): Update call by adding
15749 new argument cleanup_eh.
15750 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
15752 2018-01-03 Martin Liska <mliska@suse.cz>
15755 * cif-code.def (VARIADIC_THUNK): New enum value.
15756 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
15759 2018-01-03 Jan Beulich <jbeulich@suse.com>
15761 * sse.md (mov<mode>_internal): Tighten condition for when to use
15762 vmovdqu<ssescalarsize> for TI and OI modes.
15764 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15766 Update copyright years.
15768 2018-01-03 Martin Liska <mliska@suse.cz>
15771 * ipa-visibility.c (function_and_variable_visibility): Skip
15772 functions with noipa attribure.
15774 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15776 * gcc.c (process_command): Update copyright notice dates.
15777 * gcov-dump.c (print_version): Ditto.
15778 * gcov.c (print_version): Ditto.
15779 * gcov-tool.c (print_version): Ditto.
15780 * gengtype.c (create_file): Ditto.
15781 * doc/cpp.texi: Bump @copying's copyright year.
15782 * doc/cppinternals.texi: Ditto.
15783 * doc/gcc.texi: Ditto.
15784 * doc/gccint.texi: Ditto.
15785 * doc/gcov.texi: Ditto.
15786 * doc/install.texi: Ditto.
15787 * doc/invoke.texi: Ditto.
15789 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15791 * vector-builder.h (vector_builder::m_full_nelts): Change from
15792 unsigned int to poly_uint64.
15793 (vector_builder::full_nelts): Update prototype accordingly.
15794 (vector_builder::new_vector): Likewise.
15795 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
15796 (vector_builder::operator ==): Likewise.
15797 (vector_builder::finalize): Likewise.
15798 * int-vector-builder.h (int_vector_builder::int_vector_builder):
15799 Take the number of elements as a poly_uint64 rather than an
15801 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
15802 from unsigned int to poly_uint64.
15803 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
15804 (vec_perm_indices::new_vector): Likewise.
15805 (vec_perm_indices::length): Likewise.
15806 (vec_perm_indices::nelts_per_input): Likewise.
15807 (vec_perm_indices::input_nelts): Likewise.
15808 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
15809 number of elements per input as a poly_uint64 rather than an
15810 unsigned int. Use the original encoding for variable-length
15811 vectors, rather than clamping each individual element.
15812 For the second and subsequent elements in each pattern,
15813 clamp the step and base before clamping their sum.
15814 (vec_perm_indices::series_p): Handle polynomial element counts.
15815 (vec_perm_indices::all_in_range_p): Likewise.
15816 (vec_perm_indices_to_tree): Likewise.
15817 (vec_perm_indices_to_rtx): Likewise.
15818 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
15819 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
15820 (tree_vector_builder::new_binary_operation): Handle polynomial
15821 element counts. Return false if we need to know the number
15822 of elements at compile time.
15823 * fold-const.c (fold_vec_perm): Punt if the number of elements
15824 isn't known at compile time.
15826 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15828 * vec-perm-indices.h (vec_perm_builder): Change element type
15829 from HOST_WIDE_INT to poly_int64.
15830 (vec_perm_indices::element_type): Update accordingly.
15831 (vec_perm_indices::clamp): Handle polynomial element_types.
15832 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
15833 (vec_perm_indices::all_in_range_p): Likewise.
15834 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
15836 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
15837 polynomial vec_perm_indices element types.
15838 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
15839 * fold-const.c (fold_vec_perm): Likewise.
15840 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
15841 * tree-vect-generic.c (lower_vec_perm): Likewise.
15842 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
15843 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
15844 element type to HOST_WIDE_INT.
15846 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15847 Alan Hayward <alan.hayward@arm.com>
15848 David Sherwood <david.sherwood@arm.com>
15850 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
15851 rather than an int. Use plus_constant.
15852 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
15853 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
15855 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15856 Alan Hayward <alan.hayward@arm.com>
15857 David Sherwood <david.sherwood@arm.com>
15859 * calls.c (emit_call_1, expand_call): Change struct_value_size from
15860 a HOST_WIDE_INT to a poly_int64.
15862 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15863 Alan Hayward <alan.hayward@arm.com>
15864 David Sherwood <david.sherwood@arm.com>
15866 * calls.c (load_register_parameters): Cope with polynomial
15867 mode sizes. Require a constant size for BLKmode parameters
15868 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
15869 forces a parameter to be padded at the lsb end in order to
15870 fill a complete number of words, require the parameter size
15871 to be ordered wrt UNITS_PER_WORD.
15873 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15874 Alan Hayward <alan.hayward@arm.com>
15875 David Sherwood <david.sherwood@arm.com>
15877 * reload1.c (spill_stack_slot_width): Change element type
15878 from unsigned int to poly_uint64_pod.
15879 (alter_reg): Treat mode sizes as polynomial.
15881 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15882 Alan Hayward <alan.hayward@arm.com>
15883 David Sherwood <david.sherwood@arm.com>
15885 * reload.c (complex_word_subreg_p): New function.
15886 (reload_inner_reg_of_subreg, push_reload): Use it.
15888 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15889 Alan Hayward <alan.hayward@arm.com>
15890 David Sherwood <david.sherwood@arm.com>
15892 * lra-constraints.c (process_alt_operands): Reject matched
15893 operands whose sizes aren't ordered.
15894 (match_reload): Refer to this check here.
15896 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15897 Alan Hayward <alan.hayward@arm.com>
15898 David Sherwood <david.sherwood@arm.com>
15900 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
15901 that the mode size is in the set {1, 2, 4, 8, 16}.
15903 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15904 Alan Hayward <alan.hayward@arm.com>
15905 David Sherwood <david.sherwood@arm.com>
15907 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
15908 Use plus_constant instead of gen_rtx_PLUS.
15910 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15911 Alan Hayward <alan.hayward@arm.com>
15912 David Sherwood <david.sherwood@arm.com>
15914 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
15915 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
15916 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
15917 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
15918 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
15919 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
15920 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
15921 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
15922 * config/i386/i386.c (ix86_push_rounding): ...this new function.
15923 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
15925 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
15926 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
15927 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
15928 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
15929 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
15930 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
15931 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
15932 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
15933 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
15934 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
15936 * expr.c (emit_move_resolve_push): Treat the input and result
15937 of PUSH_ROUNDING as a poly_int64.
15938 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
15939 (emit_push_insn): Likewise.
15940 * lra-eliminations.c (mark_not_eliminable): Likewise.
15941 * recog.c (push_operand): Likewise.
15942 * reload1.c (elimination_effects): Likewise.
15943 * rtlanal.c (nonzero_bits1): Likewise.
15944 * calls.c (store_one_arg): Likewise. Require the padding to be
15945 known at compile time.
15947 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15948 Alan Hayward <alan.hayward@arm.com>
15949 David Sherwood <david.sherwood@arm.com>
15951 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
15952 Use plus_constant instead of gen_rtx_PLUS.
15954 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15955 Alan Hayward <alan.hayward@arm.com>
15956 David Sherwood <david.sherwood@arm.com>
15958 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
15959 rather than an int.
15961 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15962 Alan Hayward <alan.hayward@arm.com>
15963 David Sherwood <david.sherwood@arm.com>
15965 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
15966 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
15967 via stack temporaries. Treat the mode size as polynomial too.
15969 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15970 Alan Hayward <alan.hayward@arm.com>
15971 David Sherwood <david.sherwood@arm.com>
15973 * expr.c (expand_expr_real_2): When handling conversions involving
15974 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
15975 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
15976 as a poly_uint64 too.
15978 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15979 Alan Hayward <alan.hayward@arm.com>
15980 David Sherwood <david.sherwood@arm.com>
15982 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
15984 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15985 Alan Hayward <alan.hayward@arm.com>
15986 David Sherwood <david.sherwood@arm.com>
15988 * combine.c (can_change_dest_mode): Handle polynomial
15989 REGMODE_NATURAL_SIZE.
15990 * expmed.c (store_bit_field_1): Likewise.
15991 * expr.c (store_constructor): Likewise.
15992 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
15993 and polynomial REGMODE_NATURAL_SIZE.
15994 (gen_lowpart_common): Likewise.
15995 * reginfo.c (record_subregs_of_mode): Likewise.
15996 * rtlanal.c (read_modify_subreg_p): Likewise.
15998 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15999 Alan Hayward <alan.hayward@arm.com>
16000 David Sherwood <david.sherwood@arm.com>
16002 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
16003 numbers of elements.
16005 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16006 Alan Hayward <alan.hayward@arm.com>
16007 David Sherwood <david.sherwood@arm.com>
16009 * match.pd: Cope with polynomial numbers of vector elements.
16011 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16012 Alan Hayward <alan.hayward@arm.com>
16013 David Sherwood <david.sherwood@arm.com>
16015 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
16016 in a POINTER_PLUS_EXPR.
16018 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16019 Alan Hayward <alan.hayward@arm.com>
16020 David Sherwood <david.sherwood@arm.com>
16022 * omp-simd-clone.c (simd_clone_subparts): New function.
16023 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
16024 (ipa_simd_modify_function_body): Likewise.
16026 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16027 Alan Hayward <alan.hayward@arm.com>
16028 David Sherwood <david.sherwood@arm.com>
16030 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
16031 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
16032 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
16033 (expand_vector_condition, vector_element): Likewise.
16034 (subparts_gt): New function.
16035 (get_compute_type): Use subparts_gt.
16036 (count_type_subparts): Delete.
16037 (expand_vector_operations_1): Use subparts_gt instead of
16038 count_type_subparts.
16040 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16041 Alan Hayward <alan.hayward@arm.com>
16042 David Sherwood <david.sherwood@arm.com>
16044 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
16045 (vect_compile_time_alias): ...this new function. Do the calculation
16046 on poly_ints rather than trees.
16047 (vect_prune_runtime_alias_test_list): Update call accordingly.
16049 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16050 Alan Hayward <alan.hayward@arm.com>
16051 David Sherwood <david.sherwood@arm.com>
16053 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
16055 (vect_schedule_slp_instance): Likewise.
16057 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16058 Alan Hayward <alan.hayward@arm.com>
16059 David Sherwood <david.sherwood@arm.com>
16061 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
16062 constant and extern definitions for variable-length vectors.
16063 (vect_get_constant_vectors): Note that the number of units
16064 is known to be constant.
16066 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16067 Alan Hayward <alan.hayward@arm.com>
16068 David Sherwood <david.sherwood@arm.com>
16070 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
16071 of units as polynomial. Choose between WIDE and NARROW based
16074 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16075 Alan Hayward <alan.hayward@arm.com>
16076 David Sherwood <david.sherwood@arm.com>
16078 * tree-vect-stmts.c (simd_clone_subparts): New function.
16079 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
16081 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16082 Alan Hayward <alan.hayward@arm.com>
16083 David Sherwood <david.sherwood@arm.com>
16085 * tree-vect-stmts.c (vectorizable_call): Treat the number of
16086 vectors as polynomial. Use build_index_vector for
16087 IFN_GOMP_SIMD_LANE.
16089 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16090 Alan Hayward <alan.hayward@arm.com>
16091 David Sherwood <david.sherwood@arm.com>
16093 * tree-vect-stmts.c (get_load_store_type): Treat the number of
16094 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
16095 for variable-length vectors.
16096 (vectorizable_mask_load_store): Treat the number of units as
16097 polynomial, asserting that it is constant if the condition has
16098 already been enforced.
16099 (vectorizable_store, vectorizable_load): Likewise.
16101 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16102 Alan Hayward <alan.hayward@arm.com>
16103 David Sherwood <david.sherwood@arm.com>
16105 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
16106 of units as polynomial. Punt if we can't tell at compile time
16107 which vector contains the final result.
16109 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16110 Alan Hayward <alan.hayward@arm.com>
16111 David Sherwood <david.sherwood@arm.com>
16113 * tree-vect-loop.c (vectorizable_induction): Treat the number
16114 of units as polynomial. Punt on SLP inductions. Use an integer
16115 VEC_SERIES_EXPR for variable-length integer reductions. Use a
16116 cast of such a series for variable-length floating-point
16119 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16120 Alan Hayward <alan.hayward@arm.com>
16121 David Sherwood <david.sherwood@arm.com>
16123 * tree.h (build_index_vector): Declare.
16124 * tree.c (build_index_vector): New function.
16125 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
16126 of units as polynomial, forcibly converting it to a constant if
16127 vectorizable_reduction has already enforced the condition.
16128 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
16129 to create a {1,2,3,...} vector.
16130 (vectorizable_reduction): Treat the number of units as polynomial.
16131 Choose vectype_in based on the largest scalar element size rather
16132 than the smallest number of units. Enforce the restrictions
16135 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16136 Alan Hayward <alan.hayward@arm.com>
16137 David Sherwood <david.sherwood@arm.com>
16139 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
16140 number of units as polynomial.
16142 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16143 Alan Hayward <alan.hayward@arm.com>
16144 David Sherwood <david.sherwood@arm.com>
16146 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
16147 * target.def (autovectorize_vector_sizes): Return the vector sizes
16148 by pointer, using vector_sizes rather than a bitmask.
16149 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
16150 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
16151 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
16153 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
16154 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
16155 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
16156 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
16157 * omp-general.c (omp_max_vf): Likewise.
16158 * omp-low.c (omp_clause_aligned_alignment): Likewise.
16159 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
16160 * tree-vect-loop.c (vect_analyze_loop): Likewise.
16161 * tree-vect-slp.c (vect_slp_bb): Likewise.
16162 * doc/tm.texi: Regenerate.
16163 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
16165 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
16166 the vector size as a poly_uint64 rather than an unsigned int.
16167 (current_vector_size): Change from an unsigned int to a poly_uint64.
16168 (get_vectype_for_scalar_type): Update accordingly.
16169 * tree.h (build_truth_vector_type): Take the size and number of
16170 units as a poly_uint64 rather than an unsigned int.
16171 (build_vector_type): Add a temporary overload that takes
16172 the number of units as a poly_uint64 rather than an unsigned int.
16173 * tree.c (make_vector_type): Likewise.
16174 (build_truth_vector_type): Take the number of units as a poly_uint64
16175 rather than an unsigned int.
16177 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16178 Alan Hayward <alan.hayward@arm.com>
16179 David Sherwood <david.sherwood@arm.com>
16181 * target.def (get_mask_mode): Take the number of units and length
16182 as poly_uint64s rather than unsigned ints.
16183 * targhooks.h (default_get_mask_mode): Update accordingly.
16184 * targhooks.c (default_get_mask_mode): Likewise.
16185 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
16186 * doc/tm.texi: Regenerate.
16188 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16189 Alan Hayward <alan.hayward@arm.com>
16190 David Sherwood <david.sherwood@arm.com>
16192 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
16193 * omp-general.c (omp_max_vf): Likewise.
16194 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
16195 (expand_omp_simd): Handle polynomial safelen.
16196 * omp-low.c (omplow_simd_context): Add a default constructor.
16197 (omplow_simd_context::max_vf): Change from int to poly_uint64.
16198 (lower_rec_simd_input_clauses): Update accordingly.
16199 (lower_rec_input_clauses): Likewise.
16201 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16202 Alan Hayward <alan.hayward@arm.com>
16203 David Sherwood <david.sherwood@arm.com>
16205 * tree-vectorizer.h (vect_nunits_for_cost): New function.
16206 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
16207 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
16208 (vect_analyze_slp_cost): Likewise.
16209 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
16210 (vect_model_load_cost): Likewise.
16212 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16213 Alan Hayward <alan.hayward@arm.com>
16214 David Sherwood <david.sherwood@arm.com>
16216 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
16217 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
16218 from an unsigned int * to a poly_uint64_pod *.
16219 (calculate_unrolling_factor): New function.
16220 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
16222 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16223 Alan Hayward <alan.hayward@arm.com>
16224 David Sherwood <david.sherwood@arm.com>
16226 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
16227 from an unsigned int to a poly_uint64.
16228 (_loop_vec_info::slp_unrolling_factor): Likewise.
16229 (_loop_vec_info::vectorization_factor): Change from an int
16231 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
16232 (vect_get_num_vectors): New function.
16233 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
16234 (vect_get_num_copies): Use vect_get_num_vectors.
16235 (vect_analyze_data_ref_dependences): Change max_vf from an int *
16236 to an unsigned int *.
16237 (vect_analyze_data_refs): Change min_vf from an int * to a
16239 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
16240 than an unsigned HOST_WIDE_INT.
16241 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
16242 (vect_analyze_data_ref_dependence): Change max_vf from an int *
16243 to an unsigned int *.
16244 (vect_analyze_data_ref_dependences): Likewise.
16245 (vect_compute_data_ref_alignment): Handle polynomial vf.
16246 (vect_enhance_data_refs_alignment): Likewise.
16247 (vect_prune_runtime_alias_test_list): Likewise.
16248 (vect_shift_permute_load_chain): Likewise.
16249 (vect_supportable_dr_alignment): Likewise.
16250 (dependence_distance_ge_vf): Take the vectorization factor as a
16251 poly_uint64 rather than an unsigned HOST_WIDE_INT.
16252 (vect_analyze_data_refs): Change min_vf from an int * to a
16254 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
16255 vfm1 as a poly_uint64 rather than an int. Make the same change
16256 for the returned bound_scalar.
16257 (vect_gen_vector_loop_niters): Handle polynomial vf.
16258 (vect_do_peeling): Likewise. Update call to
16259 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
16260 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
16262 * tree-vect-loop.c (vect_determine_vectorization_factor)
16263 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
16264 (vect_get_known_peeling_cost): Likewise.
16265 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
16266 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
16267 (vect_transform_loop): Likewise. Use the lowest possible VF when
16268 updating the upper bounds of the loop.
16269 (vect_min_worthwhile_factor): Make static. Return an unsigned int
16270 rather than an int.
16271 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
16272 polynomial unroll factors.
16273 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
16274 (vect_make_slp_decision): Likewise.
16275 (vect_supported_load_permutation_p): Likewise, and polynomial
16277 (vect_analyze_slp_cost): Handle polynomial vf.
16278 (vect_slp_analyze_node_operations): Likewise.
16279 (vect_slp_analyze_bb_1): Likewise.
16280 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
16281 than an unsigned HOST_WIDE_INT.
16282 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
16283 (vectorizable_load): Handle polynomial vf.
16284 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
16286 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
16288 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16289 Alan Hayward <alan.hayward@arm.com>
16290 David Sherwood <david.sherwood@arm.com>
16292 * match.pd: Handle bit operations involving three constants
16293 and try to fold one pair.
16295 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16297 * tree-vect-loop-manip.c: Include gimple-fold.h.
16298 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
16299 niters_maybe_zero parameters. Handle other cases besides a step of 1.
16300 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
16301 Add a path that uses a step of VF instead of 1, but disable it
16303 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
16304 and niters_no_overflow parameters. Update calls to
16305 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
16306 Create a new SSA name if the latter choses to use a ste other
16307 than zero, and return it via niters_vector_mult_vf_var.
16308 * tree-vect-loop.c (vect_transform_loop): Update calls to
16309 vect_do_peeling, vect_gen_vector_loop_niters and
16310 slpeel_make_loop_iterate_ntimes.
16311 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
16312 (vect_gen_vector_loop_niters): Update declarations after above changes.
16314 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
16316 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
16317 128-bit round to integer instructions.
16318 (ceil<mode>2): Likewise.
16319 (btrunc<mode>2): Likewise.
16320 (round<mode>2): Likewise.
16322 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16324 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
16325 unaligned VSX load/store on P8/P9.
16326 (expand_block_clear): Allow the use of unaligned VSX
16327 load/store on P8/P9.
16329 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16331 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
16333 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
16334 swap associated with both a load and a store.
16336 2018-01-02 Andrew Waterman <andrew@sifive.com>
16338 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
16339 * config/riscv/riscv.md (clear_cache): Use it.
16341 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
16343 * web.c: Remove out-of-date comment.
16345 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16347 * expr.c (fixup_args_size_notes): Check that any existing
16348 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
16349 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
16350 (emit_single_push_insn): ...here.
16352 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16354 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
16355 (const_vector_encoded_nelts): New function.
16356 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
16357 (const_vector_int_elt, const_vector_elt): Declare.
16358 * emit-rtl.c (const_vector_int_elt_1): New function.
16359 (const_vector_elt): Likewise.
16360 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
16361 of CONST_VECTOR_ELT.
16363 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16365 * expr.c: Include rtx-vector-builder.h.
16366 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
16367 directly on the tree encoding.
16368 (const_vector_from_tree): Likewise.
16369 * optabs.c: Include rtx-vector-builder.h.
16370 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
16371 sequence of "u" values.
16372 * vec-perm-indices.c: Include rtx-vector-builder.h.
16373 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
16374 directly on the vec_perm_indices encoding.
16376 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16378 * doc/rtl.texi (const_vector): Describe new encoding scheme.
16379 * Makefile.in (OBJS): Add rtx-vector-builder.o.
16380 * rtx-vector-builder.h: New file.
16381 * rtx-vector-builder.c: Likewise.
16382 * rtl.h (rtx_def::u2): Add a const_vector field.
16383 (CONST_VECTOR_NPATTERNS): New macro.
16384 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
16385 (CONST_VECTOR_DUPLICATE_P): Likewise.
16386 (CONST_VECTOR_STEPPED_P): Likewise.
16387 (CONST_VECTOR_ENCODED_ELT): Likewise.
16388 (const_vec_duplicate_p): Check for a duplicated vector encoding.
16389 (unwrap_const_vec_duplicate): Likewise.
16390 (const_vec_series_p): Check for a non-duplicated vector encoding.
16391 Say that the function only returns true for integer vectors.
16392 * emit-rtl.c: Include rtx-vector-builder.h.
16393 (gen_const_vec_duplicate_1): Delete.
16394 (gen_const_vector): Call gen_const_vec_duplicate instead of
16395 gen_const_vec_duplicate_1.
16396 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
16397 (gen_const_vec_duplicate): Use rtx_vector_builder.
16398 (gen_const_vec_series): Likewise.
16399 (gen_rtx_CONST_VECTOR): Likewise.
16400 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
16401 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
16402 Build a new vector rather than modifying a CONST_VECTOR in-place.
16403 (handle_special_swappables): Update call accordingly.
16404 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
16405 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
16406 Build a new vector rather than modifying a CONST_VECTOR in-place.
16407 (handle_special_swappables): Update call accordingly.
16409 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16411 * simplify-rtx.c (simplify_const_binary_operation): Use
16412 CONST_VECTOR_ELT instead of XVECEXP.
16414 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16416 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
16417 the selector elements to be different from the data elements
16418 if the selector is a VECTOR_CST.
16419 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
16420 ssizetype for the selector.
16422 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16424 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
16425 before testing each element individually.
16426 * tree-vect-generic.c (lower_vec_perm): Likewise.
16428 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16430 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
16431 * selftest-run-tests.c (selftest::run_tests): Call it.
16432 * vector-builder.h (vector_builder::operator ==): New function.
16433 (vector_builder::operator !=): Likewise.
16434 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
16435 (vec_perm_indices::all_from_input_p): New function.
16436 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
16437 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
16438 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
16439 instead of reading the VECTOR_CST directly. Detect whether both
16440 vector inputs are the same before constructing the vec_perm_indices,
16441 and update the number of inputs argument accordingly. Use the
16442 utility functions added above. Only construct sel2 if we need to.
16444 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16446 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
16447 the broadcast of the low byte.
16448 (expand_mult_highpart): Use an explicit encoding for the permutes.
16449 * optabs-query.c (can_mult_highpart_p): Likewise.
16450 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
16451 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16452 (vectorizable_bswap): Likewise.
16453 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
16454 explicit encoding for the power-of-2 permutes.
16455 (vect_permute_store_chain): Likewise.
16456 (vect_grouped_load_supported): Likewise.
16457 (vect_permute_load_chain): Likewise.
16459 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16461 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
16462 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
16463 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
16464 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
16465 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
16466 (vect_gen_perm_mask_any): Likewise.
16468 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16470 * int-vector-builder.h: New file.
16471 * vec-perm-indices.h: Include int-vector-builder.h.
16472 (vec_perm_indices): Redefine as an int_vector_builder.
16473 (auto_vec_perm_indices): Delete.
16474 (vec_perm_builder): Redefine as a stand-alone class.
16475 (vec_perm_indices::vec_perm_indices): New function.
16476 (vec_perm_indices::clamp): Likewise.
16477 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
16478 (vec_perm_indices::new_vector): New function.
16479 (vec_perm_indices::new_expanded_vector): Update for new
16480 vec_perm_indices class.
16481 (vec_perm_indices::rotate_inputs): New function.
16482 (vec_perm_indices::all_in_range_p): Operate directly on the
16483 encoded form, without computing elided elements.
16484 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
16485 encoding. Update for new vec_perm_indices class.
16486 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
16487 the given vec_perm_builder.
16488 (expand_vec_perm_var): Update vec_perm_builder constructor.
16489 (expand_mult_highpart): Use vec_perm_builder instead of
16490 auto_vec_perm_indices.
16491 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
16492 vec_perm_indices instead of auto_vec_perm_indices. Use a single
16493 or double series encoding as appropriate.
16494 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
16495 vec_perm_indices instead of auto_vec_perm_indices.
16496 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16497 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16498 (vect_permute_store_chain): Likewise.
16499 (vect_grouped_load_supported): Likewise.
16500 (vect_permute_load_chain): Likewise.
16501 (vect_shift_permute_load_chain): Likewise.
16502 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
16503 (vect_transform_slp_perm_load): Likewise.
16504 (vect_schedule_slp_instance): Likewise.
16505 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16506 (vectorizable_mask_load_store): Likewise.
16507 (vectorizable_bswap): Likewise.
16508 (vectorizable_store): Likewise.
16509 (vectorizable_load): Likewise.
16510 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
16511 vec_perm_indices instead of auto_vec_perm_indices. Use
16512 tree_to_vec_perm_builder to read the vector from a tree.
16513 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
16514 vec_perm_builder instead of a vec_perm_indices.
16515 (have_whole_vector_shift): Use vec_perm_builder and
16516 vec_perm_indices instead of auto_vec_perm_indices. Leave the
16517 truncation to calc_vec_perm_mask_for_shift.
16518 (vect_create_epilog_for_reduction): Likewise.
16519 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
16520 from auto_vec_perm_indices to vec_perm_indices.
16521 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
16522 instead of changing individual elements.
16523 (aarch64_vectorize_vec_perm_const): Use new_vector to install
16524 the vector in d.perm.
16525 * config/arm/arm.c (expand_vec_perm_d::perm): Change
16526 from auto_vec_perm_indices to vec_perm_indices.
16527 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
16528 instead of changing individual elements.
16529 (arm_vectorize_vec_perm_const): Use new_vector to install
16530 the vector in d.perm.
16531 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
16532 Update vec_perm_builder constructor.
16533 (rs6000_expand_interleave): Likewise.
16534 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
16535 (rs6000_expand_interleave): Likewise.
16537 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16539 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
16540 to qimode could truncate the indices.
16541 * optabs.c (expand_vec_perm_var): Likewise.
16543 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16545 * Makefile.in (OBJS): Add vec-perm-indices.o.
16546 * vec-perm-indices.h: New file.
16547 * vec-perm-indices.c: Likewise.
16548 * target.h (vec_perm_indices): Replace with a forward class
16550 (auto_vec_perm_indices): Move to vec-perm-indices.h.
16551 * optabs.h: Include vec-perm-indices.h.
16552 (expand_vec_perm): Delete.
16553 (selector_fits_mode_p, expand_vec_perm_var): Declare.
16554 (expand_vec_perm_const): Declare.
16555 * target.def (vec_perm_const_ok): Replace with...
16556 (vec_perm_const): ...this new hook.
16557 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
16558 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
16559 * doc/tm.texi: Regenerate.
16560 * optabs.def (vec_perm_const): Delete.
16561 * doc/md.texi (vec_perm_const): Likewise.
16562 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
16563 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
16564 expand_vec_perm for constant permutation vectors. Assert that
16565 the mode of variable permutation vectors is the integer equivalent
16566 of the mode that is being permuted.
16567 * optabs-query.h (selector_fits_mode_p): Declare.
16568 * optabs-query.c: Include vec-perm-indices.h.
16569 (selector_fits_mode_p): New function.
16570 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
16571 is defined, instead of checking whether the vec_perm_const_optab
16572 exists. Use targetm.vectorize.vec_perm_const instead of
16573 targetm.vectorize.vec_perm_const_ok. Check whether the indices
16574 fit in the vector mode before using a variable permute.
16575 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
16576 vec_perm_indices instead of an rtx.
16577 (expand_vec_perm): Replace with...
16578 (expand_vec_perm_const): ...this new function. Take the selector
16579 as a vec_perm_indices rather than an rtx. Also take the mode of
16580 the selector. Update call to shift_amt_for_vec_perm_mask.
16581 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
16582 Use vec_perm_indices::new_expanded_vector to expand the original
16583 selector into bytes. Check whether the indices fit in the vector
16584 mode before using a variable permute.
16585 (expand_vec_perm_var): Make global.
16586 (expand_mult_highpart): Use expand_vec_perm_const.
16587 * fold-const.c: Includes vec-perm-indices.h.
16588 * tree-ssa-forwprop.c: Likewise.
16589 * tree-vect-data-refs.c: Likewise.
16590 * tree-vect-generic.c: Likewise.
16591 * tree-vect-loop.c: Likewise.
16592 * tree-vect-slp.c: Likewise.
16593 * tree-vect-stmts.c: Likewise.
16594 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
16596 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
16597 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
16598 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
16599 (aarch64_vectorize_vec_perm_const): ...this new function.
16600 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16601 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16602 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
16603 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
16604 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16605 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16606 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
16608 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
16609 check for NEON modes.
16610 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
16611 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
16612 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
16613 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
16615 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
16616 the old VEC_PERM_CONST conditions.
16617 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
16618 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
16619 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
16620 (ia64_vectorize_vec_perm_const_ok): Merge into...
16621 (ia64_vectorize_vec_perm_const): ...this new function.
16622 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
16623 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
16624 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
16625 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
16626 * config/mips/mips.c (mips_expand_vec_perm_const)
16627 (mips_vectorize_vec_perm_const_ok): Merge into...
16628 (mips_vectorize_vec_perm_const): ...this new function.
16629 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
16630 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
16631 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
16632 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
16633 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
16634 (rs6000_expand_vec_perm_const): Delete.
16635 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
16637 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16638 (altivec_expand_vec_perm_const_le): Take each operand individually.
16639 Operate on constant selectors rather than rtxes.
16640 (altivec_expand_vec_perm_const): Likewise. Update call to
16641 altivec_expand_vec_perm_const_le.
16642 (rs6000_expand_vec_perm_const): Delete.
16643 (rs6000_vectorize_vec_perm_const_ok): Delete.
16644 (rs6000_vectorize_vec_perm_const): New function.
16645 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
16646 an element count and rtx array.
16647 (rs6000_expand_extract_even): Update call accordingly.
16648 (rs6000_expand_interleave): Likewise.
16649 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
16650 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
16651 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
16652 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
16653 (rs6000_expand_vec_perm_const): Delete.
16654 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16655 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16656 (altivec_expand_vec_perm_const_le): Take each operand individually.
16657 Operate on constant selectors rather than rtxes.
16658 (altivec_expand_vec_perm_const): Likewise. Update call to
16659 altivec_expand_vec_perm_const_le.
16660 (rs6000_expand_vec_perm_const): Delete.
16661 (rs6000_vectorize_vec_perm_const_ok): Delete.
16662 (rs6000_vectorize_vec_perm_const): New function. Remove stray
16663 reference to the SPE evmerge intructions.
16664 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
16665 an element count and rtx array.
16666 (rs6000_expand_extract_even): Update call accordingly.
16667 (rs6000_expand_interleave): Likewise.
16668 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
16669 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
16671 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16673 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16675 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
16676 vector mode and that that mode matches the mode of the data
16678 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
16679 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
16680 directly using expand_vec_perm_1 when forcing selectors into
16682 (expand_vec_perm_var): New function, split out from expand_vec_perm.
16684 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16686 * optabs-query.h (can_vec_perm_p): Delete.
16687 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
16688 * optabs-query.c (can_vec_perm_p): Split into...
16689 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
16690 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
16691 particular selector is valid.
16692 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16693 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16694 (vect_grouped_load_supported): Likewise.
16695 (vect_shift_permute_load_chain): Likewise.
16696 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
16697 (vect_transform_slp_perm_load): Likewise.
16698 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16699 (vectorizable_bswap): Likewise.
16700 (vect_gen_perm_mask_checked): Likewise.
16701 * fold-const.c (fold_ternary_loc): Likewise. Don't take
16702 implementations of variable permutation vectors into account
16703 when deciding which selector to use.
16704 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
16705 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
16706 with a false third argument.
16707 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
16708 to test whether the constant selector is valid and can_vec_perm_var_p
16709 to test whether a variable selector is valid.
16711 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16713 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
16714 * optabs-query.c (can_vec_perm_p): Likewise.
16715 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
16716 instead of vec_perm_indices.
16717 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
16718 (vect_gen_perm_mask_checked): Likewise,
16719 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
16720 (vect_gen_perm_mask_checked): Likewise,
16722 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16724 * optabs-query.h (qimode_for_vec_perm): Declare.
16725 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
16726 (qimode_for_vec_perm): ...this new function.
16727 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
16729 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16731 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
16732 does not have a conditional at the top.
16734 2018-01-02 Richard Biener <rguenther@suse.de>
16736 * ipa-inline.c (big_speedup_p): Fix expression.
16738 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
16741 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
16744 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
16748 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
16749 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
16750 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
16751 cond_taken_branch_cost 3->4.
16753 2018-01-01 Jakub Jelinek <jakub@redhat.com>
16755 PR tree-optimization/83581
16756 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
16757 TODO_cleanup_cfg if any changes have been made.
16759 PR middle-end/83608
16760 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
16761 convert_modes if target mode has the right side, but different mode
16764 PR middle-end/83609
16765 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
16766 last argument when extracting from CONCAT. If either from_real or
16767 from_imag is NULL, use expansion through memory. If result is not
16768 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
16769 the parts directly to inner mode, if even that fails, use expansion
16772 PR middle-end/83623
16773 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
16774 check for bswap in mode rather than HImode and use that in expand_unop
16777 Copyright (C) 2018 Free Software Foundation, Inc.
16779 Copying and distribution of this file, with or without modification,
16780 are permitted in any medium without royalty provided the copyright
16781 notice and this notice are preserved.