1 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
4 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
5 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
6 * dojump.c (do_jump): Likewise.
7 * expr.c (expand_expr_real_2): Check operand type's sign.
8 * fold-const.c (const_unop): Handle ABSU_EXPR.
9 (fold_abs_const): Likewise.
10 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
11 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
12 (strip_sign_op_1): Likesise.
13 * match.pd: Add new pattern to generate ABSU_EXPR.
14 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
15 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
16 * tree-eh.c (operation_could_trap_helper_p): Likewise.
17 * tree-inline.c (estimate_operator_cost): Likewise.
18 * tree-pretty-print.c (dump_generic_node): Likewise.
19 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
20 * tree.def (ABSU_EXPR): New.
22 2018-06-16 Jakub Jelinek <jakub@redhat.com>
25 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
26 documented as preserved for backward compatibility only.
27 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
29 PR rtl-optimization/86108
30 * bb-reorder.c (create_forwarder_block): Renamed to ...
31 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
32 jump from new landing pad to the second part.
33 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
36 2018-06-15 Jakub Jelinek <jakub@redhat.com>
39 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
40 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
41 Only call store_expr for halves if the mode is the same.
44 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
45 Fix up comment formatting.
47 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
49 * typed-splay-tree.h (typed_splay_tree::remove): New function.
50 (typed_splay_tree::closure,
51 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
52 (typed_splay_tree::typed_splay_tree,
53 typed_splay_tree::operator =): Declared private.
54 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
55 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
56 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
57 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
58 typed_splay_tree::splay_tree_splay,
59 typed_splay_tree::splay_tree_foreach_helper,
60 typed_splay_tree::splay_tree_insert,
61 typed_splay_tree::splay_tree_remove,
62 typed_splay_tree::splay_tree_lookup,
63 typed_splay_tree::splay_tree_predecessor,
64 typed_splay_tree::splay_tree_successor,
65 typed_splay_tree::splay_tree_min,
66 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
67 (typed_splay_tree::root, typed_splay_tree::comp,
68 typed_splay_tree::delete_key,
69 typed_splay_tree::delete_value): New data members.
70 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
71 typed_splay_tree::remove.
73 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
75 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
76 -mginv and -mno-ginv to the assembler.
77 * config/mips/mips.opt (-mcrc): New option.
79 * doc/invoke.text (-mcrc): Document.
82 2018-06-15 Nick Clifton <nickc@redhat.com>
85 * tree.c (escaped_string): New class. Converts an unescaped
86 string into its escaped equivalent.
87 (warn_deprecated_use): Use the new class to convert the
88 deprecation message, if present.
89 (test_escaped_strings): New self test.
90 (test_c_tests): Add test_escaped_strings.
91 * doc/extend.texi (deprecated): Add a note that the
92 deprecation message is affected by the -fmessage-length
93 option, and that control characters will be escaped.
94 (#pragma GCC error): Document this pragma.
95 (#pragma GCC warning): Likewise.
96 * doc/invoke.texi (-fmessage-length): Document this option's
97 effect on the #warning and #error preprocessor directives and
98 the deprecated attribute.
100 2018-06-15 Richard Biener <rguenther@suse.de>
102 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
103 here, also noting vector size used.
104 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
105 size used in MSG_OPTIMIZED_LOCATIONS dump.
106 (pass_slp_vectorize::execute): Adjust.
108 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
111 * config/arc/arc.c (arc_return_address_register): Fix
114 2018-06-15 Richard Biener <rguenther@suse.de>
117 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
118 leave useless conversion stripping to force_gimple_operand_gsi.
119 (gimplify_build2): Likewise.
120 (gimplify_build1): Likewise.
122 2018-06-15 Richard Biener <rguenther@suse.de>
125 * tree-cfg.c (move_stmt_op): unshare invariant addresses
126 before adjusting their block.
128 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
130 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
131 multilibs for *-*-rtems*.
132 * config/riscv/t-rtems: New file.
134 2018-06-14 Jakub Jelinek <jakub@redhat.com>
137 * match.pd ((A +- CST1) +- CST2): Punt if last resort
138 unsigned_type_for returns NULL.
141 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
142 subregs of multi-word pseudos unless the float mode has word size.
144 2018-06-14 Richard Biener <rguenther@suse.de>
147 * tree-vect-generic.c (build_word_mode_vector_type): Remove
148 duplicate and harmful type_hash_canon.
149 * tree.c (type_hash_canon): Assert we didn't find ourselves.
151 2018-06-14 Richard Biener <rguenther@suse.de>
154 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
157 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
159 * config/rtems.h (STDINT_LONG32): Define.
161 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
162 Prachi Godbole <prachi.godbole@imgtec.com>
164 * config/mips/mips-cpus.def: Define P6600.
165 * config/mips/mips-tables.opt: Regenerate.
166 * config/mips/mips.c (mips_ucbranch_type): New enum.
167 (mips_rtx_cost_data): Add support for P6600.
168 (mips_issue_rate): Likewise.
169 (mips_multipass_dfa_lookahead): Likewise.
170 (mips_avoid_hazard): Likewise.
171 (mips_reorg_process_insns): Likewise.
172 (mips_classify_branch_p6600): New function.
173 * config/mips/mips.h (TUNE_P6600): New define.
174 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
175 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
176 * config/mips/mips.md: Include p6600.md.
177 (processor): Add p6600.
178 * config/mips/p6600.md: New file.
179 * doc/invoke.texi: Add p6600 to supported architectures.
181 2018-06-13 Martin Sebor <msebor@redhat.com>
183 PR tree-optimization/86114
184 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
186 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
188 2018-06-13 Richard Biener <rguenther@suse.de>
190 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
191 Properly set vector type of the intermediate stmt.
192 * tree-vect-stmts.c (vectorizable_operation): The destination
193 var always has vectype_out type.
195 2018-06-13 Jeff Law <law@redhat.com>
197 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
198 integer 0 for argument to print_rtl_with_bb.
199 (rl78_reorg): Likewise.
201 2018-06-13 David Malcolm <dmalcolm@redhat.com>
203 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
204 from rtx to rtx_insn *.
205 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
207 (add_sched_insns_for_speculation): Likewise for local "target",
208 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
209 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
210 from rtx_insn ** to rtx_code_label **.
211 (reorg_emit_nops): Likewise.
212 (c6x_reorg): Likewise for local "call_labels".
213 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
215 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
216 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
217 the loops over LABEL_REFS.
218 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
220 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
221 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
222 (split_branches): Strengthen local "olabel" from rtx to
223 rtx_insn *, adding a safe_as_a cast.
224 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
226 (add_insn_after): Likewise for first two params.
227 (add_insn_before): Likewise.
228 (remove_insn): Likewise for param.
229 (emit_pattern_before_noloc): Likewise for second and third params.
230 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
231 (emit_call_insn_before_noloc): Likewise.
232 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
234 (emit_barrier_before): Likewise.
235 (emit_label_before): Strengthen "label" param from "rtx" to
236 "rtx_code_label *". Strengthen "before" param from "rtx" to
238 (emit_insn_after_1): Strengthen "after" param from "rtx" to
240 (emit_pattern_after_noloc): Likewise.
241 (emit_insn_after_noloc): Likewise.
242 (emit_jump_insn_after_noloc): Likewise.
243 (emit_call_insn_after_noloc): Likewise.
244 (emit_debug_insn_after_noloc): Likewise.
245 (emit_barrier_after): Likewise.
246 (emit_label_after): Likewise for both params.
247 (emit_pattern_after_setloc): Likewise for "after" param. Convert
248 "loc" param from "int" to "location_t".
249 (emit_insn_after_setloc): Likewise.
250 (emit_jump_insn_after_setloc): Likewise.
251 (emit_call_insn_after_setloc): Likewise.
252 (emit_debug_insn_after_setloc): Likewise.
253 (emit_pattern_before_setloc): Likewise for "before" param. Convert
254 "loc" param from "int" to "location_t".
255 (emit_pattern_before): Convert NULL_RTX to NULL.
256 (emit_insn_before_setloc): Convert "loc" param from "int" to
258 (emit_jump_insn_before_setloc): Likewise.
259 (emit_call_insn_before_setloc): Likewise.
260 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
261 rtx_insn *. Convert "loc" param from "int" to "location_t".
262 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
263 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
264 Convert 3rd param from "int" to "location_t".
265 (emit_barrier_before, emit_barrier_after, next_real_insn):
266 Strengthen param from rtx to rtx_insn *.
267 (emit_label_before): Strengthen 1st param from "rtx" to
268 "rtx_code_label *". Strengthen 2nd param from "rtx" to
270 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
271 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
272 Strengthen 2nd param from "rtx" to "rtx_insn *".
273 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
274 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
275 Likewise. Convert 3rd param from "int" to "location_t".
276 (emit_label_after): Strengthen 1st param from "rtx" to
278 (next_real_insn, remove_insn): Strengthen param from "rtx" to
280 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
281 from "rtx" to "rtx_insn *".
283 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
285 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
286 bodies streamed in with -Q.
287 * dumpfile.c (dump_files): Add lto-stream-out dump file.
288 * dumpfile.h (tree_dump_index): Add lto_stream_out.
289 * gimple-streamer-out.c: Include gimple-pretty-print.h
290 (output_bb): Dump stmts streamed.
291 * lto-section-out.c: Include print-tree.h
292 (lto_begin_section): Dump sections created.
293 (lto_output_decl_index): Dump decl encoded.
294 * lto-streamer-out.c: Include print-tree.h
295 (create_output_block): Dump output block created.
296 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
297 (output_function): Dump function output.
298 (output_constructor): Dump constructor streamed.
299 (write_global_stream): Output indexes encoded.
300 (produce_asm_for_decls): Dump streams encoded.
301 * lto-streamer.c (streamer_dump_file): New global var.
302 * lto-streamer.h (streamer_dump_file): Declare.
303 * passes.c (ipa_write_summaries): Initialize streamer dump.
304 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
307 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
310 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
311 offsets for register save directives. Emit a second batch of save
312 directives, if need be, when the function accesses prior frames.
314 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
316 * config/arc/fpu.md (fmasf4): Force operand to register.
319 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
321 * config/arc/arc-protos.h (arc_pad_return): Remove.
322 * config/arc/arc.c (machine_function): Remove force_short_suffix
324 (arc_print_operand): Adjust printing of '&'.
325 (arc_verify_short): Remove conditional printing of short suffix.
326 (arc_final_prescan_insn): Remove reference to size_reason.
327 (pad_return): New function.
328 (arc_reorg): Call pad_return.
329 (arc_pad_return): Remove.
330 (arc_init_machine_status): Remove reference to force_short_suffix.
331 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
332 (attr length): When attribute iscompact is true force to 2
333 regardless; in the case of maybe check if we want to force the
334 instruction to have 4 bytes length.
335 (nopv): Change it to generate 4 byte long nop as well.
336 (blockage): New pattern.
337 (simple_return): Remove call to arc_pad_return.
338 (p_return_i): Likewise.
340 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
342 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
344 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
346 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
349 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
351 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
352 for ARC700 and ARCv2.
354 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
357 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
358 operands[2] instead of operands[1].
361 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
363 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
364 case, check whether the outer register overlaps an unallocatable
365 register, not just whether it fits the required class.
367 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
369 * poly-int.h (can_div_trunc_p): Add new overload in which all values
371 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
372 (memrefs_conflict_p): Likewise.
373 (init_alias_analysis): Likewise.
374 * cfgexpand.c (expand_debug_expr): Likewise.
375 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
376 * cse.c (fold_rtx): Likewise.
377 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
378 * expr.c (emit_block_move_hints): Likewise.
379 (clear_storage_hints, push_block, emit_push_insn): Likewise.
380 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
381 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
382 (emit_group_store): Likewise.
383 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
384 to read the PRE/POST_MODIFY increment.
385 * calls.c (store_one_arg): Use strip_offset.
386 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
388 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
390 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
391 (simplify_binary_operation_1): Extend CONST_INT handling to
393 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
394 than a HOST_WIDE_INT.
395 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
397 (adjust_mems, add_stores): Update accodingly.
398 (vt_canonicalize_addr): Track polynomial offsets.
399 (emit_note_insn_var_location): Likewise.
400 (vt_add_function_parameter): Likewise.
401 (vt_initialize): Likewise.
403 2018-06-12 Jeff Law <law@redhat.com>
405 * config.gcc (alpha*-*-freebsd*): Remove.
406 * config/alpha/freebsd.h: Remove.
408 2018-06-12 David Malcolm <dmalcolm@redhat.com>
411 * spellcheck-tree.c (levenshtein_distance): Rename to...
412 (get_edit_distance): ...this, and update for underlying renaming.
413 * spellcheck-tree.h (levenshtein_distance): Rename to...
414 (get_edit_distance): ...this.
415 * spellcheck.c (levenshtein_distance): Rename to...
416 (get_edit_distance): ...this. Convert from Levenshtein distance
417 to Damerau-Levenshtein distance by supporting transpositions of
418 adjacent characters. Rename "v1" to "v_next" and "v0" to
420 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
421 (selftest::test_edit_distance_unit_test_oneway): ...this, and
422 update for underlying renaming.
423 (selftest::levenshtein_distance_unit_test): Rename to...
424 (selftest::test_get_edit_distance_unit): ...this, and update for
426 (selftest::test_find_closest_string): Add example from PR 69968
427 where transposition helps
428 (selftest::test_metric_conditions): Update for renaming.
429 (selftest::test_metric_conditions): Likewise.
430 (selftest::spellcheck_c_tests): Likewise.
431 * spellcheck.h (levenshtein_distance): Rename both overloads to...
432 (get_edit_distance): ...this.
433 (best_match::consider): Update for renaming.
435 2018-06-12 Martin Sebor <msebor@redhat.com>
437 PR tree-optimization/85259
438 * builtins.c (compute_objsize): Handle constant offsets.
439 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
440 true iff a warning has been issued.
441 * gimple.h (gimple_nonartificial_location): New function.
442 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
443 gimple_nonartificial_location and handle -Wno-system-headers.
444 (handle_builtin_stxncpy): Same.
446 2018-06-12 Martin Sebor <msebor@redhat.com>
449 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
451 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
453 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
454 BUILTIN_VEC_XST entries for pointer to double and long long.
456 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
459 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
461 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
464 2018-06-12 Martin Liska <mliska@suse.cz>
466 * doc/options.texi: Document IntegerRange.
468 2018-06-12 Martin Liska <mliska@suse.cz>
470 * config/i386/i386.opt: Make MPX-related options as Deprecated.
471 * opt-functions.awk: Handle Deprecated flag.
472 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
474 (read_cmdline_option): Report warning for a deprecated option.
475 * opts.h (struct cl_option): Add new field cl_deprecated.
476 (CL_ERR_DEPRECATED): New.
478 2018-06-12 Martin Liska <mliska@suse.cz>
480 * doc/options.texi: Document Deprecated option flag.
482 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
484 * config/arc/arc-arch.h (arc_extras): New enum.
485 (arc_cpu_t):Add field extra.
486 (arc_cpu_types): Consider the extras.
487 * config/arc/arc-cpus.def: Add extras info.
488 * config/arc/arc-opts.h (processor_type): Consider extra field.
489 * config/arc/arc.c (arc_override_options): Handle extra field.
491 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
493 * config/arc/arc-arch.h: Update ARC_OPTX macro.
494 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
496 * config/arc/arc.c (arc_init): Update pic warning.
497 (irq_range): Update irq range parsing warnings.
498 (arc_override_options): Update various warning messages.
499 (arc_handle_aux_attribute): Likewise.
501 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
503 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
505 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
507 * doc/sourcebuild.texi: Document usage of line number 0 in verify
508 compiler messages directives.
510 2018-06-12 Matthew Fortune <mfortune@gmail.com>
512 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
513 * config/mips/mips-tables.opt: Regenerate.
514 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
516 * doc/invoke.texi: Document -march=i6500.
518 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
520 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
521 (i6400_gpmul): Add cpu_unit.
522 (i6400_gpdiv): Likewise.
523 (i6400_msa_add_d): Update reservations.
524 (i6400_msa_int_add) Likewise.
525 (i6400_msa_short_logic3) Likewise.
526 (i6400_msa_short_logic2) Likewise.
527 (i6400_msa_short_logic) Likewise.
528 (i6400_msa_move) Likewise.
529 (i6400_msa_cmp) Likewise.
530 (i6400_msa_short_float2) Likewise.
531 (i6400_msa_div_d) Likewise.
532 (i6400_msa_long_logic1) Likewise.
533 (i6400_msa_long_logic2) Likewise.
534 (i6400_msa_mult) Likewise.
535 (i6400_msa_long_float2) Likewise.
536 (i6400_msa_long_float4) Likewise.
537 (i6400_msa_long_float5) Likewise.
538 (i6400_msa_long_float8) Likewise.
539 (i6400_fpu_fadd): Include frint type.
540 (i6400_fpu_store): New define_insn_reservation.
541 (i6400_fpu_load): Likewise.
542 (i6400_fpu_move): Likewise.
543 (i6400_fpu_fcmp): Likewise.
544 (i6400_fpu_fmadd): Likewise.
545 (i6400_int_mult): Include imul3nc type and update reservation.
546 (i6400_int_div): Include idiv3 type and update reservation.
547 (i6400_int_load): Update to check type not move_type.
548 (i6400_int_store): Likewise.
549 (i6400_int_prefetch): Set zero latency.
551 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
553 * gcc.c: Document new %@{...} sequence.
554 (LINK_COMMAND_SPEC): Use it for the -L switches.
555 (cpp_unique_options): Use it for the -I switches.
556 (at_file_argbuf): New global variable.
557 (in_at_file): Likewise.
558 (alloc_args): Create at_file_argbuf.
559 (clear_args): Truncate at_file_argbuf.
560 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
561 (open_at_file): New function.
562 (close_at_file): Likewise.
563 (create_at_file): Delete.
564 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
567 (validate_switches_from_spec): Deal with %@{...} sequence.
568 (validate_switches): Likewise.
569 (driver::finalize): Call clear_args.
571 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
573 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
575 2018-06-11 Martin Sebor <msebor@redhat.com>
577 * doc/invoke.texi (-Wall): List -Wc++17-compat.
578 (Wno-class-memaccess): Add @opindex.
579 (Wno-templates, Wno-multiple-inheritance): Same.
580 (Wno-virtual-inheritance, Wno-namespaces): Same.
581 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
582 (Wno-format-overflow, Wno-format-truncation): Same.
583 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
584 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
585 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
586 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
587 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
588 (Wno-misspelled-isr): Same.
590 2018-06-11 Martin Sebor <msebor@redhat.com>
592 * PR tree-optimization/86083
593 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
595 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
597 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
599 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
602 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
603 on the correct operand.
604 (*movdi_internal64): Ditto.
606 2018-06-11 Martin Liska <mliska@suse.cz>
608 PR tree-optimization/86089
609 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
611 2018-06-11 Julia Koval <julia.koval@intel.com>
613 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
614 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
615 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
617 2018-06-11 Olivier Hainque <hainque@adacore.com>
619 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
620 for Ada with strict dwarf2.
622 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
625 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
628 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
630 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
632 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
634 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
637 2018-06-08 Martin Liska <mliska@suse.cz>
639 * tree-cfg.h (debug_function): Fix argument type to match
642 2018-06-08 Martin Liska <mliska@suse.cz>
644 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
645 Remove usage of MPX-related (and removed) fields.
646 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
648 2018-06-08 David Malcolm <dmalcolm@redhat.com>
650 * cfg.c (debug): Use TDF_NONE rather than 0.
651 * cfghooks.c (debug): Likewise.
652 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
653 (struct dump_option_value_info): Convert to...
654 (struct kv_pair): ...this template type.
655 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
657 (optinfo_verbosity_options): Likewise.
658 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
660 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
661 than int for "optgroup_flags" param.
662 (dump_generic_expr_loc): Use dump_flags_t rather than int for
664 (dump_dec): Likewise.
665 (dump_finish): Use TDF_NONE rather than 0.
666 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
667 rather than int for "optgroup_flags" param. Use TDF_NONE rather
668 than 0. Update for change to option_ptr.
669 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
670 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
671 0. Update for changes to optinfo_verbosity_options and
673 (opt_info_switch_p): Convert optgroup_flags from int to
675 (dump_basic_block): Use dump_flags_t rather than int
676 for "dump_kind" param.
677 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
678 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
679 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
680 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
681 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
682 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
683 TDF_NONE): Convert from macros to...
684 (enum dump_flag): ...this new enum.
685 (dump_flags_t): Update to use enum.
686 (operator|, operator&, operator~, operator|=, operator&=):
687 Implement for dump_flags_t.
688 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
689 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
690 Convert from macros to...
691 (enum optgroup_flag): ...this new enum.
692 (optgroup_flags_t): New typedef.
693 (operator|, operator|=): Implement for optgroup_flags_t.
694 (struct dump_file_info): Convert field "alt_flags" to
695 dump_flags_t. Convert field "optgroup_flags" to
697 (dump_basic_block): Use dump_flags_t rather than int for param.
698 (dump_generic_expr_loc): Likewise.
699 (dump_dec): Likewise.
700 (dump_register): Convert param "optgroup_flags" to
702 (opt_info_enable_passes): Likewise.
703 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
705 * gimple-pretty-print.c (debug): Likewise.
706 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
707 (merged_store_group::apply_stores): Likewise.
708 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
709 * gimple.c (verify_gimple_pp): Likewise.
710 * graphite-poly.c (print_pbb_body): Likewise.
711 * passes.c (pass_manager::register_one_dump_file): Convert
712 local "optgroup_flags" to optgroup_flags_t.
713 * print-tree.c (print_node): Use TDF_NONE rather than 0.
715 (debug_body): Likewise.
716 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
718 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
720 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
721 (convert_mult_to_fma): Likewise.
722 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
723 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
724 * tree-vect-data-refs.c (dump_lower_bound): Convert param
725 "dump_kind" to dump_flags_t.
727 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
729 * config/rs6000/rs6000.c (min, max): Delete.
731 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
733 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
734 -mabi=spe and -mabi=no-spe.
736 2018-06-08 Martin Liska <mliska@suse.cz>
738 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
739 where we expect an existing summary.
741 2018-06-08 Martin Liska <mliska@suse.cz>
743 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
744 * ipa-inline.h (estimate_edge_growth): Likewise.
746 2018-06-08 Martin Liska <mliska@suse.cz>
748 * cgraph.c (function_version_hasher::hash): Use
749 cgraph_node::get_uid ().
750 (function_version_hasher::equal):
751 * cgraph.h (cgraph_node::get_uid): New method.
752 * ipa-inline.c (update_caller_keys): Use
753 cgraph_node::get_uid ().
754 (update_callee_keys): Likewise.
755 * ipa-utils.c (searchc): Likewise.
756 (ipa_reduced_postorder): Likewise.
757 * lto-cgraph.c (input_node): Likewise.
758 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
759 * symbol-summary.h (symtab_insertion): Likewise.
760 (symtab_removal): Likewise.
761 (symtab_duplication): Likewise.
762 * tree-pretty-print.c (dump_function_header): Likewise.
763 * tree-sra.c (convert_callers_for_node): Likewise.
765 2018-06-08 Martin Liska <mliska@suse.cz>
767 * cgraph.c (symbol_table::create_edge): Always assign a new
769 (symbol_table::free_edge): Do not recycle numbers.
770 * cgraph.h (cgraph_edge::get): New method.
771 * symbol-summary.h (symtab_removal): Use it.
772 (symtab_duplication): Likewise.
773 (call_summary::hashable_uid): Remove.
775 2018-06-08 Martin Liska <mliska@suse.cz>
777 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
778 (initialize_growth_caches): Remove.
779 (free_growth_caches): Likewise.
780 (do_estimate_edge_time): Use edge_growth_cache.
781 (do_estimate_edge_size): Likewise.
782 (do_estimate_edge_hints): Likewise.
783 * ipa-inline.c (reset_edge_caches): Likewise.
784 (recursive_inlining): Likewise.
785 (inline_small_functions): Likewise.
786 * ipa-inline.h (initialize_growth_caches): Remove.
787 (estimate_edge_size): Likewise.
788 (estimate_edge_time): Likewise.
789 (estimate_edge_hints): Likewise.
790 (reset_edge_growth_cache): Likewise.
791 * symbol-summary.h (call_summary::remove): New method.
793 2018-06-08 Martin Liska <mliska@suse.cz>
795 * ipa-cp.c (class edge_clone_summary): New summary.
796 (grow_edge_clone_vectors): Remove.
797 (ipcp_edge_duplication_hook): Remove.
798 (class edge_clone_summary_t): New call_summary class.
799 (ipcp_edge_removal_hook): Remove.
800 (edge_clone_summary_t::duplicate): New function.
801 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
802 (create_specialized_node): Likewise.
803 (ipcp_driver): Initialize edge_clone_summaries and do not
806 2018-06-08 Martin Liska <mliska@suse.cz>
808 * symbol-summary.h (get): New function.
809 (call_summary::m_initialize_when_cloning): New class member.
811 2018-06-08 Martin Liska <mliska@suse.cz>
813 * cgraph.c (cgraph_node::remove): Do not recycle uid.
814 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
815 (symbol_table::allocate_cgraph_symbol): Do not set uid.
816 * passes.c (uid_hash_t): Record removed_nodes by their uids.
817 (remove_cgraph_node_from_order): Use the removed_nodes set.
818 (do_per_function_toporder): Likwise.
819 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
820 instead of summary_uid.
821 (symtab_removal): Likewise.
822 (symtab_duplication): Likewise.
824 2018-06-08 Martin Liska <mliska@suse.cz>
826 * ipa-cp.c (ipcp_store_bits_results): Use
827 ipcp_transformation_sum.
828 (ipcp_store_vr_results): Likewise.
829 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
831 (ipcp_transformation_initialize): ... this.
832 (ipa_set_node_agg_value_chain):
833 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
834 (write_ipcp_transformation_info): Likewise.
835 (read_ipcp_transformation_info): Likewise.
836 (ipcp_update_bits): Likewise.
837 (ipcp_update_vr): Likewise.
838 (ipcp_transform_function): Likewise.
839 * ipa-prop.h: Rename ipcp_transformation_summary to
841 (class ipcp_transformation_t): New function summary.
842 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
843 (ipa_get_agg_replacements_for_node): Likewise.
845 2018-06-08 Martin Liska <mliska@suse.cz>
847 * ipa-pure-const.c (struct funct_state_d): Do it class instead
849 (class funct_state_summary_t): New function_summary class.
850 (has_function_state): Remove.
851 (get_function_state): Likewise.
852 (set_function_state): Likewise.
853 (add_new_function): Likewise.
854 (funct_state_summary_t::insert): New function.
855 (duplicate_node_data): Remove.
856 (remove_node_data): Remove.
857 (funct_state_summary_t::duplicate): New function.
858 (register_hooks): Create new funct_state_summaries.
859 (pure_const_generate_summary): Use it.
860 (pure_const_write_summary): Likewise.
861 (pure_const_read_summary): Likewise.
862 (propagate_pure_const): Likewise.
863 (propagate_nothrow): Likewise.
864 (dump_malloc_lattice): Likewise.
865 (propagate_malloc): Likewise.
866 (execute): Do not register hooks, just remove summary
868 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
871 2018-06-08 Martin Liska <mliska@suse.cz>
873 * ipa-reference.c (remove_node_data): Remove.
874 (duplicate_node_data): Likewise.
875 (class ipa_ref_var_info_summary_t): New class.
876 (class ipa_ref_opt_summary_t): Likewise.
877 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
878 (get_reference_optimization_summary): Use
879 ipa_ref_opt_sum_summaries.
880 (set_reference_vars_info): Remove.
881 (set_reference_optimization_summary): Likewise.
882 (ipa_init): Create summaries.
883 (init_function_info): Use function summary.
884 (ipa_ref_opt_summary_t::duplicate): New function.
885 (ipa_ref_opt_summary_t::remove): New function.
886 (get_read_write_all_from_node): Fix GNU coding style.
887 (propagate): Use function summary.
888 (write_node_summary_p): Fix GNU coding style.
889 (stream_out_bitmap): Likewise.
890 (ipa_reference_read_optimization_summary): Use function summary.
891 (ipa_reference_c_finalize): Do not release hooks.
893 2018-06-08 Martin Liska <mliska@suse.cz>
895 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
896 (analyze_function_body): Extract multiple calls of get_create.
897 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
898 * ipa-inline.c (recursive_inlining): Use ::get method.
899 * ipa-inline.h (estimate_edge_growth): Likewise.
901 2018-06-08 Martin Liska <mliska@suse.cz>
903 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
905 (hsa_function_summary::hsa_function_summary): Use the new enum
907 (hsa_gpu_implementation_p): Use hsa_summaries::get.
908 * hsa-gen.c (hsa_get_host_function): Likewise.
909 (get_brig_function_name): Likewise.
910 * ipa-hsa.c (process_hsa_functions): Likewise.
911 (ipa_hsa_write_summary): Likewise.
912 * symbol-summary.h (symtab_duplication): Use ::get function/
915 2018-06-08 Martin Liska <mliska@suse.cz>
917 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
919 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
920 (hsa_register_kernel): Likewise.
921 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
922 * hsa-gen.c (hsa_get_host_function): Likewise.
923 (get_brig_function_name): Likewise.
924 (generate_hsa): Likewise.
925 (pass_gen_hsail::execute): Likewise.
926 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
927 (devirtualization_time_bonus): Likewise.
928 (ipcp_propagate_stage): Likewise.
929 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
930 (edge_set_predicate): Likewise.
931 (evaluate_conditions_for_known_args): Likewise.
932 (evaluate_properties_for_edge): Likewise.
933 (ipa_fn_summary::reset): Likewise.
934 (ipa_fn_summary_t::duplicate): Likewise.
935 (dump_ipa_call_summary): Likewise.
936 (ipa_dump_fn_summary): Likewise.
937 (analyze_function_body): Likewise.
938 (compute_fn_summary): Likewise.
939 (estimate_edge_devirt_benefit): Likewise.
940 (estimate_edge_size_and_time): Likewise.
941 (estimate_calls_size_and_time): Likewise.
942 (estimate_node_size_and_time): Likewise.
943 (inline_update_callee_summaries): Likewise.
944 (remap_edge_change_prob): Likewise.
945 (remap_edge_summaries): Likewise.
946 (ipa_merge_fn_summary_after_inlining): Likewise.
947 (ipa_update_overall_fn_summary): Likewise.
948 (read_ipa_call_summary): Likewise.
949 (inline_read_section): Likewise.
950 (write_ipa_call_summary): Likewise.
951 (ipa_fn_summary_write): Likewise.
952 (ipa_free_fn_summary): Likewise.
953 * ipa-hsa.c (process_hsa_functions): Likewise.
954 (ipa_hsa_write_summary): Likewise.
955 (ipa_hsa_read_section): Likewise.
956 * ipa-icf.c (sem_function::merge): Likewise.
957 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
958 (do_estimate_edge_time): Likewise.
959 (estimate_size_after_inlining): Likewise.
960 (estimate_growth): Likewise.
961 (growth_likely_positive): Likewise.
962 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
963 (inline_call): Likewise.
964 * ipa-inline.c (caller_growth_limits): Likewise.
965 (can_inline_edge_p): Likewise.
966 (can_inline_edge_by_limits_p): Likewise.
967 (compute_uninlined_call_time): Likewise.
968 (compute_inlined_call_time): Likewise.
969 (want_inline_small_function_p): Likewise.
970 (edge_badness): Likewise.
971 (update_caller_keys): Likewise.
972 (update_callee_keys): Likewise.
973 (recursive_inlining): Likewise.
974 (inline_small_functions): Likewise.
975 (inline_to_all_callers_1): Likewise.
976 (dump_overall_stats): Likewise.
977 (early_inline_small_functions): Likewise.
978 (early_inliner): Likewise.
979 * ipa-inline.h (estimate_edge_growth): Likewise.
980 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
981 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
982 * ipa-prop.h (IPA_NODE_REF): Likewise.
983 (IPA_EDGE_REF): Likewise.
984 * ipa-pure-const.c (malloc_candidate_p): Likewise.
985 (propagate_malloc): Likewise.
986 * ipa-split.c (execute_split_functions): Likewise.
987 * symbol-summary.h: Rename get to get_create.
989 (get_create): Likewise.
990 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
992 2018-06-08 Martin Liska <mliska@suse.cz>
994 * symbol-summary.h (release): Move definition out of class
996 (symtab_removal): Likewise.
997 (symtab_duplication): Likewise.
999 2018-06-08 Martin Liska <mliska@suse.cz>
1001 * symbol-summary.h (function_summary): Move constructor
1002 implementation out of class declaration.
1003 (release): Likewise.
1004 (symtab_insertion): Likewise.
1005 (symtab_removal): Likewise.
1006 (symtab_duplication): Likewise.
1009 2018-06-08 Martin Liska <mliska@suse.cz>
1011 * Makefile.in: Remove support for MPX (macros, related functions,
1012 fields in cgraph_node, ...).
1013 * builtin-types.def (BT_BND): Likewise.
1014 (BT_FN_BND_CONST_PTR): Likewise.
1015 (BT_FN_CONST_PTR_BND): Likewise.
1016 (BT_FN_VOID_PTR_BND): Likewise.
1017 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
1018 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
1019 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
1020 (expand_builtin_mempcpy_with_bounds): Likewise.
1021 (expand_builtin_memset_with_bounds): Likewise.
1022 (expand_builtin_memset_args): Likewise.
1023 (std_expand_builtin_va_start): Likewise.
1024 (expand_builtin): Likewise.
1025 (expand_builtin_with_bounds): Likewise.
1026 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
1027 (DEF_LIB_BUILTIN_CHKP): Likewise.
1028 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
1029 (DEF_CHKP_BUILTIN): Likewise.
1030 (BUILT_IN_MEMCPY): Likewise.
1031 (BUILT_IN_MEMMOVE): Likewise.
1032 (BUILT_IN_MEMPCPY): Likewise.
1033 (BUILT_IN_MEMSET): Likewise.
1034 (BUILT_IN_STPCPY): Likewise.
1035 (BUILT_IN_STRCAT): Likewise.
1036 (BUILT_IN_STRCHR): Likewise.
1037 (BUILT_IN_STRCPY): Likewise.
1038 (BUILT_IN_STRLEN): Likewise.
1039 (BUILT_IN_MEMCPY_CHK): Likewise.
1040 (BUILT_IN_MEMMOVE_CHK): Likewise.
1041 (BUILT_IN_MEMPCPY_CHK): Likewise.
1042 (BUILT_IN_MEMSET_CHK): Likewise.
1043 (BUILT_IN_STPCPY_CHK): Likewise.
1044 (BUILT_IN_STRCAT_CHK): Likewise.
1045 (BUILT_IN_STRCPY_CHK): Likewise.
1046 * calls.c (store_bounds): Likewise.
1047 (emit_call_1): Likewise.
1048 (special_function_p): Likewise.
1049 (maybe_warn_nonstring_arg): Likewise.
1050 (initialize_argument_information): Likewise.
1051 (finalize_must_preallocate): Likewise.
1052 (compute_argument_addresses): Likewise.
1053 (expand_call): Likewise.
1054 * cfgexpand.c (expand_call_stmt): Likewise.
1055 (expand_return): Likewise.
1056 (expand_gimple_stmt_1): Likewise.
1057 (pass_expand::execute): Likewise.
1058 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
1059 (cgraph_node::remove): Likewise.
1060 (cgraph_node::dump): Likewise.
1061 (cgraph_node::verify_node): Likewise.
1062 * cgraph.h (chkp_function_instrumented_p): Likewise.
1063 (symtab_node::get_alias_target): Likewise.
1064 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
1065 (cgraph_local_p): Likewise.
1066 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
1067 (cgraph_edge::rebuild_references): Likewise.
1068 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
1069 (walk_polymorphic_call_targets): Likewise.
1070 (cgraph_node::expand_thunk): Likewise.
1071 (symbol_table::output_weakrefs): Likewise.
1072 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
1073 (ix86_handle_option): Likewise.
1074 * config/i386/constraints.md: Likewise.
1075 * config/i386/i386-builtin-types.def (BND): Likewise.
1079 * config/i386/i386-builtin.def (BDESC_END): Likewise.
1080 (BDESC_FIRST): Likewise.
1082 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
1083 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
1084 * config/i386/i386.c (enum reg_class): Likewise.
1085 (ix86_target_string): Likewise.
1086 (ix86_option_override_internal): Likewise.
1087 (ix86_conditional_register_usage): Likewise.
1088 (ix86_valid_target_attribute_inner_p): Likewise.
1089 (ix86_set_indirect_branch_type): Likewise.
1090 (ix86_set_current_function): Likewise.
1091 (ix86_function_arg_regno_p): Likewise.
1092 (init_cumulative_args): Likewise.
1093 (ix86_function_arg_advance): Likewise.
1094 (ix86_function_arg): Likewise.
1095 (ix86_pass_by_reference): Likewise.
1096 (ix86_function_value_regno_p): Likewise.
1097 (ix86_function_value_1): Likewise.
1098 (ix86_function_value_bounds): Likewise.
1099 (ix86_return_in_memory): Likewise.
1100 (ix86_setup_incoming_vararg_bounds): Likewise.
1101 (ix86_va_start): Likewise.
1102 (indirect_thunk_need_prefix): Likewise.
1103 (print_reg): Likewise.
1104 (ix86_print_operand): Likewise.
1105 (ix86_expand_call): Likewise.
1106 (ix86_output_function_return): Likewise.
1107 (reg_encoded_number): Likewise.
1108 (BDESC_VERIFYS): Likewise.
1109 (ix86_init_mpx_builtins): Likewise.
1110 (ix86_init_builtins): Likewise.
1111 (ix86_emit_cmove): Likewise.
1112 (ix86_emit_move_max): Likewise.
1113 (ix86_expand_builtin): Likewise.
1114 (ix86_builtin_mpx_function): Likewise.
1115 (ix86_get_arg_address_for_bt): Likewise.
1116 (ix86_load_bounds): Likewise.
1117 (ix86_store_bounds): Likewise.
1118 (ix86_load_returned_bounds): Likewise.
1119 (ix86_store_returned_bounds): Likewise.
1120 (ix86_class_likely_spilled_p): Likewise.
1121 (ix86_hard_regno_mode_ok): Likewise.
1122 (x86_order_regs_for_local_alloc): Likewise.
1123 (ix86_mitigate_rop): Likewise.
1124 (ix86_bnd_prefixed_insn_p): Likewise.
1125 (ix86_mpx_bound_mode): Likewise.
1126 (ix86_make_bounds_constant): Likewise.
1127 (ix86_initialize_bounds): Likewise.
1128 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
1129 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
1130 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
1131 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
1132 (TARGET_CHKP_BOUND_MODE): Likewise.
1133 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
1134 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
1135 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
1136 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
1137 * config/i386/i386.h (TARGET_MPX): Likewise.
1138 (TARGET_MPX_P): Likewise.
1139 (VALID_BND_REG_MODE): Likewise.
1140 (FIRST_BND_REG): Likewise.
1141 (LAST_BND_REG): Likewise.
1142 (enum reg_class): Likewise.
1143 (BND_REG_P): Likewise.
1144 (BND_REGNO_P): Likewise.
1145 (BNDmode): Likewise.
1146 (ADJUST_INSN_LENGTH): Likewise.
1147 * config/i386/i386.md: Likewise.
1148 * config/i386/i386.opt: Likewise.
1149 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
1150 (defined): Likewise.
1151 (LINK_MPX): Likewise.
1152 (MPX_SPEC): Likewise.
1153 (LIBMPX_SPEC): Likewise.
1154 (LIBMPXWRAPPERS_SPEC): Likewise.
1155 (CHKP_SPEC): Likewise.
1156 * config/i386/predicates.md: Likewise.
1157 * dbxout.c (dbxout_type): Likewise.
1158 * doc/extend.texi: Likewise.
1159 * doc/invoke.texi: Likewise.
1160 * doc/md.texi: Likewise.
1161 * doc/tm.texi: Likewise.
1162 * doc/tm.texi.in: Likewise.
1163 * dwarf2out.c (is_base_type): Likewise.
1164 (gen_formal_types_die): Likewise.
1165 (gen_subprogram_die): Likewise.
1166 (gen_type_die_with_usage): Likewise.
1167 (gen_decl_die): Likewise.
1168 (dwarf2out_late_global_decl): Likewise.
1169 * expr.c (expand_assignment): Likewise.
1170 (emit_storent_insn): Likewise.
1171 (store_expr_with_bounds): Likewise.
1172 (store_expr): Likewise.
1173 (expand_expr_real_1): Likewise.
1174 * expr.h (store_expr_with_bounds): Likewise.
1175 * function.c (use_register_for_decl): Likewise.
1176 (struct bounds_parm_data): Likewise.
1177 (assign_parms_augmented_arg_list): Likewise.
1178 (assign_parm_find_entry_rtl): Likewise.
1179 (assign_parm_is_stack_parm): Likewise.
1180 (assign_parm_load_bounds): Likewise.
1181 (assign_bounds): Likewise.
1182 (assign_parms): Likewise.
1183 (expand_function_start): Likewise.
1184 * gcc.c (CHKP_SPEC): Likewise.
1185 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
1186 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
1187 (wrestrict_dom_walker::check_call): Likewise.
1188 * gimple.c (gimple_build_call_from_tree): Likewise.
1189 * gimple.h (enum gf_mask): Likewise.
1190 (gimple_call_with_bounds_p): Likewise.
1191 (gimple_call_set_with_bounds): Likewise.
1192 * gimplify.c (gimplify_init_constructor): Likewise.
1193 * ipa-cp.c (initialize_node_lattices): Likewise.
1194 (propagate_constants_across_call): Likewise.
1195 (find_more_scalar_values_for_callers_subset): Likewise.
1196 * ipa-hsa.c (process_hsa_functions): Likewise.
1197 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
1198 * ipa-icf.c (sem_function::merge): Likewise.
1199 * ipa-inline.c (early_inliner): Likewise.
1200 * ipa-pure-const.c (warn_function_noreturn): Likewise.
1201 (warn_function_cold): Likewise.
1202 (propagate_pure_const): Likewise.
1203 * ipa-ref.h (enum GTY): Likewise.
1204 * ipa-split.c (find_retbnd): Likewise.
1205 (consider_split): Likewise.
1206 (split_function): Likewise.
1207 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
1208 * ipa.c (walk_polymorphic_call_targets): Likewise.
1209 (symbol_table::remove_unreachable_nodes): Likewise.
1210 (process_references): Likewise.
1211 (cgraph_build_static_cdtor_1): Likewise.
1212 * lto-cgraph.c (lto_output_node): Likewise.
1213 (output_refs): Likewise.
1214 (compute_ltrans_boundary): Likewise.
1215 (input_overwrite_node): Likewise.
1216 (input_node): Likewise.
1217 (input_cgraph_1): Likewise.
1218 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
1219 * passes.c (pass_manager::execute_early_local_passes): Likewise.
1220 (class pass_chkp_instrumentation_passes): Likewise.
1221 (make_pass_chkp_instrumentation_passes): Likewise.
1222 * passes.def: Likewise.
1223 * rtl.h (struct GTY): Likewise.
1224 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
1225 * stor-layout.c (layout_type): Likewise.
1226 * symtab.c: Likewise.
1227 * target.def: Likewise.
1228 * targhooks.c (default_chkp_bound_type): Likewise.
1229 (default_chkp_bound_mode): Likewise.
1230 (default_builtin_chkp_function): Likewise.
1231 (default_chkp_function_value_bounds): Likewise.
1232 (default_chkp_make_bounds_constant): Likewise.
1233 (default_chkp_initialize_bounds): Likewise.
1234 * targhooks.h (default_chkp_bound_type): Likewise.
1235 (default_chkp_bound_mode): Likewise.
1236 (default_builtin_chkp_function): Likewise.
1237 (default_chkp_function_value_bounds): Likewise.
1238 (default_chkp_make_bounds_constant): Likewise.
1239 (default_chkp_initialize_bounds): Likewise.
1240 * toplev.c (compile_file): Likewise.
1241 (process_options): Likewise.
1242 * tree-core.h (DEF_BUILTIN): Likewise.
1243 (DEF_BUILTIN_CHKP): Likewise.
1244 * tree-inline.c (declare_return_variable): Likewise.
1245 (remap_gimple_stmt): Likewise.
1246 (copy_bb): Likewise.
1247 (initialize_inlined_parameters): Likewise.
1248 (expand_call_inline): Likewise.
1249 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
1250 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
1251 (make_pass_ipa_chkp_produce_thunks): Likewise.
1252 (make_pass_chkp): Likewise.
1253 (make_pass_chkp_opt): Likewise.
1254 (make_pass_chkp_instrumentation_passes): Likewise.
1255 * tree-pretty-print.c (dump_generic_node): Likewise.
1256 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
1257 * tree-ssa-dce.c (propagate_necessity): Likewise.
1258 (eliminate_unnecessary_stmts): Likewise.
1259 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
1260 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
1261 * tree-ssa-sccvn.h: Likewise.
1262 * tree-ssa-strlen.c (get_string_length): Likewise.
1263 (valid_builtin_call): Likewise.
1264 (adjust_last_stmt): Likewise.
1265 (handle_builtin_strchr): Likewise.
1266 (handle_builtin_strcpy): Likewise.
1267 (handle_builtin_stxncpy): Likewise.
1268 (handle_builtin_memcpy): Likewise.
1269 (handle_builtin_strcat): Likewise.
1270 (strlen_check_and_optimize_stmt): Likewise.
1271 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
1272 * tree-streamer-in.c: Likewise.
1273 * tree-streamer.c (record_common_node): Likewise.
1274 * tree.c (tree_code_size): Likewise.
1275 (wide_int_to_tree_1): Likewise.
1276 (type_contains_placeholder_1): Likewise.
1277 (build_common_tree_nodes): Likewise.
1278 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
1279 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
1280 (POINTER_BOUNDS_P): Likewise.
1281 (BOUNDED_TYPE_P): Likewise.
1282 (BOUNDED_P): Likewise.
1283 (CALL_WITH_BOUNDS_P): Likewise.
1284 (pointer_bounds_type_node): Likewise.
1285 * value-prof.c (gimple_ic): Likewise.
1286 * var-tracking.c (vt_add_function_parameters): Likewise.
1287 * varasm.c (make_decl_rtl): Likewise.
1288 (assemble_start_function): Likewise.
1289 (output_constant): Likewise.
1290 (maybe_assemble_visibility): Likewise.
1291 * varpool.c (ctor_for_folding): Likewise.
1292 * chkp-builtins.def: Remove.
1293 * ipa-chkp.c: Remove.
1294 * ipa-chkp.h: Remove.
1295 * rtl-chkp.c: Remove.
1296 * rtl-chkp.h: Remove.
1297 * tree-chkp-opt.c: Remove.
1298 * tree-chkp.c: Remove.
1299 * tree-chkp.h: Remove.
1301 2018-06-07 Carl Love <cel@us.ibm.com>
1303 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
1304 vextract_fp_from_shortl): Add BE support.
1306 2018-06-07 Paul Koning <ni1d@arrl.net>
1308 * compare-elim.c (try_merge_compare): Don't merge compare if
1309 address contains a side effect.
1310 (try_eliminate_compare): Likewise.
1312 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
1314 * config.gcc: Support "tremont".
1315 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
1316 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1318 * config/i386/i386.c (m_TREMONT): Define.
1319 (processor_target_table): Add "tremont".
1320 (PTA_TREMONT): Define.
1321 (ix86_lea_outperforms): Add TARGET_TREMONT.
1322 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
1323 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
1324 and M_INTEL_GOLDMONT_PLUS.
1325 (fold_builtin_cpu): Add "tremont".
1326 (ix86_add_stmt_cost): Add TARGET_TREMONT.
1327 (ix86_option_override_internal): Add "tremont".
1328 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
1329 (processor_type): Add PROCESSOR_TREMONT.
1330 * config/i386/x86-tune.def: Add m_TREMONT.
1331 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
1333 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1335 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
1336 symbol defined for msp430i* devices to be lower case.
1338 gcc/testsuite/gcc.target/msp430/
1339 * msp430i-device-symbol.c: New test.
1340 * msp430f-device-symbol.c: New test.
1341 * msp430.h: New test header file.
1343 2018-06-07 Richard Biener <rguenther@suse.de>
1345 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
1346 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
1347 Properly wrap signed arithmetic if overflow wraps.
1349 2018-06-07 Jakub Jelinek <jakub@redhat.com>
1351 PR tree-optimization/69615
1352 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
1353 of a cast from a same precision integral SSA_NAME in a bb dominated
1354 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
1355 cast to utype if rhs2 has already a compatible type.
1357 2018-06-07 Richard Biener <rguenther@suse.de>
1359 PR tree-optimization/85935
1360 * graphite-scop-detection.c (find_params_in_bb): Analyze
1361 condition operands with respect to the correct loop. Assert
1362 the analysis doesn't fail.
1364 2018-06-04 Carl Love <cel@us.ibm.com>
1366 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
1367 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
1368 as it is slightly cheaper.
1369 (first_match_or_eos_index_<mode>):
1370 Calculate index using natural element order.
1371 (first_match_index_<mode>):
1372 Calculate index using natural element order.
1373 (first_match_or_eos_index_<mode>):
1374 Calculate index using natural order.
1375 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
1376 for BE and LE modes.
1377 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
1378 P9V_BUILTIN_VCLZLSBB_V16QI.
1379 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
1382 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
1384 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
1385 indentation and line wrap for many prototypes. Add missing
1386 @smallexample directives around block of prototypes for vec_xl and
1389 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
1391 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
1392 track if we pass or return IEEE 128-bit floating point.
1393 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
1394 C++ mangling that is compatible with GCC 8.1.
1395 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
1396 (init_cumulative_args): Note if we pass or return IEEE 128-bit
1397 floating point types.
1398 (rs6000_function_arg_advance_1): Likewise.
1399 (rs6000_mangle_type): Optionally generate mangled names that match
1400 what GCC 8.1 generated for IEEE 128-bit floating point types.
1401 (rs6000_globalize_decl_name): If we have an external function that
1402 passes or returns IEEE 128-bit types, generate a weak reference
1403 from the mangled name used in GCC 8.1 to the current mangled
1405 (rs6000_init_builtins): Make __ibm128 use the long double type if
1406 long double is IBM extended double. Make __float128 use the long
1407 double type if long double is IEEE 128-bit.
1410 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
1411 macro for __ibm128 built-in functions.
1412 (PACK_IF): Add __ibm128 pack/unpack functions.
1413 (UNPACK_IF): Likewise.
1414 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
1415 enable long double built-in functions if long double is IEEE
1416 128-bit floating point.
1417 (rs6000_invalid_builtin): Update long double built-in function
1419 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
1420 functions, adjust the built-in function to use the long double
1421 built-in function if __ibm128 and long double are the same type.
1422 * doc/extend.texi (PowerPC builtins): Update documention for
1423 __builtin_{,un}pack_longdouble. Add documentation for
1424 __builtin_{,un}pack_ibm128.
1426 2018-06-06 Jim Wilson <jimw@sifive.com>
1428 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
1429 (struct machine_function): New field interrupt_mode.
1430 (riscv_handle_type_attribute): New function. Add forward declaration.
1431 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
1432 (riscv_expand_epilogue): Check interrupt_mode field.
1433 (riscv_set_current_function): Check interrupt attribute args and
1434 set interrupt_mode field.
1435 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
1436 (riscv_sret, riscv_uret): New.
1437 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
1438 new arguments to interrupt attribute.
1440 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
1443 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
1444 Don't handle -mcpu=power8 if -mpower9-vector is also used.
1446 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
1448 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
1449 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
1450 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1451 several redundant entries.
1453 2018-06-06 David Malcolm <dmalcolm@redhat.com>
1455 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
1456 type from "rtx" to "rtx_insn *".
1457 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
1458 for local "call_insn", removing cast.
1459 (ix86_expand_call): Likewise, introducing a "call_insn" local.
1461 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
1463 PR tree-optimization/86066
1464 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
1465 for BIT_INSERT_EXPR stores.
1467 2018-06-06 Richard Biener <rguenther@suse.de>
1469 PR tree-optimization/86062
1470 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1471 component refs ontop
1472 of to be offsetted base.
1474 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1476 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
1477 to be static and remove check on interrupt attribute name.
1479 gcc/testsuite/gcc.target/msp430/
1480 * function-attributes-4.c: New test.
1481 * static-interrupts.c: New test.
1483 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
1485 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
1486 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
1488 2018-06-05 Steve Ellcey <sellcey@cavium.com>
1491 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
1493 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
1494 Remove second argument, change how error is called.
1495 (aarch64_layout_arg): Remove second argument from
1496 aarch64_err_no_fpadvsimd call.
1497 (aarch64_init_cumulative_args): Ditto.
1498 (aarch64_gimplify_va_arg_expr): Ditto.
1499 * config/aarch64/aarch64.md (mov<mode>): Ditto.
1501 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
1503 * config/i386/i386.md (simple_return_indirect_internal): New expander.
1504 (*simple_return_indirect_internal<mode>): Rename from
1505 simple_return_indirect_internal. Use W mode iterator.
1506 (rstorssp): New expander.
1507 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
1508 (clrssbsy): New expander.
1509 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
1511 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
1513 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
1515 (cmse_check_pointed_object): Likewise.
1517 2018-06-05 Martin Liska <mliska@suse.cz>
1519 PR gcov-profile/47618
1520 * doc/invoke.texi: Document how -fprofile-dir format
1523 2018-06-05 Richard Biener <rguenther@suse.de>
1525 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
1526 removal pretend DOM info isn't available so we do not update
1527 it and only remove edges, not dominated blocks. Actually free
1528 DOM info in case we removed something. Remove unreachable blocks.
1529 (mfb_keep_latches): Work with either DOM info or marked backedges.
1530 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
1531 first. Mark backedges if DOM info isn't available.
1532 (Re-)compute DOM info after cleanup_control_flow_pre.
1534 2018-06-05 Richard Biener <rguenther@suse.de>
1536 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
1537 (locus_discrim_hasher::hash): Adjust.
1538 (locus_discrim_hasher::equal): Likewise.
1539 (next_discriminator_for_locus): Work on line directly.
1540 (same_line_p): Pass in expanded locus1 as well.
1541 (assign_discriminators): Avoid redundant location expansions.
1543 2018-06-05 Richard Biener <rguenther@suse.de>
1545 PR tree-optimization/86046
1546 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
1547 if required after clearing TREE_ADDRESSABLE.
1549 2018-06-05 Richard Biener <rguenther@suse.de>
1551 PR tree-optimization/86047
1552 * tree-ssa-loop.c (for_each_index): Glob handling of all
1553 decls and constants and really handle all of them.
1555 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1558 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
1559 qualifier_void_pointer and qualifier_const_void_pointer.
1560 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
1561 (arm_init_builtins): Handle the above.
1562 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
1563 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
1564 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
1567 2018-06-05 Martin Liska <mliska@suse.cz>
1569 * auto-profile.c (read_autofdo_file): Do not use
1570 gcov_ctr_summary struct.
1571 (afdo_callsite_hot_enough_for_early_inline): Likewise.
1572 * coverage.c (struct counts_entry): Likewise.
1573 (read_counts_file): Read just single summary entry.
1574 (get_coverage_counts): Use gcov_summary struct.
1575 * coverage.h (get_coverage_counts): Likewise.
1576 * gcov-dump.c (dump_working_sets): Likewise.
1577 (tag_summary): Dump just single summary.
1578 * gcov-io.c (gcov_write_summary): Write just histogram
1580 (gcov_read_summary): Read just single summary.
1581 (compute_working_sets): Use gcov_summary struct.
1582 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
1583 of GCOV_COUNTERS_SUMMABLE.
1584 (GCOV_COUNTERS_SUMMABLE): Remove.
1585 (GCOV_FIRST_VALUE_COUNTER): Replace with
1586 GCOV_COUNTER_V_INTERVAL.
1587 (struct gcov_ctr_summary): Remove.
1588 (struct gcov_summary): Directly use fields of former
1590 (compute_working_sets): Use gcov_summary struct.
1591 * gcov.c (read_count_file): Do not use ctrs fields.
1592 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
1594 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
1596 * profile.c: Likewise.
1597 * profile.h: Likewise.
1599 2018-06-05 Martin Liska <mliska@suse.cz>
1601 PR gcov-profile/84846
1602 * gcov.c (output_lines): Print working directory only
1603 in intermediate format.
1605 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
1607 * config/s390/s390-builtin-types.def: Add void function type.
1608 * config/s390/s390-builtins.def: Use the function type for the
1611 2018-06-04 Jim Wilson <jimw@sifive.com>
1613 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
1615 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
1616 and maybe_eh_return. Change regno to unsigned int. Use new args to
1617 handle EH_RETURN_DATA_REGNO registers properly.
1618 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
1619 (riscv_expand_epilogue): Update comment. Change argument name and
1620 type. Update code to use new name and type. Pass new args to
1621 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
1623 * config/riscv/riscv.md (NORMAL_RETURN): New.
1624 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
1625 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
1626 (eh_return): Call gen_eh_return_internal and emit barrier.
1627 (eh_return_internal): Call riscv_expand_epilogue.
1629 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
1631 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
1632 bit_insertion field and declare can_be_merged_into method.
1633 (merged_store_group::can_be_merged_into): New method.
1634 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
1635 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
1636 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
1638 2018-06-04 Richard Biener <rguenther@suse.de>
1640 PR tree-optimization/85955
1641 * builtins.c (fold_builtin_sincos): Convert pointers to
1642 destination to appropriate type before dereferencing.
1644 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
1646 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
1648 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
1650 * expr.c (expand_expr_real_1): Force the operand into memory if
1651 its TYPE_MODE is BLKmode and if there is no integer mode for
1652 the number of bits being extracted.
1654 2018-06-04 Jakub Jelinek <jakub@redhat.com>
1658 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
1659 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
1661 2018-06-04 Richard Biener <rguenther@suse.de>
1663 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
1664 (cleanup_tree_cfg_noloop): ... single caller. Do
1665 start_recording_case_labels later.
1667 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
1669 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
1670 to _IMMINTRIN_H_INCLUDED.
1671 * config/i386/pconfigintrin.h: Ditto.
1672 * config/i386/waitpkgintrin.h: Ditto.
1673 * config/i386/immintrin.h: Add includes for sgxintrin.h,
1674 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
1675 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
1676 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
1677 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
1678 waitpkgintrin.h and cldemoteintrin.h.
1680 2018-06-04 Richard Biener <rguenther@suse.de>
1682 PR tree-optimization/86038
1683 * tracer.c (find_best_successor): Check probability for
1684 being initialized, bail out if not.
1686 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
1689 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
1690 of bits to ignore when comparing architectures.
1692 2018-06-04 Jakub Jelinek <jakub@redhat.com>
1694 PR tree-optimization/69615
1695 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
1696 maximum or minimum of the type, try to merge it also as if
1697 range1 is + [-, x - 1] or + [x + 1, -].
1700 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
1702 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
1704 PR tree-optimization/86034
1705 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
1706 the unsigned bitfield type in a bit insertion sequence if it does not
1707 have a larger precision than the bitfield size.
1708 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
1710 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
1712 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
1714 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
1716 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
1717 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
1718 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
1719 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
1721 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
1723 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
1724 Disable -fdelete-null-pointer-checks for ELF toolchain.
1726 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
1727 Kito Cheng <kito.cheng@gmail.com>
1729 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
1730 (nds32le-*-*, nds32be-*-*): Integrate checking process.
1731 (nds32*-*-*): Add glibc and uclibc conditions.
1732 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
1733 (TARGET_EXCEPT_UNWIND_INFO): Define.
1734 * config/nds32/elf.h: New file.
1735 * config/nds32/linux.h: New file.
1736 * config/nds32/nds32-elf.opt: New file.
1737 * config/nds32/nds32-linux.opt: New file.
1738 * config/nds32/nds32-fp-as-gp.c
1739 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
1740 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
1742 (nds32_asm_file_end): Ditto.
1743 (nds32_print_operand): Ditto.
1744 (nds32_insert_attributes): Ditto.
1745 (nds32_init_libfuncs): New function.
1746 (TARGET_HAVE_TLS): Define.
1747 (TARGET_INIT_LIBFUNCS): Define.
1748 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
1750 (TARGET_ELF): Apply different mcmodel setting.
1751 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
1752 been migrated into elf.h and linux.h files.
1753 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
1754 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
1755 (mcmodel): The content has been migrated into nds32-elf.opt and
1756 nds32-linux.opt files.
1757 * config/nds32/t-elf: New file.
1758 * config/nds32/t-linux: New file.
1760 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
1761 Shiva Chen <shiva0217@gmail.com>
1763 * config/nds32/constants.md (unspec_volatile_element): Add
1764 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
1765 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
1767 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
1768 (make_pass_nds32_fp_as_gp): Declare.
1769 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
1771 (nds32_asm_function_end_prologue): Remove unused asm output.
1772 (nds32_asm_function_begin_epilogue): Remove unused asm output.
1773 (nds32_asm_file_start): Output necessary fp_as_gp information.
1774 (nds32_option_override): Adjust register usage.
1775 (nds32_expand_prologue): Consider fp_as_gp situation.
1776 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
1777 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
1780 (simple_return): Ditto.
1781 (omit_fp_begin): Output special directive for fp_as_gp.
1782 (omit_fp_end): Output special directive for fp_as_gp.
1783 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
1784 mforbid-fp-as-gp): New options.
1786 2018-06-01 Mark Wielaard <mark@klomp.org>
1788 * dwarf2out.c (dwarf2out_finish): Remove generation of
1789 DW_AT_loclists_base.
1791 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
1793 * gimple-ssa-store-merging.c: Include gimple-fold.h.
1794 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
1795 (struct merged_store_group): Add bit_insertion field.
1796 (dump_char_array): Use standard hexadecimal format.
1797 (merged_store_group::merged_store_group): Set bit_insertion to false.
1798 (merged_store_group::apply_stores): Use optimal buffer size. Deal
1799 with BIT_INSERT_EXPR stores. Move up code updating the mask and
1800 also print the mask in the dump file.
1801 (pass_store_merging::gate): Minor tweak.
1802 (imm_store_chain_info::coalesce_immediate): Fix wrong association
1803 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
1804 stores with INTEGER_CST stores.
1805 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
1806 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
1807 and use it throughout. Generate bit insertion sequences if need be.
1808 (pass_store_merging::process_store): Remove redundant condition.
1809 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
1811 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
1813 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
1814 the 128-bit floating point types. Fix function comment.
1816 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1818 * config/aarch64/aarch64-simd.md
1819 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
1821 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
1824 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
1826 PR tree-optimization/85989
1827 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
1829 (backprop::intersect_uses): Check it when deciding whether this
1830 is a backedge reference.
1831 (backprop::process_block): Add each phi to m_visited_phis
1832 after visiting it, then clear it at the end.
1834 2018-06-01 Richard Biener <rguenther@suse.de>
1836 * tree-vectorizer.h (vect_dr_stmt): New function.
1837 (vect_get_load_cost): Adjust.
1838 (vect_get_store_cost): Likewise.
1839 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
1840 Use vect_dr_stmt instead of DR_SMTT.
1841 (vect_record_base_alignments): Likewise.
1842 (vect_calculate_target_alignment): Likewise.
1843 (vect_compute_data_ref_alignment): Likewise and make static.
1844 (vect_update_misalignment_for_peel): Likewise.
1845 (vect_verify_datarefs_alignment): Likewise.
1846 (vector_alignment_reachable_p): Likewise.
1847 (vect_get_data_access_cost): Likewise. Pass down
1848 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
1849 (vect_get_peeling_costs_all_drs): Likewise.
1850 (vect_peeling_hash_get_lowest_cost): Likewise.
1851 (vect_enhance_data_refs_alignment): Likewise.
1852 (vect_find_same_alignment_drs): Likewise.
1853 (vect_analyze_data_refs_alignment): Likewise.
1854 (vect_analyze_group_access_1): Likewise.
1855 (vect_analyze_group_access): Likewise.
1856 (vect_analyze_data_ref_access): Likewise.
1857 (vect_analyze_data_ref_accesses): Likewise.
1858 (vect_vfa_segment_size): Likewise.
1859 (vect_small_gap_p): Likewise.
1860 (vectorizable_with_step_bound_p): Likewise.
1861 (vect_prune_runtime_alias_test_list): Likewise.
1862 (vect_analyze_data_refs): Likewise.
1863 (vect_supportable_dr_alignment): Likewise.
1864 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
1865 (vect_gen_prolog_loop_niters): Likewise.
1866 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
1867 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
1869 (vect_recog_mask_conversion_pattern): Likewise.
1870 (vect_try_gather_scatter_pattern): Likewise.
1871 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
1872 to vect_get_store_cost.
1873 (vect_get_store_cost): Get stmt_info instead of DR.
1874 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
1875 (vect_get_load_cost): Get stmt_info instead of DR.
1877 2018-06-01 Richard Biener <rguenther@suse.de>
1880 * gimple-fold.c (var_decl_component_p): Also allow offsetted
1881 vars wrapped in MEM_REFs.
1883 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
1885 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
1886 Fix subreg tests so that we only return a choice between
1887 GENERAL_REGS and FP_REGS if the original classes included both.
1889 2018-06-01 Richard Biener <rguenther@suse.de>
1892 * tree-ssa-structalias.c (get_function_part_constraint):
1893 Handle NULL fi->decl.
1894 (find_func_aliases_for_call): Properly handle indirect
1895 fi from direct call.
1896 (find_func_clobbers): Likewise.
1897 (ipa_pta_execute): Likewise.
1898 (create_variable_info_for): For functions that are ifunc_resolver
1899 resolve to a varinfo that contains the result of the resolver call.
1900 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
1903 2018-05-31 Michael Collison <michael.collison@arm.com>
1905 * config/aarch64/aarch64.md:
1906 (*fix_to_zero_extenddfdi2): New pattern.
1907 * gcc.target/aarch64/fix_extend1.c: New testcase.
1909 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
1913 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
1914 and BUILT_IN_STRNCMP_EQ.
1915 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
1916 BUILT_IN_STRNCMP_EQ.
1917 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
1918 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
1919 (gimple_fold_builtin): Likewise.
1920 * tree-ssa-strlen.c (compute_string_length): New function.
1921 (determine_min_obsize): New function.
1922 (handle_builtin_string_cmp): New function to handle calls to
1923 string compare functions.
1924 (strlen_optimize_stmt): Add handling to builtin string compare
1926 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
1927 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
1928 * tree.c (build_common_builtin_nodes): Add new defines of
1929 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
1931 2018-05-31 Jakub Jelinek <jakub@redhat.com>
1934 * bb-reorder.c (pass_partition_blocks::gate): Return false for
1935 functions with naked attribute.
1937 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
1939 * config/i386/sse.md (avx_vec_concat<mode>):
1940 Substitute concat_tg_mode mode attribute with xtg_mode.
1941 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
1942 (concat_tg_mode): Remove mode attribute.
1944 2018-05-31 Martin Sebor <msebor@redhat.com>
1947 * calls.c (alloc_max_size): Correct a logic error/typo.
1948 Treat excessive arguments as infinite. Warn for invalid arguments.
1949 * doc/invoke.texi (-Walloc-size-larger-than): Update.
1951 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
1954 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
1955 and movx for Haswell.
1957 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
1958 Cesar Philippidis <cesar@codesourcery.com>
1961 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
1962 when emitting error on private/firstprivate reductions.
1963 * omp-low.c (lower_omp_target): Avoid reference-type processing
1964 on pointers for firstprivate clause.
1966 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
1968 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
1971 * config/aarch64/aarch64-simd.md
1972 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
1973 (aarch64_ld1_x3_<mode>): Likewise
1974 (aarch64_st1x2<VALLDIF:mode>): Likewise
1975 (aarch64_st1_x2_<mode>): Likewise
1976 (aarch64_st1x3<VALLDIF:mode>): Likewise
1977 (aarch64_st1_x3_<mode>): Likewise
1978 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
1979 (vld1_s8_x3): Likewise.
1980 (vld1_u16_x3): Likewise.
1981 (vld1_s16_x3): Likewise.
1982 (vld1_u32_x3): Likewise.
1983 (vld1_s32_x3): Likewise.
1984 (vld1_u64_x3): Likewise.
1985 (vld1_s64_x3): Likewise.
1986 (vld1_f16_x3): Likewise.
1987 (vld1_f32_x3): Likewise.
1988 (vld1_f64_x3): Likewise.
1989 (vld1_p8_x3): Likewise.
1990 (vld1_p16_x3): Likewise.
1991 (vld1_p64_x3): Likewise.
1992 (vld1q_u8_x3): Likewise.
1993 (vld1q_s8_x3): Likewise.
1994 (vld1q_u16_x3): Likewise.
1995 (vld1q_s16_x3): Likewise.
1996 (vld1q_u32_x3): Likewise.
1997 (vld1q_s32_x3): Likewise.
1998 (vld1q_u64_x3): Likewise.
1999 (vld1q_s64_x3): Likewise.
2000 (vld1q_f16_x3): Likewise.
2001 (vld1q_f32_x3): Likewise.
2002 (vld1q_f64_x3): Likewise.
2003 (vld1q_p8_x3): Likewise.
2004 (vld1q_p16_x3): Likewise.
2005 (vld1q_p64_x3): Likewise.
2006 (vst1_s64_x2): Likewise.
2007 (vst1_u64_x2): Likewise.
2008 (vst1_f64_x2): Likewise.
2009 (vst1_s8_x2): Likewise.
2010 (vst1_p8_x2): Likewise.
2011 (vst1_s16_x2): Likewise.
2012 (vst1_p16_x2): Likewise.
2013 (vst1_s32_x2): Likewise.
2014 (vst1_u8_x2): Likewise.
2015 (vst1_u16_x2): Likewise.
2016 (vst1_u32_x2): Likewise.
2017 (vst1_f16_x2): Likewise.
2018 (vst1_f32_x2): Likewise.
2019 (vst1_p64_x2): Likewise.
2020 (vst1q_s8_x2): Likewise.
2021 (vst1q_p8_x2): Likewise.
2022 (vst1q_s16_x2): Likewise.
2023 (vst1q_p16_x2): Likewise.
2024 (vst1q_s32_x2): Likewise.
2025 (vst1q_s64_x2): Likewise.
2026 (vst1q_u8_x2): Likewise.
2027 (vst1q_u16_x2): Likewise.
2028 (vst1q_u32_x2): Likewise.
2029 (vst1q_u64_x2): Likewise.
2030 (vst1q_f16_x2): Likewise.
2031 (vst1q_f32_x2): Likewise.
2032 (vst1q_f64_x2): Likewise.
2033 (vst1q_p64_x2): Likewise.
2034 (vst1_s64_x3): Likewise.
2035 (vst1_u64_x3): Likewise.
2036 (vst1_f64_x3): Likewise.
2037 (vst1_s8_x3): Likewise.
2038 (vst1_p8_x3): Likewise.
2039 (vst1_s16_x3): Likewise.
2040 (vst1_p16_x3): Likewise.
2041 (vst1_s32_x3): Likewise.
2042 (vst1_u8_x3): Likewise.
2043 (vst1_u16_x3): Likewise.
2044 (vst1_u32_x3): Likewise.
2045 (vst1_f16_x3): Likewise.
2046 (vst1_f32_x3): Likewise.
2047 (vst1_p64_x3): Likewise.
2048 (vst1q_s8_x3): Likewise.
2049 (vst1q_p8_x3): Likewise.
2050 (vst1q_s16_x3): Likewise.
2051 (vst1q_p16_x3): Likewise.
2052 (vst1q_s32_x3): Likewise.
2053 (vst1q_s64_x3): Likewise.
2054 (vst1q_u8_x3): Likewise.
2055 (vst1q_u16_x3): Likewise.
2056 (vst1q_u32_x3): Likewise.
2057 (vst1q_u64_x3): Likewise.
2058 (vst1q_f16_x3): Likewise.
2059 (vst1q_f32_x3): Likewise.
2060 (vst1q_f64_x3): Likewise.
2061 (vst1q_p64_x3): Likewise.
2063 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2065 * config/msp430/msp430.c (msp430_output_labelref): Prepend
2066 user_label_prefix to name.
2068 * tree-core.h: Update comment about the format of NAME string
2069 passed to handler in attribute_spec.
2071 * config/msp430/msp430.md: Remove erroneous subreg expression from
2072 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
2073 zero_extend{q,h}isi2.
2075 2018-05-30 Borislav Petkov <bp@suse.de>
2077 * doc/extend.texi: Document some architecture specific
2078 constraints and sort entries.
2080 2018-05-30 Martin Sebor <msebor@redhat.com>
2083 * builtins.c (expand_builtin_stpcpy_1): New function.
2084 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
2085 only if the former succeeds.
2087 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2089 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
2092 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2094 * doc/invoke.texi (-flinker-output): Document
2096 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2098 * passes.c (ipa_write_summaries): Only modify statements if body
2100 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
2101 incrementally linking.
2102 (ipa_passes): Likewise.
2103 * lto-cgraph.c (lto_output_node): When incrementally linking do not
2104 pass down resolution info.
2105 * common.opt (flag_incremental_link): Update info.
2106 * gcc.c (plugin specs): Turn flinker-output=* to
2107 -plugin-opt=-linker-output-known
2108 * toplev.c (compile_file): Also cut compilation when doing incremental
2110 * flag-types. (enum lto_partition_model): Add
2111 LTO_LINKER_OUTPUT_NOLTOREL.
2112 (invoke.texi): Add -flinker-output docs.
2113 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
2114 link same way as WPA; do not stream in dead initializers.
2116 * dwarf2out.c (dwarf2out_die_ref_for_decl,
2117 darf2out_register_external_decl): Support incremental link.
2119 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2121 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
2123 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2125 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
2126 it down to simple_object_copy_lto_debug_sections.
2127 (run_gcc): Determine incremental LTO link time and configure
2128 lto1 into non-wpa mode, disable renaming of debug sections.
2130 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
2132 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
2133 descriptions of various incorrectly documented functions.
2135 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
2138 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
2139 address check not strict.
2141 2018-05-30 Richard Biener <rguenther@suse.de>
2143 PR tree-optimization/85964
2144 * tracer.c (better_p): Drop initialized count check, we only
2145 call the function with initialized counts now.
2146 (find_best_successor): Do find a best edge if one
2147 has uninitialized count.
2148 (find_best_predecessor): Likewise. Do BB frequency check only
2149 if count is initialized.
2151 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
2153 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
2154 (aarch64_ldrstr_offset_compare): New.
2155 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
2156 load/store orderings.
2157 (aarch64_gen_adjusted_ldpstp): Likewise.
2159 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
2161 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
2162 Check for subset of GENERAL_REGS and FP_REGS.
2163 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
2166 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
2168 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
2169 and wi::to_poly_offset. Add the current offset and then check
2170 whether the sum fits, rather than using an unchecked addition of
2171 a checked term. Check for a shwi rather than a uhwi.
2172 * expr.c (get_bit_range): Use tree_to_poly_uint64.
2173 (store_constructor): Use poly_int_tree_p.
2174 (expand_expr_real_1): Likewise.
2175 * function.c (assign_temp): Likewise.
2176 * fold-const.c (const_binop): Use poly_int_tree_p and
2178 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
2180 * ipa-icf-gimple.c (func_checker::compare_operand): Use
2181 to_poly_offset for MEM offsets.
2182 * ipa-icf.c (sem_variable::equals): Likewise.
2183 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
2184 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
2185 wi::to_poly_offset for BIT_FIELD_REF offsets.
2186 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
2188 * var-tracking.c (emit_note_insn_var_location): Use
2189 tree_to_poly_uint64.
2191 2018-05-29 Jim Wilson <jimw@sifive.com>
2193 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
2195 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
2198 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
2199 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
2201 (sse4_1_round<mode>2): Use nonimmediate_operand
2202 for operand 1 predicate.
2204 2018-05-29 Martin Sebor <msebor@redhat.com>
2205 Richard Biener <rguenther@suse.de>
2208 * calls.c (get_size_range): Call determine_value_range instead
2209 of get_value_range..
2210 * tree-vrp.h (determine_value_range): Declared new function.
2211 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
2213 2018-05-29 Richard Biener <rguenther@suse.de>
2215 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
2216 sure to use non-pattern stmts for get_earlier_stmt arguments.
2217 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
2218 called on pattern stmts.
2219 (get_later_stmt): Likewise.
2221 2018-05-29 Martin Liska <mliska@suse.cz>
2223 PR gcov-profile/85759
2224 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
2227 2018-05-29 Jakub Jelinek <jakub@redhat.com>
2229 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
2231 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
2235 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
2236 VEC_PACK_FLOAT_EXPR): New tree codes.
2237 * tree-pretty-print.c (op_code_prio): Handle
2238 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
2239 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
2240 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
2241 * tree-inline.c (estimate_operator_cost): Likewise.
2242 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
2243 * fold-const.c (const_binop): Likewise.
2244 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
2245 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
2246 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
2247 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
2248 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
2249 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
2250 * expr.c (expand_expr_real_2): Likewise.
2251 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
2252 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
2253 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
2255 * optabs.c (expand_widen_pattern_expr): For
2256 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
2257 sign from result type rather than operand's type.
2258 (expand_binop_directly): For vec_packu_float_optab and
2259 vec_packs_float_optab allow result type to be different from operand's
2261 * optabs-tree.c (optab_for_tree_code): Handle
2262 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
2263 VEC_PACK_FLOAT_EXPR. Formatting fixes.
2264 * tree-vect-generic.c (expand_vector_operations_1): Handle
2265 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
2266 VEC_PACK_FLOAT_EXPR.
2267 * tree-vect-stmts.c (supportable_widening_operation): Handle
2269 (supportable_narrowing_operation): Handle FLOAT_EXPR.
2270 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
2271 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
2272 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
2273 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
2275 (vec_pack<floatprefix>_float_<mode>): New expander.
2276 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
2278 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
2279 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
2280 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
2281 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
2282 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
2284 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
2285 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
2286 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
2287 VEC_PACK_FLOAT_EXPR): Document.
2289 2018-05-29 Richard Biener <rguenther@suse.de>
2291 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
2293 (stmt_vec_info_vec): Make pointer.
2294 (init_stmt_vec_info_vec): Remove.
2295 (free_stmt_vec_info_vec): Likewise.
2296 (set_stmt_vec_info_vec): New function.
2297 (free_stmt_vec_infos): Likewise.
2298 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
2299 (set_vinfo_for_stmt): Likewise.
2300 (get_earlier_stmt): Likewise.
2301 (get_later_stmt): Likewise.
2302 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
2303 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
2304 (vec_info::~vec_info): Free stmt_vec_infos.
2305 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
2306 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
2307 (pass_slp_vectorize::execute): Likewise.
2308 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
2309 (free_stmt_vec_info_vec): Likewise.
2310 (set_stmt_vec_info_vec): New function.
2311 (free_stmt_vec_infos): Likewise.
2312 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
2313 the global stmt_vec_info_vec.
2314 * tree-parloops.c (gather_scalar_reductions): Use
2315 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
2318 2018-05-29 Richard Biener <rguenther@suse.de>
2320 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
2322 2018-05-29 Martin Liska <mliska@suse.cz>
2323 David Malcolm <dmalcolm@redhat.com>
2325 * vec.c (test_reverse): New.
2326 (vec_c_tests): Add new test.
2327 * vec.h (vl_ptr>::reverse): New function.
2329 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
2331 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
2333 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
2336 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
2338 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
2340 2018-05-28 Richard Biener <rguenther@suse.de>
2342 PR tree-optimization/85933
2343 * tree-vect-data-refs.c (vect_record_base_alignments): Only
2344 look at stmts marked as vectorizable.
2346 2018-05-28 Richard Biener <rguenther@suse.de>
2348 PR tree-optimization/85934
2349 * tree-vect-generic.c (expand_vector_operations_1): Hoist
2350 vector boolean check before scalar optimization.
2352 2018-05-28 Jakub Jelinek <jakub@redhat.com>
2354 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
2357 2018-05-28 Mark Wielaard <mark@klomp.org>
2359 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
2360 if it is an expression containing a minus sign.
2362 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
2364 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
2366 2018-05-27 Paul Koning <ni1d@arrl.net>
2368 * config/pdp11/pdp11.md (truncsihi2): Remove.
2370 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
2371 Chung-Ju Wu <jasonwucj@gmail.com>
2373 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
2375 (unaligned_store_dw): Ditto.
2376 * config/nds32/nds32-memory-manipulation.c
2377 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
2378 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
2379 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
2380 (emit_setmem_word_loop): Rename to ...
2381 (emit_setmem_doubleword_loop): ... this.
2382 (nds32_gen_dup_4_byte_to_word_value): New function.
2383 (nds32_gen_dup_8_byte_to_double_word_value): New function.
2384 (nds32_expand_setmem_loop): Refine implementation.
2385 (nds32_expand_setmem_loop_v3m): Ditto.
2386 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
2389 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
2391 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
2393 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
2395 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
2396 (nds32_init_machine_status): Initialize machine->attr_naked_p and
2397 machine->attr_no_prologue_p.
2398 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
2399 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
2400 (nds32_expand_epilogue): Consider attr_naked_p.
2401 (nds32_expand_epilogue_v3pop): Likewise.
2402 (nds32_can_use_return_insn): Likewise.
2403 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
2404 attr_no_prologue_p fields.
2405 * config/nds32/nds32.opt (mret-in-naked-func): New option.
2407 2018-05-27 Jakub Jelinek <jakub@redhat.com>
2410 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
2412 * config/i386/sse.md
2413 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
2415 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
2417 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
2419 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
2421 (*<floatsuffix>floatv2div2sf2): Rename to ...
2422 (*float<floatunssuffix>v2div2sf2): ... this.
2423 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
2424 (float<floatunssuffix>v2div2sf2_mask): ... this.
2425 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
2426 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
2427 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
2429 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
2431 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
2433 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
2435 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
2437 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
2439 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
2440 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
2441 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
2442 gen_ufix_truncv8dfv8si2.
2443 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
2444 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
2445 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
2446 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
2447 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
2448 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
2449 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
2450 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
2452 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
2456 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
2458 2018-05-25 Jim Wilson <jimw@sifive.com>
2460 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
2461 * config/riscv/riscv.c (struct machine_function): Add
2462 interrupt_handler_p and attribute_checked_p fields.
2463 (riscv_attribute_table): Add interrupt.
2464 (riscv_interrupt_type_p): New.
2465 (riscv_save_reg_p): Save extra regs for interrupt handler.
2466 (riscv_use_save_libcall): Return false for interrupt handler.
2467 (riscv_first_stack_step): Add forward declaration.
2468 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
2469 for interrupt handler with large frame. Use it for saved reg list.
2470 (riscv_expand_prologue): Move flag_stack_usage_info support to
2471 eliminate duplication.
2472 (riscv_expand_epilogue): Generate mret for interrupt handler.
2473 (riscv_epilogue_uses): New.
2474 (riscv_can_use_return_insn): Return false for interrupt handler.
2475 (riscv_function_ok_for_sibcall): Likewise.
2476 (riscv_set_current_function): Add interrupt handler support.
2477 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
2478 * config/riscv/riscv.md (UNSPECV_MRET): New.
2480 (riscv_frflags, riscv_fsflags): Use tab after opcode.
2482 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
2484 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
2486 PR tree-optimization/85712
2487 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
2488 this candidate has already been replaced in-situ by a copy.
2490 2018-05-25 Jason Merrill <jason@redhat.com>
2492 PR c++/80485 - inline function non-zero address.
2493 * symtab.c (nonzero_address): Check DECL_COMDAT.
2495 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
2498 * config/alpha/alpha.md (ashlsi3): New insn pattern.
2499 (*ashlsi_se): Rename from *ashldi_se. Define as sign
2500 extension of SImode operation. Use const123_operand predicate.
2501 (*saddsi_1): Remove.
2502 (*saddl_se_1): Ditto.
2504 (*ssubl_se_1): Ditto.
2505 * config/alpha/predicates.md (const123_operand): New predicate.
2506 * config/alpha/constraints.md (P): Use IN_RANGE.
2508 2018-05-25 Richard Biener <rguenther@suse.de>
2510 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
2512 (ref_maybe_used_by_stmt_p): Likewise.
2513 (stmt_may_clobber_ref_p): Likewise.
2514 (stmt_may_clobber_ref_p_1): Likewise.
2515 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
2517 (ref_maybe_used_by_stmt_p): Likewise.
2518 (stmt_may_clobber_ref_p): Likewise.
2519 (stmt_may_clobber_ref_p_1): Likewise.
2520 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
2521 the alias oracle to disambiguate DRs with stmts DR analysis
2523 (vect_analyze_data_refs): Do not give up on not analyzable
2524 DRs for BB vectorization. Remove code truncating the dataref
2527 2018-05-25 Jakub Jelinek <jakub@redhat.com>
2530 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
2531 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
2532 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
2534 2018-05-25 Richard Biener <rguenther@suse.de>
2536 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
2537 function, combining stmt data ref gathering and fatal analysis
2539 (vect_analyze_data_refs): Remove now redudnant code and simplify.
2540 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
2541 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
2542 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
2543 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
2545 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2547 PR tree-optimization/85720
2548 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
2549 SCC if all partitions are builtins.
2550 (version_loop_by_alias_check): New parameter. Generate cancelable
2551 runtime alias check if all partitions are builtins.
2552 (distribute_loop): Update call to above function.
2554 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2556 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
2557 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
2558 (parm_default_def_partition_arg): Ditto.
2559 (set_parm_default_def_partition): Ditto.
2560 (get_parm_default_def_partitions): Ditto and make it static.
2561 (get_undefined_value_partitions): Ditto and make it static.
2562 (remove_ssa_form): Refactor call to init_var_map here.
2563 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
2564 computation for loop region.
2565 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
2566 (register_default_def): Delete.
2567 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
2568 (parm_default_def_partition_arg): Ditto.
2569 (set_parm_default_def_partition): Ditto.
2570 (get_parm_default_def_partitions): Ditto and make it static.
2571 (get_undefined_value_partitions): Ditto and make it static.
2572 (coalesce_with_default, coalesce_with_default): Update comment.
2573 (create_coalesce_list_for_region): New func factored out from
2574 create_outofssa_var_map.
2575 (populate_coalesce_list_for_outofssa): New func factored out from
2576 create_outofssa_var_map and coalesce_ssa_name.
2577 (create_outofssa_var_map): Delete.
2578 (coalesce_ssa_name): Refactor to support live range computation.
2579 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
2580 (get_parm_default_def_partitions): Delete.
2581 (get_undefined_value_partitions): Ditto.
2582 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
2583 computation for loop region.
2584 (new_tree_live_info, loe_visit_block): Ditto.
2585 (live_worklist, set_var_live_on_entry): Ditto.
2586 (calculate_live_on_exit, verify_live_on_entry): Ditto.
2587 * tree-ssa-live.h (struct _var_map): New fields.
2588 (init_var_map): Change decl.
2589 (region_contains_p): New.
2591 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2593 * tree-ssa-live.h (live_merge_and_clear): Delete.
2595 2018-05-25 Richard Biener <rguenther@suse.de>
2598 * tree-dump.c (dequeue_and_dump): Remove access to removed
2599 operand 2 of a SWITCH_EXPR.
2601 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2603 * doc/sourcebuild.texi (vect_double_cond_arith): Include
2604 multiplication and division.
2605 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
2606 (cond_udiv@var{m}, cond_umod@var{m}): Document.
2607 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
2608 (cond_udiv_optab, cond_umod_optab): New optabs.
2609 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
2610 (IFN_COND_RDIV): New internal functions.
2611 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
2612 TRUNC_MOD_EXPR and RDIV_EXPR.
2613 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
2614 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
2616 (SVE_INT_BINARY): Include mult.
2617 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
2618 (optab, sve_int_op): Handle mult.
2619 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
2621 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
2622 for SVE_INT_BINARY_SD.
2624 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2626 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
2627 (optab, sve_int_op): Handle div and udiv.
2628 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
2629 for SVE_INT_BINARY_SD.
2630 (*<optab><mode>3): New insn for the same.
2632 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2634 * tree-vect-patterns.c: Include predict.h.
2635 (vect_recog_divmod_pattern): Restrict check for division support
2636 to when optimizing for size.
2638 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2640 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
2641 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
2642 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
2643 (gimple_match_op::set_op): Likewise.
2644 (gimple_resimplify4): Declare.
2645 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
2646 (expr::gen_transform): Likewise.
2647 (decision_tree::gen): Generate a simplification routine for 4 operands.
2648 * gimple-match-head.c (gimple_simplify): Add an overload for
2649 4 operands. In the top-level function, handle up to 4 call
2650 arguments and call gimple_resimplify4.
2651 (gimple_resimplify4): New function.
2652 (build_call_internal): Pass a fourth operand.
2653 (maybe_push_to_seq): Likewise.
2654 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
2655 Fold VEC_COND_EXPRs of an operation and a default value into
2656 an IFN_COND_* function if possible.
2657 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
2659 (SVE_COND_FP_BINARY): Include them.
2660 (optab, sve_fp_op): Handle them.
2661 (SVE_INT_BINARY_REV): New code iterator.
2662 (SVE_COND_FP_BINARY_REV): New int iterator.
2663 (commutative): New int attribute.
2664 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
2666 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
2668 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
2669 (*cond_<optab><mode>): New patterns for reversed operands.
2671 2018-05-25 Richard Biener <rguenther@suse.de>
2673 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
2674 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
2675 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
2676 (STMT_VINFO_GROUPED_ACCESS): Adjust.
2677 * tree-vect-data-refs.c (everywhere): Adjust users.
2678 * tree-vect-loop.c (everywhere): Likewise.
2679 * tree-vect-slp.c (everywhere): Likewise.
2680 * tree-vect-stmts.c (everywhere): Likewise.
2681 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
2683 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2685 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
2687 (gcc_cv_as_section_exclude): ... this.
2688 Try Solaris as #exclude syntax.
2689 * configure: Regenerate.
2690 * config.in: Regenerate.
2691 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
2693 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
2694 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
2696 * varasm.c (default_elf_asm_named_section): Don't check if
2697 HAVE_GAS_SECTION_EXCLUDE is defined.
2699 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2701 * doc/md.texi: Update the documentation of the cond_* optabs
2702 to mention the new final operand. Fix GET_MODE_NUNITS call.
2703 Describe the scalar case too.
2704 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
2705 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
2707 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
2708 (get_conditional_internal_fn): Update comment.
2709 * tree-vect-loop.c (vectorizable_reduction): Pass the original
2710 accumulator value as a final argument to conditional functions.
2711 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
2712 a define_expand and add an "else" operand. Assert for now that
2713 the else operand is equal to operand 2. Use SVE_INT_BINARY and
2714 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
2715 (*cond_<optab><mode>): New patterns.
2716 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
2717 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
2718 (UNSPEC_COND_EOR): Delete.
2719 (optab): Remove associated mappings.
2720 (SVE_INT_BINARY): New code iterator.
2721 (sve_int_op): Remove int attribute and add "minus" to the code
2723 (SVE_COND_INT_OP): Delete.
2724 (SVE_COND_FP_OP): Rename to...
2725 (SVE_COND_FP_BINARY): ...this.
2727 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2729 * optabs.c (can_reuse_operands_p): New function.
2730 (maybe_legitimize_operands): Try to reuse the results for
2733 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
2735 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
2736 Add {q} suffix to insn mnemonic.
2738 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2740 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
2741 (msp430_warn_func_return): New.
2743 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
2745 * fold-const.c (tree_nonzero_bits): New function.
2746 * fold-const.h (tree_nonzero_bits): Likewise.
2747 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
2748 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
2750 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
2754 * varasm.c (assemble_alias): Check ifunc_resolver only on
2757 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
2760 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
2761 when memory input operand is handled.
2763 2018-05-24 Luis Machado <luis.machado@linaro.org>
2765 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
2767 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
2769 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
2771 * match.pd: Delay FMA folds until after vectorization.
2773 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
2776 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
2777 address check not strict.
2779 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
2781 * gimple-match.h (gimple_match_op): New class.
2782 (mprts_hook): Replace parameters with a gimple_match_op *.
2783 (maybe_build_generic_op): Likewise.
2784 (gimple_simplified_result_is_gimple_val): Replace parameters with
2785 a const gimple_match_op *.
2786 (gimple_simplify): Replace code_helper * and tree * parameters with
2787 a gimple_match_op * parameter.
2788 (gimple_resimplify1): Replace code_helper *, tree and tree *
2789 parameters with a gimple_match_op * parameter.
2790 (gimple_resimplify2): Likewise.
2791 (gimple_resimplify3): Likewise.
2792 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
2793 parameters with a gimple_match_op * parameter.
2794 * gimple-match-head.c (gimple_simplify): Change prototypes of
2795 auto-generated functions to take a gimple_match_op * instead of
2796 separate code_helper * and tree * parameters. Make the same
2797 change in the top-level overload and update calls to the
2798 gimple_resimplify routines. Update calls to the auto-generated
2799 functions and to maybe_push_res_to_seq in the publicly-facing
2800 operation-specific gimple_simplify overloads.
2801 (gimple_match_op::MAX_NUM_OPS): Define.
2802 (gimple_resimplify1): Replace rcode and ops with a single res_op
2803 parameter. Update call to gimple_simplify.
2804 (gimple_resimplify2): Likewise.
2805 (gimple_resimplify3): Likewise.
2806 (mprts_hook): Replace parameters with a gimple_match_op *.
2807 (maybe_build_generic_op): Likewise.
2808 (build_call_internal): Replace type, nargs and ops with
2809 a gimple_match_op *.
2810 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
2811 with a single gimple_match_op *. Update calls to mprts_hook,
2812 build_call_internal and gimple_simplified_result_is_gimple_val.
2813 Factor out code that is common to the tree_code and combined_fn cases.
2814 * genmatch.c (expr::gen_transform): Replace tem_code and
2815 tem_ops with a gimple_match_op called tem_op. Update calls
2816 to the gimple_resimplify functions and maybe_push_res_to_seq.
2817 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
2818 res_ops. Update call to the gimple_resimplify functions.
2819 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
2820 (decision_tree::gen): Make the functions take a gimple_match_op *
2821 called res_op instead of separate res_code and res_ops parameters.
2822 Update call accordingly.
2823 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
2824 and ops with a single res_op parameter. Update calls to
2825 maybe_build_generic_op and maybe_push_res_to_seq.
2826 (fold_stmt_1): Update calls to gimple_simplify and
2827 replace_stmt_with_simplification.
2828 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
2829 and gimple_simplified_result_is_gimple_val.
2830 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
2832 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
2833 with a gimple_match_op *.
2834 (vn_nary_build_or_lookup): Likewise. Update call to
2835 vn_nary_build_or_lookup_1.
2836 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
2837 gimple_match_op *. Update calls to the gimple_resimplify routines
2838 and to gimple_simplified_result_is_gimple_val.
2839 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
2840 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
2841 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
2842 (visit_nary_op): Likewise.
2843 (visit_reference_op_load): Likewise.
2845 2018-05-23 Luis Machado <luis.machado@linaro.org>
2847 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
2848 modifier for printing the step amount.
2850 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
2853 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
2856 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
2858 * doc/sourcebuild.texi (Endianness): New subsubsection.
2860 2018-05-23 Luis Machado <luis.machado@linaro.org>
2862 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2863 <prefetch_dynamic_strides>: New const bool field.
2864 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2865 prefetch_dynamic_strides.
2866 (exynosm1_prefetch_tune): Likewise.
2867 (thunderxt88_prefetch_tune): Likewise.
2868 (thunderx_prefetch_tune): Likewise.
2869 (thunderx2t99_prefetch_tune): Likewise.
2870 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
2872 (aarch64_override_options_internal): Update to set
2873 PARAM_PREFETCH_DYNAMIC_STRIDES.
2874 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
2875 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
2876 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
2877 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
2878 prefetch-dynamic-strides setting.
2880 2018-05-23 Luis Machado <luis.machado@linaro.org>
2882 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2883 <minimum_stride>: New const int field.
2884 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2885 minimum_stride field defaulting to -1.
2886 (exynosm1_prefetch_tune): Likewise.
2887 (thunderxt88_prefetch_tune): Likewise.
2888 (thunderx_prefetch_tune): Likewise.
2889 (thunderx2t99_prefetch_tune): Likewise.
2890 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
2891 <default_opt_level>: Set to 3.
2892 (aarch64_override_options_internal): Update to set
2893 PARAM_PREFETCH_MINIMUM_STRIDE.
2894 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
2895 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
2896 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
2897 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
2898 stride is constant and is below the minimum stride threshold.
2900 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
2902 * config/arm/arm-cpus.in (mode26): Delete.
2903 (armv4): Delete mode26 reference.
2904 * config/arm/arm.c (arm_configure_build_target): Delete use of
2907 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
2909 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
2911 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
2912 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
2914 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
2915 (floatunsdidf2): Ditto.
2917 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
2919 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
2920 (fixuns_trunc<mode>si2_avx512f): Ditto.
2921 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
2922 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
2923 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
2925 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
2927 PR rtl-optimization/79985
2928 * df-scan.c (df_insn_refs_collect): Remove special case for
2929 global registers and asm statements.
2931 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
2933 * extend.texi (Global Register Variables): Rewrite the bullet list.
2934 Note that the register is available for allocation. Note that access
2935 via inline asm must use constraints. Add note about async-signal
2936 handlers. Remove paragraph about automagic register selection.
2938 2018-05-23 Richard Biener <rguenther@suse.de>
2940 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
2941 of fixed offset from memset VN.
2943 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
2945 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
2947 (alloc_cand_and_find_basis): Initialize first_interp field.
2948 (slsr_process_mul): Modify first_interp field.
2949 (slsr_process_add): Likewise.
2950 (slsr_process_cast): Modify first_interp field for each new
2952 (slsr_process_copy): Likewise.
2953 (dump_candidate): Dump first_interp field.
2954 (replace_mult_candidate): Process all interpretations, not just
2956 (replace_rhs_if_not_dup): Likewise.
2957 (replace_one_candidate): Likewise.
2959 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
2961 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
2963 (aarch64_needs_frame_chain): New function.
2964 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
2966 2018-05-23 Sudakshina Das <sudi.das@arm.com>
2969 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
2970 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
2971 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
2972 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
2973 as true for strict-align.
2974 (aarch64_can_inline_p): Perform checks even when callee has no
2975 attributes to check for strict alignment.
2976 * doc/extend.texi (AArch64 Function Attributes): Document
2978 * doc/invoke.texi: (AArch64 Options): Likewise.
2980 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
2982 PR tree-optimization/85853
2983 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
2984 the handling of the root of the node to...
2985 (vect_slp_analyze_node_operations_1): ...this new function,
2986 and run the whole thing with the child nodes' def types
2987 set according to their SLP node's def type.
2989 2018-05-23 Richard Biener <rguenther@suse.de>
2992 * tree-data-ref.c (create_runtime_alias_checks): Defer
2993 and ignore overflow warnings.
2995 2018-05-23 Yury Gribov <tetra2005@gmail.com>
2997 PR tree-optimization/85822
2998 * tree-vrp.c (is_masked_range_test): Fix handling of negative
3001 2018-05-23 Richard Biener <rguenther@suse.de>
3003 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
3004 memset constants via native_interpret_expr.
3006 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
3009 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
3011 (cgraph_node::create_alias): Likewise.
3012 (cgraph_node::get_availability): Check ifunc_resolver instead
3013 of looking up ifunc attribute.
3014 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
3015 * varasm.c (do_assemble_alias): Likewise.
3016 (assemble_alias): Likewise.
3017 (default_binds_local_p_3): Likewise.
3018 * cgraph.h (cgraph_node): Add ifunc_resolver.
3019 (cgraph_node::only_called_directly_or_aliased_p): Return false
3021 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
3023 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
3024 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
3025 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
3026 instead of looking up ifunc attribute.
3028 2018-05-22 Luis Machado <luis.machado@linaro.org>
3030 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
3032 2018-05-22 Martin Sebor <msebor@redhat.com>
3035 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
3036 only when expasion succeeds.
3037 (expand_builtin_strcmp): Same.
3038 (expand_builtin_strncmp): Same.
3040 2018-05-22 Martin Sebor <msebor@redhat.com>
3042 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
3044 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3045 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3047 * config/aarch64/aarch64-ldpstp.md: Replace uses of
3048 aarch64_mem_pair_operand with memory_operand and delete operand swapping
3050 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3051 Add check for legitimate_address.
3052 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
3053 (aarch64_swap_ldrstr_operands): New.
3054 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
3057 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3058 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3060 * config/aarch64/aarch64.md: New patterns to generate stp
3062 (store_pair_sw, store_pair_dw): New patterns to generate stp for
3063 single words and double words.
3064 (load_pair_sw, load_pair_dw): Likewise.
3065 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
3067 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
3069 * config/aarch64/aarch64-ldpstp.md: Modify peephole
3070 for different mode ldpstp and add peephole for merged zero stores.
3072 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3074 (aarch64_gen_store_pair): Rename calls to match new patterns.
3075 (aarch64_gen_load_pair): Rename calls to match new patterns.
3076 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
3077 (load_pair<DREG:mode><DREG2:mode>): ... This.
3078 (store_pair<mode>): Rename to...
3079 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
3080 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
3082 (V_INT_EQUIV): Handle SImode.
3083 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
3086 2018-05-22 Martin Sebor <msebor@redhat.com>
3089 * calls.c (maybe_warn_nonstring_arg): Use string length to set
3090 or ajust the presumed bound on an operation to avoid unnecessary
3093 2018-05-22 Martin Sebor <msebor@redhat.com>
3095 PR tree-optimization/85826
3096 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
3097 assuming that a DECL necesarily has a constant size.
3099 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
3102 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
3104 2018-05-22 Richard Biener <rguenther@suse.de>
3106 PR tree-optimization/85834
3107 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
3108 non-constant and non-zero memset arguments.
3110 2018-05-22 Martin Liska <mliska@suse.cz>
3113 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
3115 2018-05-22 Richard Biener <rguenther@suse.de>
3117 PR tree-optimization/85863
3118 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
3119 comparisons when vectype is specified.
3120 (vectorizable_condition): Do not specify vectype for
3121 vect_is_simple_cond when SLP vectorizing.
3123 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
3126 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
3127 define __ibm128 as long double.
3128 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
3129 as a distinct type when IEEE 128-bit support is enabled.
3130 (init_float128_ieee): Fix up conversions between IFmode and IEEE
3131 128-bit types to use the correct functions.
3132 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
3133 convert between 128-bit floating point types that have different
3134 modes but the same representation, instead of using gen_lowpart to
3136 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
3138 (IFKF_reg): New attributes to give the register constraints for
3140 (extend<mode>tf2_internal): New insns to mark an explicit
3141 conversion between 128-bit floating point types that have a
3142 different mode but share the same representation.
3144 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
3146 PR tree-optimization/85814
3147 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
3148 a null return from get_strinfo when unsharing the next
3149 strinfo in the chain.
3151 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
3154 * varasm.c (weak_finish): Clean up weak_decls.
3156 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3158 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
3159 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
3160 UNSPEC_UADALP values.
3161 * config/aarch64/iterators.md (ABAL): New int iterator.
3164 (sur): Add mappings for the above.
3165 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
3167 (aarch64_<sur>abal<mode>_4): Likewise.
3168 (aarch64_<sur>adalp<mode>_3): Likewise.
3169 (<sur>sadv16qi): New define_expand.
3171 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
3173 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
3174 (*movdf_internal): Ditto.
3175 (*rcpsf2_sse): Ditto.
3176 (*rsqrtsf2_sse): Ditto.
3177 (*sqrt<mode>2_sse): Ditto.
3179 2018-05-21 Tamar Christina <tamar.christina@arm.com>
3181 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
3183 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
3184 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
3185 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
3186 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
3188 * config/aarch64/arm_neon.h: Likewise.
3189 * config/aarch64/iterators.md (VQ_I): New.
3191 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
3193 * config.gcc: Add arc/t-multilib-linux to tmake_file for
3195 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
3198 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
3200 * config/nds32/constraints.md (S): New constraint.
3201 * config/nds32/nds32.md (call_internal): Use constraint S.
3202 (call_value_internal): Likewise.
3203 (sibcall_internal): Likewise.
3204 (sibcall_value_internal): Likewise.
3206 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
3207 Chung-Ju Wu <jasonwucj@gmail.com>
3209 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
3212 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
3213 Chung-Ju Wu <jasonwucj@gmail.com>
3215 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
3216 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
3217 (nds32_rtx_costs_impl): Simplify.
3218 (nds32_address_cost_impl): Simplify.
3219 (nds32_init_rtx_costs): New function.
3220 (nds32_rtx_costs_speed_prefer): Likewise.
3221 (nds32_rtx_costs_size_prefer): Likewise.
3222 (nds32_address_cost_speed_prefer): Likewise.
3223 (nds32_address_cost_speed_fwprop): Likewise.
3224 (nds32_address_cost_size_prefer): Likewise.
3225 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
3226 * config/nds32/nds32.c (nds32_option_override): Use
3227 nds32_init_rtx_costs function.
3229 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
3231 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
3232 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
3233 (TARGET_PIPELINE_N8): Likewise.
3234 (TARGET_PIPELINE_N10): Likewise.
3235 (TARGET_PIPELINE_N13): Likewise.
3236 (TARGET_PIPELINE_GRAYWOLF): Likewise.
3238 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
3240 * config/nds32/nds32-fpu.md: Update copyright year.
3242 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3244 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
3246 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3248 * config/nds32/nds32.c
3249 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
3250 * config/nds32/nds32.opt (minline-asm-r15): New option.
3252 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3254 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
3256 * config/nds32/nds32.md (abssi2): New pattern.
3258 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
3260 * config/i386/i386.md (rex64namesuffix): New mode attribute.
3261 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
3262 Merge insn pattern from sse_cvtsi2ss<round_name> and
3263 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
3264 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
3265 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
3266 using SWI48 mode iterator.
3267 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
3268 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
3269 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
3270 pattern from sse_cvttss2si<round_saeonly_name>
3271 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
3272 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
3273 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
3274 using SWI48 mode iterator.
3275 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
3276 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
3277 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
3278 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
3279 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
3280 using SWI48 mode iterator.
3281 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
3282 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
3283 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
3284 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
3285 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
3286 SWI48 mode iterator.
3287 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
3288 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
3289 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
3290 pattern from sse_cvttsd2si<round_saeonly_name>
3291 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
3293 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3295 * config/nds32/nds32-md-auxiliary.c
3296 (nds32_valid_smw_lwm_base_p): Refine.
3297 (nds32_output_smw_single_word): Refine.
3298 (nds32_output_smw_double_word): New.
3299 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
3301 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3303 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
3304 (nds32_output_stack_pop): Refine.
3305 (nds32_expand_unaligned_load): Refine.
3306 (nds32_expand_unaligned_store): Refine.
3308 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3309 Chung-Ju Wu <jasonwucj@gmail.com>
3311 * config/nds32/constants.md: Add TP_REGNUM constant.
3312 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
3313 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
3315 * config/nds32/nds32-doubleword.md: Consider flag_pic.
3316 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
3317 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
3318 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
3319 and PIC code generation.
3320 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
3322 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
3324 * config/nds32/nds32.md: Support TLS and PIC.
3325 * config/nds32/nds32.c: Support TLS and PIC.
3326 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
3327 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
3330 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3332 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
3333 mode with E_ prefix.
3335 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3336 Chung-Ju Wu <jasonwucj@gmail.com>
3338 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
3339 * config/nds32/nds32-md-auxiliary.c
3340 (symbolic_reference_mentioned_p): New.
3341 (nds32_legitimize_ict_address): New.
3342 (nds32_expand_ict_move): New.
3343 (nds32_indirect_call_referenced_p): New.
3344 (nds32_symbol_binds_local_p): Delete.
3345 (nds32_long_call_p): Modify.
3346 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
3347 * config/nds32/nds32-protos.h
3348 (symbolic_reference_mentioned_p): Declare.
3349 (nds32_legitimize_ict_address): Declare.
3350 (nds32_expand_ict_move): Declare.
3351 (nds32_indirect_call_referenced_p): Declare.
3352 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
3353 (nds32_relax_group): Use nds32_ict_const_p as condition.
3354 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
3355 (nds32_asm_file_start): Output ict_model directive in asm code.
3356 (nds32_legitimate_address_p): Consider indirect call.
3357 (nds32_print_operand): Consider indirect call.
3358 (nds32_print_operand_address): Consider indirect call.
3359 (nds32_insert_attributes): Handle "indirect_call" attribute.
3360 (TARGET_LEGITIMATE_ADDRESS_P): Define.
3361 (TARGET_LEGITIMATE_CONSTANT_P): Define.
3362 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
3363 (TARGET_DELEGITIMIZE_ADDRESS): Define.
3364 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
3365 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
3366 (TARGET_ICT_MODEL_SMALL): Define.
3367 (TARGET_ICT_MODEL_LARGE): Define.
3368 * config/nds32/nds32.md (movsi): Consider ict model.
3369 (call, call_value): Consider ict model.
3370 (sibcall, sibcall_value): Consider ict model.
3371 * config/nds32/nds32.opt (mict-model): New option.
3372 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
3375 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
3376 Monk Chiang <sh.chiang04@gmail.com>
3377 Jim Wilson <jimw@sifive.com>
3379 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
3380 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
3381 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
3382 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
3383 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
3384 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
3385 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
3386 compute save_libcall_adjustment properly.
3387 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
3388 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
3389 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
3390 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
3391 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
3392 (ABI_SPEC): Handle mabi=ilp32e.
3393 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
3394 (RVE): Add RVE mask.
3395 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
3396 <-march>: Add rv32e as an example.
3398 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
3401 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
3402 (intra_create_variable_infos): Handle C++ constructors.
3404 2018-05-18 Martin Liska <mliska@suse.cz>
3406 * passes.def: Remove a redundant pass.
3408 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
3411 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
3413 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3415 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
3417 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
3419 (armv2, armv2a, armv3, armv3m): Delete architectures.
3420 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
3421 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
3422 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
3424 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
3425 (*mulsidi3adddi): Likewise.
3426 (mulsidi3): Likewise.
3427 (*mulsidi3_nov6): Likewise.
3428 (umulsidi3): Likewise.
3429 (umulsidi3_nov6): Likewise.
3430 (umaddsidi4): Likewise.
3431 (*umulsidi3adddi): Likewise.
3432 (smulsi3_highpart): Likewise.
3433 (*smulsi3_highpart_nov6): Likewise.
3434 (umulsi3_highpart): Likewise.
3435 (*umulsi3_highpart_nov6): Likewise.
3436 * config/arm/arm.h (arm_arch3m): Delete.
3437 * config/arm/arm.c (arm_arch3m): Delete.
3438 (arm_option_override_internal): Update armv3-related comment.
3439 (arm_configure_build_target): Delete use of isa_bit_mode32.
3440 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
3441 (arm_rtx_costs_internal): Delete check of arm_arch3m.
3442 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
3444 (mulusa3): Likewise.
3445 * config/arm/arm-protos.h (arm_arch3m): Delete.
3446 * config/arm/arm-tables.opt: Regenerate.
3447 * config/arm/arm-tune.md: Likewise.
3448 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
3449 deleted architectures.
3451 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3453 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
3454 (armv5t, armv5te): New features.
3455 (ARMv5, ARMv5e): Delete fgroups.
3456 (ARMv5t, ARMv5te): Adjust for above changes.
3458 (armv5, armv5e): Delete arches.
3459 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
3461 (*call_reg_arm): Likewise.
3462 (*call_value_reg_armv5): Likewise.
3463 (*call_value_reg_arm): Likewise.
3464 (*call_symbol): Likewise.
3465 (*call_value_symbol): Likewise.
3466 (*sibcall_insn): Likewise.
3467 (*sibcall_value_insn): Likewise.
3469 (prefetch): Likewise.
3470 (define_split and define_peephole2 dependent on arm_arch5):
3472 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
3474 (TARGET_ARM_QBIT): Likewise.
3475 (TARGET_DSP_MULTIPLY): Likewise.
3476 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
3477 (arm_arch5, arm_arch5e): Delete.
3478 (arm_arch5t, arm_arch5te): Declare.
3479 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
3480 (arm_arch5t): Declare.
3481 (arm_option_reconfigure_globals): Update for the above.
3482 (arm_options_perform_arch_sanity_checks): Update comment, replace
3483 use of arm_arch5 with arm_arch5t.
3484 (use_return_insn): Likewise.
3485 (arm_emit_call_insn): Likewise.
3486 (output_return_instruction): Likewise.
3487 (arm_final_prescan_insn): Likewise.
3488 (arm_coproc_builtin_available): Likewise.
3489 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
3490 arm_arch5e with arm_arch5t and arm_arch5te.
3491 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
3492 (arm_arch5t, arm_arch5te): Declare.
3493 * config/arm/arm-tables.opt: Regenerate.
3494 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
3495 * config/arm/t-multilib: Likewise.
3496 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
3497 instead of arm_arch5.
3498 (*call_reg_thumb1): Likewise.
3499 (*call_value_reg_thumb1_v5): Likewise.
3500 (*call_value_reg_thumb1): Likewise.
3501 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
3503 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
3505 2018-05-18 Martin Liska <mliska@suse.cz>
3507 PR gcov-profile/84846
3508 * doc/gcov.texi: Document -t option of gcov tool.
3510 2018-05-18 Martin Liska <mliska@suse.cz>
3512 PR gcov-profile/84846
3513 * gcov.c (print_usage): Add new -t option.
3514 (process_args): Handle the option.
3515 (generate_results): Use stdout as output when requested by
3518 2018-05-18 Martin Liska <mliska@suse.cz>
3520 PR gcov-profile/84846
3521 * coverage.c (coverage_init): Write PWD to .gcno file.
3522 * doc/gcov.texi: Document how working directory is printed.
3523 * gcov-dump.c (dump_gcov_file): Print PWD.
3524 * gcov.c (output_intermediate_file): Likewise.
3525 (read_graph_file): Read PWD string.
3526 (output_lines): Print PWD.
3528 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3531 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
3532 for retval and return false if all args to phi are zero.
3534 2018-05-18 Richard Biener <rguenther@suse.de>
3536 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
3538 (evrp_dom_walker::before_dom_children): Call it.
3540 2018-05-18 Richard Biener <rguenther@suse.de>
3542 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
3543 results when processing array refs with variable index.
3545 2018-05-18 Toon Moene <toon@moene.org>
3547 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
3548 directly after that of -floop-interchange. Indicate that both
3549 options are enabled by default when specifying -O3.
3551 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3553 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
3554 iterator. Delete separate integer-mode vec_set<mode> expander.
3555 (aarch64_simd_vec_setv2di): Delete.
3556 (vec_setv2di): Delete.
3557 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
3558 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
3559 the "w, r" alternative.
3561 2018-05-18 Martin Liska <mliska@suse.cz>
3563 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
3564 * tree-pass.h (make_pass_lower_switch_O0): New function.
3565 * tree-switch-conversion.c (node_has_low_bound): Remove.
3566 (node_has_high_bound): Likewise.
3567 (node_is_bounded): Likewise.
3568 (class pass_lower_switch): Make it a template type and create
3570 (pass_lower_switch::execute): Add template argument.
3571 (make_pass_lower_switch): New function.
3572 (make_pass_lower_switch_O0): New function.
3573 (do_jump_if_equal): Remove.
3574 (emit_case_nodes): Simplify to just handle all 3 cases and leave
3575 all the hard work to tree optimization passes.
3577 2018-05-18 Martin Liska <mliska@suse.cz>
3579 * dbgcnt.c (limit_low): Renamed from limit.
3580 (limit_high): New variable.
3581 (dbg_cnt_is_enabled): Check for upper limit.
3582 (dbg_cnt): Adjust dumping.
3583 (dbg_cnt_set_limit_by_index): Add new argument for high
3585 (dbg_cnt_set_limit_by_name): Likewise.
3586 (dbg_cnt_process_single_pair): Parse new format.
3587 (dbg_cnt_process_opt): Use strtok.
3588 (dbg_cnt_list_all_counters): Remove 'value' and add
3590 * doc/invoke.texi: Document changes.
3592 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
3594 * doc/sourcebuild.texi (scalar_all_fma): Document.
3595 * tree.def (FMA_EXPR): Delete.
3596 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
3597 * internal-fn.c (ternary_direct): New macro.
3598 (expand_ternary_optab_fn): Likewise.
3599 (direct_ternary_optab_supported_p): Likewise.
3600 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
3601 * builtins.c (fold_builtin_fma): Delete.
3602 (fold_builtin_3): Don't call it.
3603 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
3604 * expr.c (expand_expr_real_2): Likewise.
3605 * fold-const.c (operand_equal_p): Likewise.
3606 (fold_ternary_loc): Likewise.
3607 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
3608 * gimple.c (DEFTREECODE): Likewise.
3609 * gimplify.c (gimplify_expr): Likewise.
3610 * optabs-tree.c (optab_for_tree_code): Likewise.
3611 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
3612 * tree-eh.c (operation_could_trap_p): Likewise.
3613 (stmt_could_throw_1_p): Likewise.
3614 * tree-inline.c (estimate_operator_cost): Likewise.
3615 * tree-pretty-print.c (dump_generic_node): Likewise.
3616 (op_code_prio): Likewise.
3617 * tree-ssa-loop-im.c (stmt_cost): Likewise.
3618 * tree-ssa-operands.c (get_expr_operands): Likewise.
3619 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
3620 * fold-const-call.h (fold_fma): Delete.
3621 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
3622 CFN_FNMA and CFN_FNMS.
3624 * genmatch.c (combined_fn): New enum.
3625 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
3626 (commutative_op): New function.
3627 (commutate): Use it. Handle more than 2 operands.
3628 (dt_operand::gen_gimple_expr): Use commutative_op.
3629 (parser::parse_expr): Allow :c to be used with non-binary
3630 operators if the commutative operand is known.
3631 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
3632 CFN_FMS, CFN_FNMA and CFN_FNMS.
3633 (backprop::process_assign_use): Remove FMA_EXPR handling.
3634 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
3635 (gen_hsa_fma): New function.
3636 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
3637 IFN_FNMA and IFN_FNMS.
3638 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
3639 * gimple-fold.h (follow_all_ssa_edges): Declare.
3640 * gimple-fold.c (follow_all_ssa_edges): New function.
3641 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
3642 gimple_build interface and use follow_all_ssa_edges to fold the result.
3643 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
3644 instead of checking for optabs directly.
3645 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
3646 rather than FMA_EXPRs.
3647 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
3648 call to IFN_FMA instead of an FMA_EXPR.
3650 2018-05-17 Jim Wilson <jimw@sifive.com>
3652 * expr.c (do_tablejump): When converting index to Pmode, if we have a
3653 sign extended promoted subreg, and the range does not have the sign bit
3654 set, then do a sign extend.
3656 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
3657 test, check for sign extended subreg and/or constant operands, and
3658 do a sign extend in that case.
3660 2018-05-17 Steve Ellcey <sellcey@cavium.com>
3662 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
3663 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
3665 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
3666 Change logics_shift_reg to logics_shift_imm.
3667 (thunderx2t99_fp_loadpair_basic): Delete.
3668 (thunderx2t99_fp_storepair_basic): Delete.
3669 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
3670 (thunderx2t99_asimd_polynomial): Delete.
3671 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
3672 and neon_fp_mul_d_scalar_q.
3673 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
3674 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
3675 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
3676 (thunderx2t99_asimd_lut): Add missing tbl types.
3677 (thunderx2t99_asimd_ext): Delete.
3678 (thunderx2t99_asimd_load1_1_mult): Delete.
3679 (thunderx2t99_asimd_load1_2_mult): Delete.
3680 (thunderx2t99_asimd_load1_ldp): New.
3681 (thunderx2t99_asimd_load1): New.
3682 (thunderx2t99_asimd_load2): Add missing *load2* types.
3683 (thunderx2t99_asimd_load3): New.
3684 (thunderx2t99_asimd_load4): New.
3685 (thunderx2t99_asimd_store1_1_mult): Delete.
3686 (thunderx2t99_asimd_store1_2_mult): Delete.
3687 (thunderx2t99_asimd_store2_mult): Delete.
3688 (thunderx2t99_asimd_store2_onelane): Delete.
3689 (thunderx2t99_asimd_store_stp): New.
3690 (thunderx2t99_asimd_store1): New.
3691 (thunderx2t99_asimd_store2): New.
3692 (thunderx2t99_asimd_store3): New.
3693 (thunderx2t99_asimd_store4): New.
3695 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
3697 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
3698 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
3700 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
3701 Segher Boessenkool <segher@kernel.crashing.org>
3704 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
3707 2018-05-17 Richard Biener <rguenther@suse.de>
3709 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
3710 for pruning loop and prune defs feeding only already visited PHIs.
3712 2018-05-17 Richard Biener <rguenther@suse.de>
3714 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
3716 2018-05-17 Bin Cheng <bin.cheng@arm.com>
3717 Richard Biener <rguenther@suse.de>
3719 PR tree-optimization/85793
3720 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
3721 for VMAT_ELEMENTWISE.
3723 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
3725 * internal-fn.h (lookup_internal_fn): Declare
3726 * internal-fn.c (lookup_internal_fn): New function.
3727 * gimple.c (gimple_build_call_from_tree): Handle calls to
3729 * gimple-pretty-print.c (dump_gimple_call): Print "." before
3730 internal function names.
3731 * tree-pretty-print.c (dump_generic_node): Likewise.
3732 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
3734 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
3736 * gimple-fold.h (gimple_build): Make the function forms take
3737 combined_fn rather than built_in_function.
3738 (gimple_simplify): Likewise.
3739 * gimple-match-head.c (gimple_simplify): Likewise.
3740 * gimple-fold.c (gimple_build): Likewise.
3741 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
3742 rather than gimple_build_call_internal.
3743 (get_initial_defs_for_reduction): Likewise.
3744 (vect_create_epilog_for_reduction): Likewise.
3745 (vectorizable_live_operation): Likewise.
3747 2018-05-17 Martin Liska <mliska@suse.cz>
3749 * gimple-ssa-sprintf.c (format_directive): Do not use
3750 space in between 'G_' and '('.
3752 2018-05-17 Jakub Jelinek <jakub@redhat.com>
3755 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
3756 even if the mask is not all ones.
3759 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
3761 (ix86_gimple_fold_builtin): Likewise.
3764 * config/i386/i386.c: Include tree-vector-builder.h.
3765 (ix86_vector_shift_count): New function.
3766 (ix86_fold_builtin): Fold shift builtins by scalar count.
3767 (ix86_gimple_fold_builtin): Likewise.
3769 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
3770 _mm512_setzero): New intrinsics.
3772 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
3773 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3775 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
3776 code generation for cases where splatting a value is not useful.
3777 * simplify-rtx.c (simplify_ternary_operation): Simplify
3778 vec_merge across a vec_duplicate and a paradoxical subreg forming
3779 a vector mode to a vec_concat.
3781 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
3783 * config.gcc: Support "goldmont-plus".
3784 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
3786 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3787 PROCESSOR_GOLDMONT_PLUS.
3788 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
3789 (processor_target_table): Add "goldmont-plus".
3790 (PTA_GOLDMONT_PLUS): Define.
3791 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
3792 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
3793 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
3794 (fold_builtin_cpu): Add "goldmont-plus".
3795 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
3796 (ix86_option_override_internal): Add "goldmont-plus".
3797 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
3798 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
3799 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
3800 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
3802 2018-05-17 Richard Biener <rguenther@suse.de>
3804 PR tree-optimization/85757
3805 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
3806 remove defs that only feed that PHI from further processing.
3808 2018-05-16 Jim Wilson <jimw@sifive.com>
3810 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
3812 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
3814 2018-05-16 Mark Wielaard <mark@klomp.org>
3816 * dwarf2out.c (count_index_strings): New function.
3817 (output_indirect_strings): Call count_index_strings and generate
3818 header for dwarf_version >= 5.
3820 2018-05-16 Mark Wielaard <mark@klomp.org>
3822 * dwarf2out.c (dwarf_FORM): New function.
3823 (set_indirect_string): Use dwarf_FORM.
3824 (reset_indirect_string): Likewise.
3825 (size_of_die): Likewise.
3826 (value_format): Likewise.
3827 (output_die): Likewise.
3828 (add_skeleton_AT_string): Likewise.
3829 (output_macinfo_op): Likewise.
3830 (index_string): Likewise.
3831 (output_index_string_offset): Likewise.
3832 (output_index_string): Likewise.
3833 (count_index_strings): Likewise.
3835 2018-05-16 Carl Love <cel@us.ibm.com>
3837 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
3838 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
3840 2018-05-16 Martin Jambor <mjambor@suse.cz>
3842 * ipa-prop.c (ipa_free_all_edge_args): Remove.
3843 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
3845 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
3847 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
3848 (fnma<mode>4): Likewise.
3849 (fms<mode>4): Likewise.
3850 (fnms<mode>4): Likewise.
3851 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
3852 (aarch64_fnma<mode>4): Likewise.
3853 (aarch64_fms<mode>4): Likewise.
3854 (aarch64_fnms<mode>4): Likewise.
3855 (aarch64_fnmadd<mode>4): Likewise.
3857 2018-05-16 Jason Merrill <jason@redhat.com>
3859 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
3861 2018-05-16 Richard Biener <rguenther@suse.de>
3863 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
3864 (dump_stmt_cost): Declare.
3865 (add_stmt_cost): Dump cost we add.
3866 (add_stmt_costs): New function.
3867 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
3869 (vect_analyze_stmt): Adjust prototype.
3870 (vectorizable_condition): Likewise.
3871 (vectorizable_live_operation): Likewise.
3872 (vectorizable_reduction): Likewise.
3873 (vectorizable_induction): Likewise.
3874 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
3875 cost vector to pass to vectorizable_ and record afterwards.
3876 (vect_model_reduction_cost): Take cost vector argument and adjust.
3877 (vect_model_induction_cost): Likewise.
3878 (vectorizable_reduction): Likewise.
3879 (vectorizable_induction): Likewise.
3880 (vectorizable_live_operation): Likewise.
3881 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
3882 SLP_TREE_NUMBER_OF_VEC_STMTS.
3883 (vect_analyze_slp_cost_1): Remove.
3884 (vect_analyze_slp_cost): Likewise.
3885 (vect_slp_analyze_node_operations): Take visited args and
3886 a target cost vector. Avoid processing already visited stmt sets.
3887 (vect_slp_analyze_operations): Use a local cost vector to gather
3888 costs and register those of non-discarded instances.
3889 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
3890 (vect_schedule_slp_instance): Remove copying of
3891 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
3893 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
3894 adding cost. Record cost entry location.
3895 (vect_prologue_cost_for_slp_op): Function to compute cost of
3896 a constant or invariant generated for SLP vect in the prologue,
3897 split out from vect_analyze_slp_cost_1.
3898 (vect_model_simple_cost): Make static. Adjust for SLP costing.
3899 (vect_model_promotion_demotion_cost): Likewise.
3900 (vect_model_store_cost): Likewise, make static.
3901 (vect_model_load_cost): Likewise.
3902 (vectorizable_bswap): Add cost vector arg and adjust.
3903 (vectorizable_call): Likewise.
3904 (vectorizable_simd_clone_call): Likewise.
3905 (vectorizable_conversion): Likewise.
3906 (vectorizable_assignment): Likewise.
3907 (vectorizable_shift): Likewise.
3908 (vectorizable_operation): Likewise.
3909 (vectorizable_store): Likewise.
3910 (vectorizable_load): Likewise.
3911 (vectorizable_condition): Likewise.
3912 (vectorizable_comparison): Likewise.
3913 (can_vectorize_live_stmts): Likewise.
3914 (vect_analyze_stmt): Likewise.
3915 (vect_transform_stmt): Adjust calls to vectorizable_*.
3916 * tree-vectorizer.c: Include gimple-pretty-print.h.
3917 (dump_stmt_cost): New function.
3919 2018-05-16 Richard Biener <rguenther@suse.de>
3921 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
3922 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
3923 * tree-ssa-dse.c: Include tree-ssa-loop.h.
3924 (check_name): New callback.
3925 (dse_classify_store): Track cycles via a visited bitmap of PHI
3926 defs and simplify handling of in-loop and across loop dead stores
3927 and properly fail for loop-variant refs. Handle byte-tracking with
3928 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
3931 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
3933 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
3934 (vect_get_mask_type_for_stmt): Likewise.
3935 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
3937 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
3938 to determine the statement's vector type and the vector type that
3939 should be used for calculating nunits. Deal with cases in which
3940 the type has to be deferred.
3941 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
3942 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
3943 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
3944 (vect_determine_vf_for_stmt): New functions, split out from...
3945 (vect_determine_vectorization_factor): ...here.
3946 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
3947 (vect_get_mask_type_for_stmt): New functions, split out from
3948 vect_determine_vectorization_factor.
3950 2018-05-16 Richard Biener <rguenther@suse.de>
3952 * tree-cfg.c (verify_gimple_assign_ternary): Properly
3953 verify the [VEC_]COND_EXPR embedded comparison.
3955 2018-05-15 Martin Sebor <msebor@redhat.com>
3957 PR tree-optimization/85753
3958 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
3959 RECORD_TYPE in addition to ARRAY_TYPE.
3961 2018-05-15 Martin Sebor <msebor@redhat.com>
3964 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
3966 2018-05-15 Richard Biener <rguenther@suse.de>
3968 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
3969 add by_clobber_p one. Change algorithm to collect all defs
3970 representing uses we need to walk and try reducing them to
3971 a single one before failing.
3972 (dse_dom_walker::dse_optimize_stmt): Adjust.
3974 2018-05-13 Mark Wielaard <mark@klomp.org>
3976 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
3977 (size_of_loc_descr): Likewise.
3978 (output_loc_operands): Likewise.
3979 (output_loc_operands_raw): Likewise.
3980 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
3981 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
3982 (hash_loc_operands): Likewise.
3983 (compare_loc_operands): Likewise.
3985 2018-05-14 Mark Wielaard <mark@klomp.org>
3987 * dwarf2out.c (count_index_addrs): New function.
3988 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
3990 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3992 PR tree-optimization/83648
3993 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
3994 return value as malloc candidate.
3996 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3999 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
4000 param as true in call to suggest_attribute.
4002 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
4004 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
4005 -mreadonly-in-sdata.
4007 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4009 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
4011 (aarch64_crypto_aesd_fused): Likewise.
4013 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
4015 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
4016 (movsi_aarch64): Likewise.
4017 (load_pairsi): Likewise.
4018 (load_pairdi): Likewise.
4019 (store_pairsi): Likewise.
4020 (store_pairdi): Likewise.
4021 (load_pairsf): Likewise.
4022 (load_pairdf): Likewise.
4023 (store_pairsf): Likewise.
4024 (store_pairdf): Likewise.
4025 (zero_extend): Likewise.
4026 (trunc): Swap alternatives.
4027 (fcvt_target): Add '?' to prefer w over r.
4029 2018-05-14 Jakub Jelinek <jakub@redhat.com>
4032 * config/i386/i386.md: Disallow non-commutative arithmetics in
4033 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
4034 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
4035 in the peephole2 before it.
4037 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
4039 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
4040 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
4041 (ix86_handle_option): Handle -mcldemote.
4042 * config.gcc: New header.
4043 * config/i386/cldemoteintrin.h: New file.
4044 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
4045 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
4047 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4048 OPTION_MASK_ISA_CLDEMOTE.
4049 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
4050 (ix86_valid_target_attribute_inner_p): Ditto.
4051 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
4052 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
4053 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
4054 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
4055 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
4057 * config/i386/i386.opt: Add -mcldemote.
4058 * config/i386/x86intrin.h: New header.
4059 * doc/invoke.texi: Add -mcldemote.
4061 2018-05-14 Richard Biener <rguenther@suse.de>
4063 * doc/match-and-simplify.texi: Adjust :s documentation.
4065 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
4067 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
4068 intended memcpy size.
4069 (REORDER_45): Likewise.
4071 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
4073 * sort.cc: New file.
4074 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
4075 * vec.c (qsort_chk): Use gcc_qsort.
4076 * Makefile.in (OBJS-libcommon): Add sort.o.
4077 (build/sort.o): New target. Use it...
4078 (BUILD_RTL): ... here, and...
4079 (build/gencfn-macros): ... here, and...
4080 (build/genmatch): ... here.
4082 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4083 Chung-Ju Wu <jasonwucj@gmail.com>
4085 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
4086 * config/nds32/nds32-graywolf.md: New file.
4087 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
4088 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
4090 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
4091 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
4092 * config/nds32/nds32.md (pipeline_model): Add graywolf.
4093 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
4094 * config/nds32/pipelines.md: Include n15 settings.
4096 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4097 Chung-Ju Wu <jasonwucj@gmail.com>
4099 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
4100 * config/nds32/nds32-n13.md: New file.
4101 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
4102 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
4104 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
4105 * config/nds32/nds32.md (pipeline_model): Add n13.
4106 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
4107 * config/nds32/pipelines.md: Include n13 settings.
4109 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4110 Chung-Ju Wu <jasonwucj@gmail.com>
4112 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
4113 * config/nds32/nds32-n10.md: New file.
4114 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
4115 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
4117 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
4118 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
4119 * config/nds32/nds32.md (pipeline_model): Add n10.
4120 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
4121 * config/nds32/pipelines.md: Include n10 settings.
4123 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
4124 Kito Cheng <kito.cheng@gmail.com>
4125 Chung-Ju Wu <jasonwucj@gmail.com>
4127 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
4128 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4129 Add enum values for DSP extension instructions.
4130 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
4132 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
4133 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
4135 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
4136 * config/nds32/nds32-dspext.md: New file for DSP implementation.
4137 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
4138 * config/nds32/nds32-intrinsic.md: Likewise.
4139 * config/nds32/nds32_intrinsic.h: Likewise.
4140 * config/nds32/nds32-md-auxiliary.c: Likewise.
4141 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
4142 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
4143 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
4144 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
4145 * config/nds32/nds32-protos.h: New declarations for DSP extension.
4146 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
4147 TYPE_DMAC in switch statement.
4148 * config/nds32/nds32.c: New checking and implementation for DSP
4149 extension instructions.
4150 * config/nds32/nds32.h: Likewise.
4151 * config/nds32/nds32.md: Likewise.
4152 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
4153 * config/nds32/predicates.md: Implement new predicates for DSP
4156 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
4158 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
4159 Reformat alternatives and attributes so it is easier to identify
4160 which constraints/attributes go with which instruction.
4161 (mov<mode>_hardfloat32, FMOVE64): Likewise.
4162 (mov<mode>_softfloat32, FMOVE64): Likewise.
4163 (mov<mode>_hardfloat64, FMOVE64): Likewise.
4164 (mov<mode>_softfloat64, FMOVE64): Likewise.
4166 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
4168 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4170 (Basic PowerPC Built-in Functions): The new name of the
4171 subsection previously known as "PowerPC Built-in Functions".
4172 (Basic PowerPC Built-in Functions Available on all Configurations):
4174 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
4175 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
4176 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
4177 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
4179 2018-05-11 Martin Jambor <mjambor@suse.cz>
4182 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
4185 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4188 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
4190 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
4192 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
4193 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
4194 (ix86_handle_option): Handle -mwaitpkg.
4195 * config.gcc: New header.
4196 * config/i386/cpuid.h (bit_WAITPKG): New bit.
4197 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
4198 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
4200 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4201 OPTION_MASK_ISA_WAITPKG.
4202 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
4203 (ix86_option_override_internal): Add PTA_WAITPKG.
4204 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
4205 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
4206 IX86_BUILTIN_TPAUSE.
4207 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
4208 __builtin_ia32_umwait and __builtin_ia32_tpause.
4209 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
4210 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
4211 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
4212 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
4213 UNSPECV_TPAUSE): New.
4214 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
4215 * config/i386/i386.opt: Add -mwaitpkg.
4216 * config/i386/waitpkgintrin.h: New file.
4217 * config/i386/x86intrin.h: New header.
4218 * doc/invoke.texi: Add -mwaitpkg.
4220 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4223 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
4225 (cortex-m0): Use armv6s-m isa.
4226 (cortex-m0plus): Likewise.
4227 (cortex-m1): Likewise.
4228 (cortex-m0.small-multiply): Likewise.
4229 (cortex-m0plus.small-multiply): Likewise.
4230 (cortex-m1.small-multiply): Likewise.
4232 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
4233 Jakub Jelinek <jakub@redhat.com>
4235 PR tree-optimization/85692
4236 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
4237 source permute as well.
4239 2018-05-11 Martin Liska <mliska@suse.cz>
4242 * doc/extend.texi: Document LLVM style format for no_sanitize
4245 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
4247 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
4248 mode_supports_vsx_dform_quad to mode_supports_dq_form.
4249 (mode_supports_vsx_dform_quad): Likewise.
4250 (mode_supports_vmx_dform): Move these functions to be next to the
4251 other mode_supports functions.
4252 (mode_supports_dq_form): Likewise.
4253 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
4254 mode_supports_dq_form.
4255 (reg_offset_addressing_ok_p): Likewise.
4256 (offsettable_ok_by_alignment): Likewise.
4257 (rs6000_legitimate_offset_address_p): Likewise.
4258 (legitimate_lo_sum_address_p): Likewise.
4259 (rs6000_legitimize_address): Likewise.
4260 (rs6000_legitimize_reload_address): Likewise.
4261 (rs6000_secondary_reload_inner): Likewise.
4262 (rs6000_preferred_reload_class): Likewise.
4263 (rs6000_output_move_128bit): Likewise.
4265 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
4267 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
4268 Generate SImode target register for null target.
4269 <case IX86_BUILTIN_XGETBV>: Ditto.
4270 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
4271 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
4273 2018-05-10 Carl Love <cel@us.ibm.com>
4275 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
4276 dcbtt and dcbtstt if operands[2] is 0.
4278 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
4281 * config/i386/sse.md (usadv64qi): New expander.
4283 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
4285 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
4286 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
4287 -maltivec=be support.
4288 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
4289 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
4290 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
4291 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
4292 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
4293 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
4294 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
4295 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
4296 altivec_vsumsws): Adjust.
4297 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
4298 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
4299 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
4301 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
4302 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
4303 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
4304 (altivec_lve<VI_char>x): Delete expand.
4305 (*altivec_lve<VI_char>x_internal): Rename to...
4306 (altivec_lve<VI_char>x): ... this.
4307 (altivec_lvxl_<mode>): Delete expand.
4308 (*altivec_lvxl_<mode>_internal): Rename to ...
4309 (altivec_lvxl_<mode>): ... this.
4310 (altivec_stvxl_<mode>): Delete expand.
4311 (*altivec_stvxl_<mode>_internal): Rename to ...
4312 (altivec_stvxl_<mode>): ... this.
4313 (altivec_stve<VI_char>x): Delete expand.
4314 (*altivec_stve<VI_char>x_internal): Rename to ...
4315 (altivec_stve<VI_char>x): ... this.
4316 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
4317 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
4318 reduc_plus_scal_<mode>): Adjust.
4319 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
4321 (rs6000_cpu_cpp_builtins): Adjust.
4322 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
4323 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
4324 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
4325 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
4326 -maltivec=be support.
4327 (rs6000_split_vec_extract_var): Adjust.
4328 (rs6000_split_v4si_init): Adjust.
4329 (swap_selector_for_mode): Delete.
4330 (altivec_expand_lvx_be, altivec_expand_stvx_be,
4331 altivec_expand_stvex_be): Delete.
4332 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
4333 -maltivec=be support.
4334 (rs6000_gimple_fold_builtin): Ditto.
4335 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
4337 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
4338 (TARGET_DIRECT_MOVE_64BIT): Adjust.
4339 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
4340 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
4341 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
4342 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
4343 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
4344 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
4345 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
4346 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
4347 anonymous split): Adjust.
4348 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
4349 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
4351 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
4353 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
4354 when --with-gxx-include-dir is also specified.
4355 * configure: Regenerate.
4357 2018-05-09 Jim Wilson <jimw@sifive.com>
4360 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
4361 * config/riscv/t-withmultilib: New.
4362 * config/riscv/withmultilib.h: New.
4363 * doc/install.texi: Document RISC-V --with-multilib-list support.
4365 2018-05-09 Richard Biener <rguenther@suse.de>
4367 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
4369 (vect_bb_vectorization_profitable_p): Adjust. Compute
4370 actual scalar cost using the cost vector and the add_stmt_cost
4373 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4375 PR rtl-optimization/85645
4376 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
4377 in the REG_CFA_REGISTER note for LR, don't leave it empty.
4379 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4381 PR rtl-optimization/85645
4382 * shrink-wrap.c (spread_components): Return a boolean saying if
4383 anything was changed.
4384 (try_shrink_wrapping_separate): Iterate spread_components until
4385 nothing changes anymore.
4387 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4389 PR rtl-optimization/85645
4390 * regrename.c (build_def_use): Also kill the chains that include the
4391 destination of a REG_CFA_REGISTER note.
4393 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4395 PR rtl-optimization/85645
4396 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
4397 insn that has a REG_CFA_REGISTER note.
4399 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
4401 * cfgexpand.c (expand_clobber): New function.
4402 (expand_gimple_stmt_1): Use it.
4403 * tree-vect-stmts.c (vect_clobber_variable): New function,
4405 (vectorizable_simd_clone_call): ...here.
4406 (vectorizable_store): Emit a clobber either side of an
4407 IFN_STORE_LANES sequence.
4408 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
4410 2018-05-09 Tom de Vries <tom@codesourcery.com>
4413 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
4414 (define_insn "trap_if_false"): Add exit after trap.
4416 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
4418 PR rtl-optimization/85638
4419 * bb-reorder.c: Include common/common-target.h.
4420 (create_forwarder_block): New function extracted from...
4421 (fix_up_crossing_landing_pad): ...here. Rename into...
4422 (dw2_fix_up_crossing_landing_pad): ...this.
4423 (sjlj_fix_up_crossing_landing_pad): New function.
4424 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
4425 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
4426 from both partitions and exit the loop after one iteration.
4428 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
4431 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4433 (Basic PowerPC Built-in Functions): The new name of the
4434 subsection previously known as "PowerPC Built-in Functions".
4435 (Basic PowerPC Built-in Functions Available on all Configurations):
4437 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
4439 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
4441 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
4443 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
4446 2018-05-08 Jim Wilson <jimw@sifive.com>
4448 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
4449 (LD_EMUL_SUFFIX): New.
4450 (LINK_SPEC): Use it.
4452 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
4454 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4456 (Basic PowerPC Built-in Functions): The new name of the
4457 subsection previously known as "PowerPC Built-in Functions".
4458 (Basic PowerPC Built-in Functions Available on all Configurations):
4460 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
4462 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
4464 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
4466 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
4469 2018-05-08 Jakub Jelinek <jakub@redhat.com>
4472 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
4473 after cmpelim optimization.
4475 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
4477 * config.gcc: Support "goldmont".
4478 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
4479 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4481 * config/i386/i386.c (m_GOLDMONT): Define.
4482 (processor_target_table): Add "goldmont".
4483 (PTA_GOLDMONT): Define.
4484 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
4485 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
4486 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
4487 (fold_builtin_cpu): Add "goldmont".
4488 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
4489 (ix86_option_override_internal): Add "goldmont".
4490 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
4491 (processor_type): Add PROCESSOR_GOLDMONT.
4492 * config/i386/i386.md: Add CPU "glm".
4493 * config/i386/glm.md: New file.
4494 * config/i386/x86-tune.def: Add m_GOLDMONT.
4495 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
4497 2018-05-08 Jakub Jelinek <jakub@redhat.com>
4500 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
4502 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
4503 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
4504 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
4507 * config/i386/i386.c (ix86_fold_builtin): Handle
4508 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
4511 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
4512 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
4514 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
4517 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
4518 (check_arch): Likewise.
4519 (check_fpu): Return the result rather than printing it.
4520 (end arch): Fix operator precedence.
4521 (end cpu): Likewise.
4522 (END): Print the result from check_fpu.
4524 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
4525 Alan Hayward <alan.hayward@arm.com>
4526 David Sherwood <david.sherwood@arm.com>
4528 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
4529 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
4530 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
4531 (*fcmuo<mode>_and): New patterns.
4533 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
4535 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
4536 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
4537 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
4538 (cmp_op, sve_imm_con): New code attributes.
4539 (SVE_COND_INT_CMP, imm_con): Delete.
4540 (cmp_op): Remove above unspecs from int attribute.
4541 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
4543 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
4544 comparison-specific unspecs.
4545 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
4546 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
4547 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
4548 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
4549 (*vec_fcm<cmp_op><mode>): Rename to...
4550 (*fcm<cmp_op><mode>): ...this and adjust likewise.
4551 (*vec_fcmuo<mode>): Rename to...
4552 (*fcmuo<mode>): ...this and adjust likewise.
4553 (*pred_fcm<cmp_op><mode>): New pattern.
4554 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
4555 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
4557 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
4559 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
4560 (aarch64_emit_sve_predicated_cond): New function.
4561 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
4562 (aarch64_emit_unspec_cond_or): Replace with...
4563 (aarch64_emit_sve_or_conds): ...this new function. Use
4564 aarch64_emit_sve_ptrue_op for the individual comparisons and
4565 aarch64_emit_binop to OR them together.
4566 (aarch64_emit_inverted_unspec_cond): Replace with...
4567 (aarch64_emit_sve_inverted_cond): ...this new function. Use
4568 aarch64_emit_sve_ptrue_op for the comparison and
4569 aarch64_emit_unop to invert the result.
4570 (aarch64_expand_sve_vec_cmp_float): Update after the above
4571 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
4573 2018-05-07 Nathan Sidwell <nathan@acm.org>
4575 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
4576 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
4577 (Backwards Compatibility): Likewise.
4579 2018-05-07 Luis Machado <luis.machado@linaro.org>
4583 2018-05-07 Luis Machado <luis.machado@linaro.org>
4585 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4586 <prefetch_dynamic_strides>: New const bool field.
4587 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4588 prefetch_dynamic_strides.
4589 (exynosm1_prefetch_tune): Likewise.
4590 (thunderxt88_prefetch_tune): Likewise.
4591 (thunderx_prefetch_tune): Likewise.
4592 (thunderx2t99_prefetch_tune): Likewise.
4593 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
4595 (aarch64_override_options_internal): Update to set
4596 PARAM_PREFETCH_DYNAMIC_STRIDES.
4597 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4598 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4599 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4600 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4601 prefetch-dynamic-strides setting.
4603 2018-05-07 Luis Machado <luis.machado@linaro.org>
4605 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4606 <minimum_stride>: New const int field.
4607 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4608 minimum_stride field.
4609 (exynosm1_prefetch_tune): Likewise.
4610 (thunderxt88_prefetch_tune): Likewise.
4611 (thunderx_prefetch_tune): Likewise.
4612 (thunderx2t99_prefetch_tune): Likewise.
4613 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
4614 (aarch64_override_options_internal): Update to set
4615 PARAM_PREFETCH_MINIMUM_STRIDE.
4616 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4617 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4618 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4619 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4620 stride is constant and is below the minimum stride threshold.
4622 2018-05-07 Luis Machado <luis.machado@linaro.org>
4624 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
4627 2018-05-07 Luis Machado <luis.machado@linaro.org>
4629 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4630 <prefetch_dynamic_strides>: New const bool field.
4631 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4632 prefetch_dynamic_strides.
4633 (exynosm1_prefetch_tune): Likewise.
4634 (thunderxt88_prefetch_tune): Likewise.
4635 (thunderx_prefetch_tune): Likewise.
4636 (thunderx2t99_prefetch_tune): Likewise.
4637 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
4639 (aarch64_override_options_internal): Update to set
4640 PARAM_PREFETCH_DYNAMIC_STRIDES.
4641 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4642 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4643 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4644 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4645 prefetch-dynamic-strides setting.
4647 2018-05-07 Luis Machado <luis.machado@linaro.org>
4649 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4650 <minimum_stride>: New const int field.
4651 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4652 minimum_stride field.
4653 (exynosm1_prefetch_tune): Likewise.
4654 (thunderxt88_prefetch_tune): Likewise.
4655 (thunderx_prefetch_tune): Likewise.
4656 (thunderx2t99_prefetch_tune): Likewise.
4657 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
4658 (aarch64_override_options_internal): Update to set
4659 PARAM_PREFETCH_MINIMUM_STRIDE.
4660 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4661 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4662 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4663 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4664 stride is constant and is below the minimum stride threshold.
4666 2018-05-06 Jakub Jelinek <jakub@redhat.com>
4669 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
4670 the type is addressable. Don't force op into register if it has
4673 2018-05-05 Roland McGrath <mcgrathr@google.com>
4676 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
4677 any section for which we don't know a specific type it should have,
4678 regardless of name. Previously this was done only for the exact
4679 names ".init_array", ".fini_array", and ".preinit_array".
4680 (default_elf_asm_named_section): Add comment about
4681 relationship with default_section_type_flags and SECTION_NOTYPE.
4682 (get_section): Don't consider it a type conflict if one side has
4683 SECTION_NOTYPE and the other doesn't, as long as neither has the
4684 SECTION_BSS et al used in the default_section_type_flags logic.
4686 2018-05-05 Tom de Vries <tom@codesourcery.com>
4689 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
4690 (workaround_barsyncs): New function.
4691 (nvptx_reorg): Use workaround_barsyncs.
4692 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4693 (define_expand "nvptx_membar_cta"): New define_expand.
4694 (define_insn "*nvptx_membar_cta"): New insn.
4696 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
4698 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
4699 To improve optimization opportunities.
4700 * builtin-types.def: The new needed builtin types for the above.
4702 2018-05-04 Richard Biener <rguenther@suse.de>
4704 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
4705 * gimple-ssa-store-merging.c
4706 (imm_store_chain_info::output_merged_store): Remove redundant create,
4707 release split_store vector contents on failure.
4708 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
4709 scalar stmt vector on cache hit.
4711 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
4713 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
4715 * config.gcc (powerpc-xilinx-eabi*): Remove.
4716 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
4718 (fusion_addis_mem_combo_load): Ditto.
4719 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
4721 (rs6000_cpu_cpp_builtins): Ditto.
4722 * config/rs6000/rs6000-linux.c
4723 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
4724 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
4725 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
4727 (rs6000_setup_reg_addr_masks): Ditto.
4728 (rs6000_init_hard_regno_mode_ok): Ditto.
4729 (rs6000_option_override_internal): Ditto.
4730 (legitimate_lo_sum_address_p): Ditto.
4731 (rs6000_legitimize_address): Ditto.
4732 (rs6000_legitimize_reload_address): Ditto.
4733 (rs6000_legitimate_address_p): Ditto.
4734 (abi_v4_pass_in_fpr): Ditto.
4735 (setup_incoming_varargs): Ditto.
4736 (rs6000_gimplify_va_arg): Ditto.
4737 (rs6000_split_multireg_move): Ditto.
4738 (rs6000_savres_strategy): Ditto.
4739 (rs6000_emit_prologue_components): Ditto.
4740 (rs6000_emit_epilogue_components): Ditto.
4741 (rs6000_emit_prologue): Ditto.
4742 (rs6000_emit_epilogue): Ditto.
4743 (rs6000_elf_file_end): Ditto.
4744 (rs6000_function_value): Ditto.
4745 (rs6000_libcall_value): Ditto.
4746 * config/rs6000/rs6000.h: Ditto.
4747 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
4748 (TARGET_MINMAX): ... this. New.
4749 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
4750 * config/rs6000/rs6000.md: Remove Xilinx FP support.
4751 (*movsi_internal1_single): Delete.
4752 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
4753 mfpu=, mxilinx-fpu): Delete.
4754 * config/rs6000/singlefp.h: Delete.
4755 * config/rs6000/sysv4.h: Remove Xilinx FP support.
4756 * config/rs6000/t-rs6000: Ditto.
4757 * config/rs6000/t-xilinx: Delete.
4758 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
4759 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
4760 (VStype_simple): Delete.
4761 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
4762 * config/rs6000/xfpu.h: Delete.
4763 * config/rs6000/xfpu.md: Delete.
4764 * config/rs6000/xilinx.h: Delete.
4765 * config/rs6000/xilinx.opt: Delete.
4766 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
4767 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
4769 2018-05-04 Tom de Vries <tom@codesourcery.com>
4772 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
4775 2018-05-04 Richard Biener <rguenther@suse.de>
4778 * tree-complex.c (update_complex_assignment): We are always in SSA form.
4779 (expand_complex_div_wide): Likewise.
4780 (expand_complex_operations_1): Likewise.
4781 (expand_complex_libcall): Preserve EH info of the original stmt.
4782 (tree_lower_complex): Handle removed blocks.
4783 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
4784 on complex multiplication and division libcall builtins.
4786 2018-05-04 Richard Biener <rguenther@suse.de>
4789 * fold-const.c (negate_expr_p): Restrict negation of operand
4790 zero of a division to when we know that can happen without
4792 (fold_negate_expr_1): Likewise.
4794 2018-05-04 Jakub Jelinek <jakub@redhat.com>
4797 * real.h (real_nextafter): Declare.
4798 * real.c (real_nextafter): New function.
4799 * fold-const-call.c (fold_const_nextafter): New function.
4800 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
4801 CASE_CFN_NEXTTOWARD.
4802 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
4803 even when arg1_mode is different from arg0_mode.
4805 2018-05-03 Nathan Sidwell <nathan@acm.org>
4807 * doc/extend.texi (Deprecated Features): Remove
4809 (Backwards Compatibility): Likewise.
4810 * doc/invoke.texi (C++ Language Options): Likewise.
4811 (C++ Dialect Options): Likewise.
4813 2018-05-03 Jakub Jelinek <jakub@redhat.com>
4816 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
4817 _mm512_mask_mullox_epi64): New intrinsics.
4819 2018-05-03 Tom de Vries <tom@codesourcery.com>
4822 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
4823 dump files): Add offload-tree.
4825 2018-05-03 Richard Biener <rguenther@suse.de>
4827 PR tree-optimization/85615
4828 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
4829 to loops not nested in BBs loop father to avoid creating multi-entry
4832 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4834 PR tree-optimization/70291
4835 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
4836 arguments. Change return type to tree. Emit libcall as a new
4837 statement rather than replacing existing one when inplace_p is true.
4838 (expand_complex_multiplication_components): New function.
4839 (expand_complex_multiplication): Expand floating-point complex
4840 multiplication using the above.
4841 (expand_complex_division): Rename inner_type parameter to type.
4842 Update expand_complex_libcall call-site.
4843 (expand_complex_operations_1): Update expand_complex_multiplication
4844 and expand_complex_division call-sites.
4846 2018-05-02 Jakub Jelinek <jakub@redhat.com>
4849 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
4850 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
4851 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
4852 the highest significant bit of the shift count mask is clear. In
4853 check whether and[sq]i3 is needed verify that all significant bits
4854 of the shift count other than the highest are set.
4856 2018-05-02 Tom de Vries <tom@codesourcery.com>
4859 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
4860 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
4861 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
4862 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
4863 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
4864 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
4865 __builtin_goacc_parlevel_size.
4867 2018-05-02 Richard Biener <rguenther@suse.de>
4869 PR tree-optimization/85597
4870 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
4871 do not use split vect_get_vec_defs call but call vect_get_slp_defs
4874 2018-05-02 Tom de Vries <tom@codesourcery.com>
4877 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
4878 dump files): Add ltrans-tree.
4880 2018-05-02 Tom de Vries <tom@codesourcery.com>
4883 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
4884 dump files): Add wpa-ipa.
4886 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
4888 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
4889 powerpc*-*-linux*paired* target.
4890 * config/rs6000/750cl.h: Delete.
4891 * config/rs6000/paired.h: Delete.
4892 * config/rs6000/paired.md: Delete.
4893 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
4895 * config/rs6000/rs6000-builtin.def: Remove paired float support.
4896 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
4897 comment. Remove paired float support.
4898 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
4899 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
4901 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
4902 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
4904 * config/rs6000/rs6000.c: Remove paired float support.
4905 (paired_expand_vector_init, paired_expand_vector_move,
4906 paired_emit_vector_compare, paired_emit_vector_cond_expr,
4907 (paired_expand_lv_builtin, paired_expand_stv_builtin,
4908 paired_expand_builtin, paired_expand_predicate_builtin,
4909 paired_init_builtins): Delete.
4910 * config/rs6000/rs6000.h: Remove paired float support.
4911 * config/rs6000/rs6000.md: Remove paired float support.
4912 (move_from_CR_ov_bit): Delete.
4913 * config/rs6000/rs6000.opt (mpaired): Delete.
4914 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
4915 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
4917 2018-05-02 Richard Biener <rguenther@suse.de>
4920 * gimplify.c (gimplify_save_expr): When in SSA form allow
4921 SAVE_EXPRs to compute to SSA vars.
4923 2018-05-02 Jakub Jelinek <jakub@redhat.com>
4926 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
4927 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
4928 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
4929 clobber operands[2], instead use a new pseudo. Formatting fixes.
4931 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
4933 PR tree-optimization/85586
4934 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
4935 exit early for statements in the same group if the accesses are
4938 2018-05-02 Tom de Vries <tom@codesourcery.com>
4941 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
4944 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
4946 PR tree-optimization/85143
4947 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
4949 2018-05-01 Tom de Vries <tom@codesourcery.com>
4952 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
4953 not found" error message.
4955 2018-05-01 Tom de Vries <tom@codesourcery.com>
4958 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
4959 * vec.c (test_ordered_remove_if): New function.
4960 (vec_c_tests): Call test_ordered_remove_if.
4961 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
4962 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
4963 * tree-vect-patterns.c (vect_pattern_recog_1): Use
4964 VEC_ORDERED_REMOVE_IF.
4966 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4968 PR tree-optimization/82665
4969 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
4970 pointer subtraction where arguments come from a memchr call.
4972 2018-05-01 Jakub Jelinek <jakub@redhat.com>
4974 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
4975 --push-state --as-needed and --pop-state instead of --as-needed and
4976 --no-as-needed if ld supports it.
4977 * configure: Regenerated.
4980 * doc/install.texi2html: Replace _002d with - and _002a with * in
4981 generated html files using sed.
4983 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4986 * gcc-rich-location.c (blank_line_before_p): New function.
4987 (use_new_line): New function.
4988 (gcc_rich_location::add_fixit_insert_formatted): New function.
4989 * gcc-rich-location.h
4990 (gcc_rich_location::add_fixit_insert_formatted): New function.
4992 2018-04-30 David Malcolm <dmalcolm@redhat.com>
4994 * selftest.c (assert_streq): Rename "expected" and "actual" to
4995 "val1" and "val2". Extend NULL-handling to cover both inputs
4996 symmetrically, while still requiring both to be non-NULL for a pass.
4997 * selftest.h (assert_streq): Rename "expected" and "actual" to
4999 (ASSERT_EQ): Likewise.
5000 (ASSERT_EQ_AT): Likewise.
5001 (ASSERT_KNOWN_EQ): Likewise.
5002 (ASSERT_KNOWN_EQ_AT): Likewise.
5003 (ASSERT_NE): Likewise.
5004 (ASSERT_MAYBE_NE): Likewise.
5005 (ASSERT_MAYBE_NE_AT): Likewise.
5006 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
5007 the assertion to pass.
5008 (ASSERT_STREQ_AT): Likewise.
5010 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
5012 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
5013 interaction with -pie.
5015 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5017 * selftest.h: Fix alphabetization of per-source-file selftest
5020 2018-04-30 Jason Merrill <jason@redhat.com>
5022 PR c++/61982 - dead stores to destroyed objects.
5023 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
5026 2018-04-30 Jason Merrill <jason@redhat.com>
5028 * tree.c (build_clobber): New.
5029 * tree.h: Declare it.
5030 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
5032 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5034 * diagnostic-show-locus.c (layout::layout): Update for
5035 location_get_source_line returning a char_span.
5036 (struct char_span): Move to input.h.
5037 (struct correction): Update for fields in char_span becoming
5039 (struct source_line): Update for location_get_source_line
5040 returning a char_span.
5041 (layout::print_line): Likewise.
5042 * edit-context.c (edited_file::print_content): Likewise.
5043 (edited_file::print_diff_hunk): Likewise.
5044 (edited_file::print_run_of_changed_lines): Likewise.
5045 (edited_file::get_num_lines): Likewise.
5046 (edited_line::edited_line): Likewise.
5047 * final.c (asm_show_source): Likewise.
5048 * input.c (location_get_source_line): Convert return type
5049 from const char * to char_span, losing the final "line_len"
5051 (dump_location_info): Update for the above.
5052 (get_substring_ranges_for_loc): Likewise. Use a char_span
5053 when handling the literal within the line.
5054 (test_reading_source_line): Update for location_get_source_line
5055 returning a char_span.
5056 * input.h (class char_span): Move here from
5057 diagnostic-show-locus.c, converting from a struct to a class.
5058 Make data members private.
5059 (char_span::operator bool): New.
5060 (char_span::length): New.
5061 (char_span::get_buffer): New.
5062 (char_span::operator[]): New.
5063 (char_span::subspan): Make const.
5064 (char_span::xstrdup): New.
5065 (location_get_source_line): Convert return type from const char *
5066 to char_span, losing the final "line_size" param.
5068 2018-04-30 Jan Hubicka <jh@suse.cz>
5070 * lto-wrapper.c (ltrans_priorities): New static var.
5071 (cmp_priority): New.
5072 (run_gcc): Read priorities and if doing parallel build order
5073 the Makefile by them.
5075 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5077 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
5079 2018-04-30 Richard Biener <rguenther@suse.de>
5081 * tree-cfg.c (verify_address): Remove base argument, add
5082 flag whether to check TREE_ADDRESSABLE and do that.
5083 (verify_expr): Remove.
5084 (verify_types_in_gimple_reference): Add pieces from verify_expr.
5085 (verify_gimple_assign_single): Likewise.
5086 (verify_gimple_switch): Likewise.
5087 (verify_expr_location_1): Dereference tp once. Add (disabled)
5088 piece from verify_expr.
5089 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
5091 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5093 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
5095 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5097 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
5098 (small_data_pattern): Likewise.
5099 (arc_rewrite_small_data): Likewise.
5100 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
5101 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
5102 (get_symbol_alignment): New function.
5103 (legitimate_small_data_address_p): Likewise.
5104 (legitimate_scaled_address): Update, call
5105 legitimate_small_data_address_p.
5106 (output_sdata): New static variable.
5107 (arc_print_operand): Update how we handle small data operands.
5108 (arc_print_operand_address): Likewise.
5109 (arc_legitimate_address_p): Update, use
5110 legitimate_small_data_address_p.
5111 (arc_rewrite_small_data_p): Remove.
5112 (arc_rewrite_small_data_1): Likewise.
5113 (arc_rewrite_small_data): Likewise.
5114 (small_data_pattern): Likewise.
5115 (compact_sda_memory_operand): Update to use
5116 legitimate_small_data_address_p and get_symbol_alignment.
5117 (prepare_move_operands): Don't rewite sdata pattern.
5118 (prepare_extend_operands): Remove.
5119 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
5121 (zero_extendqisi2): Likewise.
5122 (zero_extendhisi2): Likewise.
5123 (extendqihi2): Likewise.
5124 (extendqisi2): Likewise.
5125 (extendhisi2): Likewise.
5129 * config/arc/constraints.md (Usd): Change it to memory constraint.
5131 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5133 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
5134 as source of std instructions.
5135 * config/arc/arc.md (movsi_insn): Update pattern predicate to
5136 allow 6-bit constants as source for store instructions.
5137 (movdi_insn): Update instruction pattern to allow 6-bit constants
5138 as source for store instructions.
5140 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
5142 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
5144 2018-04-30 Nathan Sidwell <nathan@acm.org>
5145 Sandra Loosemore <sandra@codesourcery.com>
5147 * dumpfile.c (dump_open): Allow '-' for stdout.
5148 * doc/invoke.texi (Developer Options): Document dump filename
5149 determination early. Document stdin/stdout selection.
5151 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
5153 Microblaze Target: PIC data text relative
5155 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
5156 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
5158 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
5159 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
5160 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
5161 New addressing mode for data-text relative position indepenedent code.
5162 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
5163 'ADDRESS_SYMBOLIC_TXT_REL'.
5164 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
5165 (microblaze_legitimate_pic_operand): Exclude function calls from
5166 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
5167 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
5169 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
5170 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
5171 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
5172 for 'address + offset'.
5173 (microblaze_expand_prologue): Add new function prologue call for
5175 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
5176 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
5177 table in case of TARGET_PIC_DATA_TEXT_REL.
5178 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
5179 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
5180 Add new macros 'UNSPEC_TEXT',
5181 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
5182 + exclude function calls from 'UNSPEC_PLT' in case of data text
5184 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
5185 new target hook for generating address diff vector tables in case of
5187 * doc/tm.texi : Regenerate.
5188 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
5189 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
5190 of addr diff vector generation.
5191 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
5192 target hook definition.
5193 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
5194 Add default function for generate_pic_addr_diff_vec -> flag_pic.
5195 * doc/invoke.texi (Add new pic option): Add new microblaze pic
5196 option for data text relative.
5198 2018-04-30 Richard Biener <rguenther@suse.de>
5200 * tree-chrec.h (evolution_function_is_constant_p): Remove
5202 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
5204 2018-04-30 Richard Biener <rguenther@suse.de>
5207 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
5209 2018-04-30 Richard Biener <rguenther@suse.de>
5211 PR tree-optimization/28364
5212 PR tree-optimization/85275
5213 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
5214 copying first exit test.
5216 2018-04-28 Mark Wielaard <mark@klomp.org>
5218 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
5220 (dwarf_AT): Handle DW_AT_addr_base.
5221 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
5223 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
5226 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
5227 (*ashl<dwi>3_doubleword_mask_1): Ditto.
5228 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
5229 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
5231 2018-04-28 Richard Biener <rguenther@suse.de>
5233 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
5234 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
5235 to reflect use. Only add interesting stmts.
5237 2018-04-27 Martin Jambor <mjambor@suse.cz>
5240 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
5241 the jump function allows for passing through aggregate values.
5243 2018-04-27 David Malcolm <dmalcolm@redhat.com>
5245 * input.h (in_system_header_at): Convert from macro to inline
5247 (from_macro_expansion_at): Likewise.
5248 (from_macro_definition_at): Likewise.
5250 2018-04-27 Jeff Law <law@redhat.com>
5252 * config.gcc: Mark tile* targets as deprecated/obsolete.
5254 2018-04-27 Richard Biener <rguenther@suse.de>
5256 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
5259 2018-04-27 Richard Biener <rguenther@suse.de>
5261 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
5263 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
5265 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
5266 with Yd constraint. Set "preferred_for_speed" attribute from
5267 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
5269 (*movdi_internal): Ditto.
5270 (movti_interunit splitters): Remove
5271 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
5272 (movdi_interunit splitters): Ditto.
5273 * config/i386/constraints.md (Ye): Remove.
5274 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
5276 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5279 * config/aarch64/constraints.md (Usg): Limit to 31.
5282 2018-04-27 Jakub Jelinek <jakub@redhat.com>
5284 PR tree-optimization/85529
5285 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
5286 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
5287 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
5288 zero extension or masking of the MSB bit.
5289 (optimize_range_tests): Add FIRST_BB argument, pass it through
5290 to optimize_range_tests_var_bound.
5291 (maybe_optimize_range_tests, reassociate_bb): Adjust
5292 optimize_range_tests callers.
5294 2018-04-26 Richard Biener <rguenther@suse.de>
5295 Jakub Jelinek <jakub@redhat.com>
5297 * cgraph.h (symbol_table): Just declare debug method here.
5298 * symtab.c (symbol_table::debug): Define.
5300 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
5302 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
5304 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
5306 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
5307 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
5308 (*movdi_internal): Substitute Yi and Yj constraint with x
5309 and Ym and Yn constraint with y constraint. Update "isa"
5310 attribute and set "preferred_for_speed" attribute from
5311 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
5312 (*movsi_internal): Ditto.
5313 (*movdf_internal): Ditto.
5314 (*movsf_internal): Ditto.
5315 (*zero_extendsidi2): Ditto.
5316 * config/i386/sse.md (vec_set<mode>_0): Ditto.
5317 (sse2_loadld): Ditto.
5318 (*vec_extract<ssevecmodelower>_0): Ditto.
5319 (*vec_extractv4si_0_zext_sse4): Ditto.
5320 (vec_concatv2di): Ditto.
5321 (*vec_dup<mode>): Ditto.
5322 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
5323 * config/i386/constraints.md (Yi): Remove.
5328 2018-04-26 Nathan Sidwell <nathan@acm.org>
5330 * dumpfile.c (dump_open): New.
5331 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
5332 (dump_finish): Detect stdio/stderr by value not name.
5334 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
5336 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
5338 2018-04-26 Tom de Vries <tom@codesourcery.com>
5341 * config/nvptx/nvptx.c (verify_neutering_jumps)
5342 (verify_neutering_labels): New function
5343 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
5345 2018-04-26 Tom de Vries <tom@codesourcery.com>
5348 * config/nvptx/nvptx.c (needs_neutering_p): New function.
5349 (nvptx_single): Use needs_neutering_p to skip over insns that do not
5352 2018-04-26 Richard Biener <rguenther@suse.de>
5353 Tom de Vries <tom@codesourcery.com>
5356 * lto-streamer-out.c (output_function): Fixup loops if required to match
5357 discovery done in the reader.
5359 2018-04-26 Richard Biener <rguenther@suse.de>
5361 PR tree-optimization/85116
5362 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
5363 have a loop exit from the single latch predecessor. Remove
5364 case of header with just condition.
5365 (ch_base::copy_headers): Exclude infinite loops from any
5367 (pass_ch::execute): Record exits.
5369 2018-04-26 Richard Biener <rguenther@suse.de>
5371 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
5372 prologue cost vector and pass it to vect_get_load_cost.
5373 (vect_get_peeling_costs_all_drs): Likewise.
5374 (vect_peeling_hash_get_lowest_cost): Likewise.
5375 (vect_enhance_data_refs_alignment): Likewise.
5377 2018-04-26 Richard Biener <rguenther@suse.de>
5380 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
5381 checking of integer<->pointer conversions.
5382 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
5383 sign-/zero-extending pointer types.
5384 (expand_omp_for_static_chunk): Likewise.
5386 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
5387 Jean Lee <xiaoyur347@gmail.com>
5389 * config/mips/mips.c (mips_asan_shadow_offset): New function.
5390 (TARGET_ASAN_SHADOW_OFFSET): Define.
5391 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
5392 true for -fsanitize=address.
5394 2018-04-25 Mark Wielaard <mark@klomp.org>
5396 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
5399 2018-04-25 Jakub Jelinek <jakub@redhat.com>
5401 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
5402 than "alu", remove explicit "memory" and "imm_disp" attributes.
5403 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
5406 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
5407 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
5408 gen_lowpart_no_emit.
5410 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
5413 * config/i386/i386.c (ix86_expand_builtin): Change memory
5414 operand to XI, extend p0 to Pmode.
5415 * config/i386/i386.md: Change unspec volatile and operand
5416 1 mode to XI, change operand 0 mode to P.
5418 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5420 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
5421 GET_MODE_MASK before any checking.
5422 (nds32_can_use_bset_p): Likewise.
5423 (nds32_can_use_btgl_p): Likewise.
5425 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5427 * config/nds32/nds32-doubleword.md: New define_split pattern for
5428 illegal register number.
5430 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5432 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
5434 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5436 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
5438 2018-04-25 Richard Biener <rguenther@suse.de>
5440 * lto-streamer.h (LTO_major_version): Bump to 8.
5442 2018-04-25 Jakub Jelinek <jakub@redhat.com>
5444 * BASE-VER: Set to 9.0.0.
5446 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
5448 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
5449 in __abskf2 and __powikf2.
5451 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5454 * config/aarch64/constraints.md (Usg, Usj): New constraints.
5455 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
5456 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
5457 Use the above on operand 2. Reindent.
5458 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
5460 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
5463 * common/config/i386/i386-common.c (ix86_handle_option): Don't
5465 * config/i386/i386.opt (mcet): Removed.
5466 * doc/install.texi: Remove -mcet documentation.
5467 * doc/invoke.texi: Likewise.
5469 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
5472 * doc/install.texi: Remove -mcet from bootstrap-cet.
5474 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5477 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
5478 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
5482 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
5483 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
5484 containing a CONST_VECTOR.
5486 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
5488 * doc/install.texi: Update newlib dependency for nvptx.
5490 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5493 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
5494 instead of INTVAL when shifting x left.
5496 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
5498 PR tree-optimization/85478
5499 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
5500 vect_grouped_store_supported for single element vectors.
5502 2018-04-24 Richard Biener <rguenther@suse.de>
5505 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
5506 load cost increase to the case of non-constant step.
5508 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5511 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
5512 destination if any_malformed_asm.
5514 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
5517 * expr.c (store_field): In the bitfield case, if the value comes from
5518 a function call and is returned in registers by means of a PARALLEL,
5519 do not change the mode of the temporary unless BLKmode and VOIDmode.
5521 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
5523 PR rtl-optimization/85423
5524 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
5525 dependencies to debug insns when the previous insn is non-debug.
5527 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
5529 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
5530 enums into a single definition.
5531 (fls): Fix predicates and printing.
5534 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
5536 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
5537 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
5538 and short u6 immediate.
5539 (check_if_valid_sleep_operand): Remove.
5540 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
5542 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5544 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
5545 flag_always_save_lp condition.
5546 * config/nds32/nds32.opt (malways-save-lp): New option.
5548 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5550 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
5551 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
5552 * config/nds32/nds32.h
5553 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
5554 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
5556 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5558 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
5559 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
5561 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5562 Chung-Ju Wu <jasonwucj@gmail.com>
5564 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
5566 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
5567 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
5569 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5571 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
5573 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5575 * config/nds32/nds32-protos.h (nds32_data_alignment,
5576 nds32_local_alignment): Declare.
5577 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
5578 nds32_local_alignment): New functions.
5579 (TARGET_CONSTANT_ALIGNMENT): Define.
5580 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
5582 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5584 * config/nds32/nds32.c
5585 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
5586 (TARGET_MODES_TIEABLE_P): Likewise.
5588 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5590 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
5593 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
5594 Chung-Ju Wu <jasonwucj@gmail.com>
5596 * config/nds32/constants.md (unspec_volatile_element): Add enum values
5597 for unaligned access.
5598 * config/nds32/nds32-intrinsic.c: Implementation of expanding
5600 * config/nds32/nds32-intrinsic.md: Likewise.
5601 * config/nds32/nds32_intrinsic.h: Likewise.
5602 * config/nds32/nds32.h (nds32_builtins): Likewise.
5603 * config/nds32/nds32.opt (munaligned-access): New option.
5604 * config/nds32/nds32.c (nds32_asm_file_start): Display
5605 flag_unaligned_access status.
5607 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
5609 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
5610 -mno-relax is present.
5611 * config/riscv/linux.h (LINK_SPEC): Ditto.
5613 2018-04-20 Martin Sebor <msebor@redhat.com>
5616 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
5618 (gimple_fold_builtin_stxcpy_chk): Same.
5619 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
5621 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
5624 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
5625 __powikf2 when long double is IEEE 128-bit.
5627 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
5629 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
5630 step to make sure stack always aligned.
5632 2018-04-20 Carl Love <cel@us.ibm.com>
5635 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
5636 size check for arg0.
5638 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
5639 Tom de Vries <tom@codesourcery.com>
5642 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
5643 Emit insns for calls too.
5644 (nvptx_find_par): Always look for worker-level predecessor insn.
5645 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
5647 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
5648 (nvptx_process_pars): Propagate frames for calls.
5650 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
5653 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
5655 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
5656 (ix86_handle_option): Don't handle OPT_mibt.
5657 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
5659 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
5661 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
5662 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
5663 (ix86_target_macros): Define __CET__ with flag_cf_protection
5664 for -fcf-protection.
5665 * config/i386/i386.c (isa2_opts): Remove -mibt.
5666 * config/i386/i386.h (TARGET_IBT): Removed.
5667 (TARGET_IBT_P): Likewise.
5668 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
5669 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
5670 * config/i386/i386.opt (mcet): Update help message.
5673 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
5674 -mcet as an alias for -mshstk.
5676 2018-04-20 Richard Biener <rguenther@suse.de>
5679 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
5680 complexity by forcing a single use of the multiply operand.
5682 2018-04-20 Martin Jambor <mjambor@suse.cz>
5685 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
5686 recursion dependency to only apply to non-clones.
5688 2018-04-20 Martin Jambor <mjambor@suse.cz>
5691 * ipa-cp.c (create_specialized_node): Check that clones of
5692 self-recursive edges exist during IPA-CP.
5694 2018-04-19 Toon Moene <toon@moene.org>
5696 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
5699 2018-04-19 Jakub Jelinek <jakub@redhat.com>
5701 PR tree-optimization/85467
5702 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
5703 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
5704 VECTOR_CST element to type.
5706 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
5709 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
5710 * config/i386/i386.md (builtin_setjmp_setup): Removed.
5711 (builtin_longjmp): Likewise.
5712 (save_stack_nonlocal): New pattern.
5713 (restore_stack_nonlocal): Likewise.
5715 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
5718 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
5719 Replace ASM_OUTPUT_LABEL with fprintf.
5721 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
5724 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
5725 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
5726 * config/i386/i386-c.c (ix86_target_macros_internal): Also
5727 define __IBT__ and __SHSTK__ for -fcf-protection.
5728 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
5730 (ix86_trampoline_init): Likewise.
5731 (x86_output_mi_thunk): Likewise.
5732 (ix86_notrack_prefixed_insn_p): Likewise.
5733 (ix86_option_override_internal): Don't disallow -fcf-protection.
5734 * config/i386/i386.md (rdssp<mode>): Also enable for
5736 (incssp<mode>): Likewise.
5737 (nop_endbr): Likewise.
5738 * config/i386/i386.opt (mcet): Change help message to built-in
5742 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
5743 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
5744 enable CET built-in functions.
5746 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
5748 * common/config/i386/i386-common.c
5749 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
5750 OPTION_MASK_ISA_MOVDIRI_UNSET,
5751 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
5752 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
5753 * config.gcc (movdirintrin.h): New header.
5754 * config/i386/cpuid.h (bit_MOVDIRI,
5755 bit_MOVDIR64B): New bits.
5756 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
5758 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
5759 (VOID, PVOID, PCVOID)): New function types.
5760 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
5761 __builtin_ia32_directstoreu_u64,
5762 __builtin_ia32_movdir64b): New builtins.
5763 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
5764 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
5766 (ix86_valid_target_attribute_inner_p): Ditto.
5767 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
5768 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
5769 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
5770 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
5771 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
5772 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
5773 (movdiri<mode>, movdir64b_<mode>): New.
5774 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
5775 * config/i386/immintrin.h: Include movdirintrin.h.
5776 * config/i386/movdirintrin.h: New file.
5777 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
5779 2018-04-19 Richard Biener <rguenther@suse.de>
5782 * cfg.c (clear_bb_flags): When loop state says we have
5783 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
5785 2018-04-19 Richard Biener <rguenther@suse.de>
5787 PR tree-optimization/84737
5788 * tree-vect-data-refs.c (vect_copy_ref_info): New function
5789 copying restrict info.
5790 (vect_setup_realignment): Use it.
5791 * tree-vectorizer.h (vect_copy_ref_info): Declare.
5792 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
5793 the first DR to all generated stores.
5794 (vectorizable_load): Likewise for loads.
5796 2018-04-19 Jakub Jelinek <jakub@redhat.com>
5798 PR tree-optimization/85446
5799 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
5800 the integral and pointer types to have the same precision.
5802 * doc/install.texi: Document --disable-cet being the default and
5805 2018-04-18 Martin Liska <mliska@suse.cz>
5807 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
5810 2018-04-18 Martin Liska <mliska@suse.cz>
5813 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5816 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5817 arguments if they are comparable.
5819 2018-04-18 Martin Liska <mliska@suse.cz>
5822 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
5825 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
5828 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
5831 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
5832 ENDBR after calling __morestack.
5834 2018-04-18 David Malcolm <dmalcolm@redhat.com>
5837 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
5838 by using gcc_base_ver to generate a gcc_driver_version, and use
5839 it when generating GCC_DRIVER_NAME.
5840 * configure: Regenerate.
5842 2018-04-18 Jakub Jelinek <jakub@redhat.com>
5845 * config.gcc: Obsolete powerpc*-*-*spe*.
5847 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5850 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
5851 (stabstr_D): Change type of unum from unsigned int to
5852 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
5855 2018-04-17 Jim Wilson <jimw@sifive.com>
5858 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
5859 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
5860 Set arg_pointer_offset after using pretend_args_size.
5862 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5864 PR rtl-optimization/85431
5865 * dse.c (record_store): Ignore zero width stores.
5868 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
5869 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
5870 __builtin_stack_restore rather than after it.
5871 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
5872 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
5873 argument instead of virtual_dynamic_stack_rtx.
5875 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
5877 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
5879 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
5880 Add note to error message to explain internal mapping of overloaded
5881 built-in function name to non-overloaded built-in function name.
5882 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
5885 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
5888 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
5889 where the inputs overlap with the output.
5891 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5893 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
5894 (=v, v) alternative and explicit "memory" attribute.
5895 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
5896 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5898 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
5899 "sselog1" type instead of "sselog".
5900 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
5901 "sselog". Remove explicit "memory" attribute.
5902 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
5903 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5905 (vec_extract_hi_v32hi): Merge all alternatives into one, use
5906 "sselog1" type instead of "sselog". Remove explicit "memory"
5908 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
5909 use "sselog1" type instead of "sselog". Remove explicit "memory"
5911 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
5912 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
5914 (vec_extract_hi_v64qi): Merge all alternatives into one, use
5915 "sselog1" type instead of "sselog". Remove explicit "memory"
5917 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
5918 use "sselog1" type instead of "sselog". Remove explicit "memory"
5922 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
5925 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
5928 2018-04-17 Martin Jambor <mjambor@suse.cz>
5931 * ipa-cp.c (create_specialized_node): Call
5932 expand_all_artificial_thunks if necessary.
5934 2018-04-17 Martin Liska <mliska@suse.cz>
5937 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
5938 in message, remote space in between '_G' and '('.
5940 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5943 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
5944 avx512f_vmcmp<mode>3<round_saeonly_name>,
5945 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
5946 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
5947 avx512f_rndscale<mode><round_saeonly_name>,
5948 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
5949 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
5950 Use %<iptr>2 instead of %2 for -masm=intel.
5951 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
5952 avx512f_vcvttss2usi<round_saeonly_name>,
5953 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
5955 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
5956 avx512f_vcvttsd2usi<round_saeonly_name>,
5957 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
5958 Use %q1 instead of %1 for -masm=intel.
5959 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
5960 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
5961 of %3 for -masm=intel.
5962 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
5964 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
5966 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
5968 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
5969 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
5971 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
5973 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
5974 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
5975 %g1 and one with %0 and %1.
5976 (avx512er_vmrcp28<mode><round_saeonly_name>,
5977 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
5979 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
5980 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
5981 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
5982 of %0 and %{%4%} for -masm=intel.
5983 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
5984 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
5985 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
5986 order of %0 and %{%5%}%{z%} for -masm=intel.
5988 2018-04-17 Jan Hubicka <jh@suse.cz>
5991 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
5993 2018-04-17 Martin Liska <mliska@suse.cz>
5996 * multiple_target.c (create_dispatcher_calls): Set apostrophes
5997 for target_clone error message. Make default implementation
5998 clone to be a local declaration.
5999 (separate_attrs): Add new argument and check for an empty
6001 (expand_target_clones): Handle it.
6002 (ipa_target_clone): Make redirection just for target_clones
6005 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
6006 Tom de Vries <tom@codesourcery.com>
6009 * omp-expand.c (expand_oacc_for): Add dummy false branch for
6010 tiled basic blocks without omp continue statements.
6012 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
6015 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
6016 vec_extract expression as having side effects to make sure it gets
6019 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
6022 * config/i386/i386.c (get_builtin_code_for_version): Check
6025 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
6028 * gcc/config.gcc: Support "skylake".
6029 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
6031 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
6032 (processor_target_table): Add "skylake".
6033 (ix86_option_override_internal): Add "skylake".
6034 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
6035 PROCESSOR_CANNONLAKE.
6036 (get_builtin_code_for_version): Fix priority for
6037 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
6038 PROCESSOR_SKYLAKE-AVX512.
6039 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
6040 (processor_type): Add PROCESSOR_SKYLAKE.
6042 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
6043 Jason Merrill <jason@redhat.com>
6046 * convert.c (convert_to_integer_1): Use direct recursion for
6047 enumeral types and types with a precision less than the number
6048 of bits in their mode.
6050 2018-04-16 Julia Koval <julia.koval@intel.com>
6053 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
6054 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
6056 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
6059 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
6060 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
6061 and -mno-direct-move.
6063 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
6066 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
6067 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
6068 Ensure negative shifts result in {0}.
6070 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
6072 PR rtl-optimization/79916
6073 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
6074 regs (if any) to define how to gnerate SD moves when LRA is in
6077 2018-04-13 Jakub Jelinek <jakub@redhat.com>
6079 PR rtl-optimization/85393
6080 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
6081 * except.c (expand_dw2_landing_pad_for_region): Make static.
6082 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
6083 a label and unconditional jump to old_bb, rather than
6084 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
6087 PR rtl-optimization/85376
6088 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
6089 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
6090 instead of a specific value.
6092 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
6093 Bin Cheng <bin.cheng@arm.com>
6095 PR tree-optimization/82965
6096 PR tree-optimization/83991
6097 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
6098 by_profile_only parameter.
6099 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
6100 information if the loop was predicted to iterate too many times.
6101 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
6103 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
6106 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
6109 2018-04-13 Martin Liska <mliska@suse.cz>
6110 Jakub Jelinek <jakub@redhat.com>
6113 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
6114 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
6115 * builtins.c (expand_builtin_memory_copy_args): Use
6116 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
6117 handle dest_addr == pc_rtx.
6119 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
6122 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
6123 asked to not generate direct moves.
6124 (fix_trunc<mode>si2_stfiwx): Similar.
6125 (fix_trunc<mode>si2_internal): Similar.
6127 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6130 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
6131 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
6132 lookup if dest in some wider mode is known to be const0_rtx and
6133 if so, record permanent equivalence for it to be ZERO_EXTEND of
6134 the narrower mode destination.
6136 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
6138 * lto-streamer-out.c (output_function): Revert 259346.
6139 * omp-expand.c (expand_oacc_for): Likewise.
6141 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
6143 PR rtl-optimization/85354
6144 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
6145 * sel-sched.c (sel_global_init): ... here.
6147 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
6150 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
6151 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
6152 mode for PE-COFF targets.
6153 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
6154 (i386_pe_asm_lto_end): Likewise.
6155 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
6156 (TARGET_ASM_LTO_END): Likewise.
6157 * config/i386/winnt.c (saved_debug_info_level): New static variable.
6158 (i386_pe_asm_lto_start): New function.
6159 (i386_pe_asm_lto_end): Likewise.
6161 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
6162 Richard Biener <rguenther@suse.de>
6165 * lto-streamer-out.c (output_function): Fix CFG loop state before
6167 * omp-expand.c (expand_oacc_for): Handle calls to internal
6168 functions like regular functions.
6170 2018-04-12 Richard Biener <rguenther@suse.de>
6173 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
6174 for the early LTO debug to properly generate references to it
6175 during DIE emission. Do not re-use that for the skeleton for
6177 (dwarf2out_early_finish): Likewise.
6179 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6182 * config/i386/sse.md
6183 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
6184 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
6185 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
6186 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
6187 and output is a reg, avoid creating invalid lowpart subreg, but
6188 instead split into a 512-bit move. Don't split if not AVX512VL,
6189 input is xmm16+ reg and output is a mem.
6190 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
6191 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
6192 xmm16+ reg and output is a mem.
6194 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6196 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
6197 also for flag_dwarf2_cfi_asm.
6199 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6201 PR rtl-optimization/85342
6202 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
6203 a bool scalar var inside of the loop instead. Don't try to update
6204 recog_data.operand after failed apply_change_group.
6206 2018-04-12 Tom de Vries <tom@codesourcery.com>
6209 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
6210 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
6211 array with flexible array member as array without given dimension.
6212 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
6213 argument for undefined param to true.
6215 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
6218 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
6219 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
6220 from PowerPC section.
6221 * config/rs6000/sysv4.opt (mcall-): Improve help text.
6222 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
6223 help text that is too long.
6224 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
6225 help text that is too long.
6226 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
6227 help text that is too long.
6229 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
6231 * config/alpha/alpha.md (stack_probe_internal): Rename
6232 from "probe_stack". Update all callers.
6234 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6236 PR rtl-optimization/84566
6237 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
6238 sched_macro_fuse_insns.
6240 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6243 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
6244 (compute_block_dependences): ... from here.
6246 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6248 PR tree-optimization/85331
6249 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
6250 from int to HOST_WIDE_INT.
6252 2018-04-11 Martin Jambor <mjambor@suse.cz>
6255 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
6256 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
6257 not the same as the source val.
6258 (cgraph_edge_brings_value_p): New parameter.
6259 (gather_edges_for_value): Pass destination value to
6260 cgraph_edge_brings_value_p.
6261 (perhaps_add_new_callers): Likewise.
6262 (get_info_about_necessary_edges): Likewise and exclude values brought
6263 only by self-recursive edges.
6264 (create_specialized_node): Redirect only clones of self-calling edges.
6265 (+self_recursive_pass_through_p): New function.
6266 (find_more_scalar_values_for_callers_subset): Use it.
6267 (find_aggregate_values_for_callers_subset): Likewise.
6268 (known_aggs_to_agg_replacement_list): Removed.
6269 (decide_whether_version_node): Re-calculate known constants for all
6270 remaining context clones.
6272 2018-04-11 Richard Biener <rguenther@suse.de>
6275 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
6276 from early DWARF output.
6277 (dwarf2out_early_finish): Output line info unconditionally into
6278 early DWARF and add reference to it.
6280 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6283 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
6284 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
6285 other than V2DFmode using iptr mode attribute.
6286 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
6288 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6290 PR rtl-optimization/84659
6291 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
6293 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6296 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
6298 (output_loc_list): Pass address of a dummy size variable even in the
6299 locview handling loop.
6300 (index_location_lists): Add comment on why skip_loc_list_entry can't
6303 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
6306 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
6309 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
6312 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
6313 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
6314 and -mstring-compare-inline-limit.
6316 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6319 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
6320 for stack clash protection in a register whenever we need it to be in
6323 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6325 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
6326 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
6328 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6331 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
6333 (mlong-double-): Ditto.
6334 * config/rs6000/sysv4.opt (msdata=): Ditto.
6335 (mtls-size=): Ditto.
6337 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
6339 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
6340 erroneous entries for
6341 "vector int vec_ldl (int, long int *)", and
6342 "vector unsigned int vec_ldl (int, unsigned long int *)".
6343 Add comments and entries for
6344 "vector bool char vec_ldl (int, bool char *)",
6345 "vector bool short vec_ldl (int, bool short *)",
6346 "vector bool int vec_ldl (int, bool int *)",
6347 "vector bool long long vec_ldl (int, bool long long *)",
6348 "vector pixel vec_ldl (int, pixel *)",
6349 "vector long long vec_ldl (int, long long *)",
6350 "vector unsigned long long vec_ldl (int, unsigned long long *)".
6351 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
6352 type tree bool_long_long_type_node and correct definition of
6353 bool_V2DI_type_node to make reference to this new type tree.
6354 (rs6000_mangle_type): Replace erroneous reference to
6355 bool_long_type_node with bool_long_long_type_node.
6356 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
6357 comments to emphasize sign distinctions for char and int types and
6358 replace RS6000_BTI_bool_long constant with
6359 RS6000_BTI_bool_long_long constant. Also add comment to restrict
6360 use of RS6000_BTI_pixel.
6361 (bool_long_type_node): Remove this macro definition.
6362 (bool_long_long_type_node): New macro definition
6364 2018-04-10 Jakub Jelinek <jakub@redhat.com>
6366 PR rtl-optimization/85300
6367 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
6368 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
6369 simplify_unary_operation fails.
6371 2018-04-10 Martin Liska <mliska@suse.cz>
6373 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
6374 cgraph_edge and ipa_ref.
6376 2018-04-10 Jakub Jelinek <jakub@redhat.com>
6380 * config/i386/sse.md
6381 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
6382 computation of the VEC_MERGE selector from mask.
6383 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
6384 Fix decoding of the VEC_MERGE selector into mask.
6386 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
6388 PR tree-optimization/85286
6389 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
6391 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
6393 * final.c (final_1): Set insn_last_address as well as
6394 insn_current_address.
6396 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6399 * explow.c (emit_stack_probe): Call validize_mem on memory location
6400 before passing it to gen_probe_stack. Create address operand and
6401 legitimize it for the probe_stack_address case.
6403 2018-04-09 Jan Hubicka <jh@suse.cz>
6406 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
6407 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
6408 * tree.c (free_lang_data_in_type): Fix handling of binfos;
6410 (free_lang_data): Rebuild type inheritance graph.
6412 2018-04-09 Martin Sebor <msebor@redhat.com>
6414 * invoke.texi (-finline-small-functions): Mention other optimization
6416 (-findirect-inlining, -fpartial-inlining): Same.
6417 (-finline-functions-called-once): Same.
6418 (-freorder-blocks-and-partition): Same.
6420 2018-04-09 Jan Hubicka <jh@suse.cz>
6423 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
6424 jumps; choose last target that matches the criteria (i.e.
6425 no partition changes for non-crossing jumps).
6426 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
6427 support for redirecting crossing jumps to non-crossing.
6429 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
6431 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
6432 also for naked functions.
6434 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
6436 * config/arc/arc.md (add_shift): New pattern.
6437 (add_shift2): Likewise.
6438 (sub_shift): Likewise.
6439 (sub_shift_cmp0_noout): Likewise.
6440 (compare_si_ashiftsi): Likewise.
6441 (xbfu_cmp0_noout): New combine pattern.
6442 (xbfu_cmp0"): Likewise.
6443 (movsi_set_cc_insn): Place the predicable variant first.
6444 (commutative_binary_cmp0_noout): Remove clobber.
6445 (commutative_binary_cmp0): New pattern.
6446 (noncommutative_binary_cmp0): Likewise.
6447 (noncommutative_binary_cmp0_noout): Likewise.
6448 (noncommutative_binary_comparison_result_used): Removed.
6449 (rsub_cmp0): New pattern.
6450 (rsub_cmp0_noout): Likewise.
6451 (extzvsi): Changed, keep only meaningful variants.
6452 (SQH, SEZ): New iterators.
6453 (SQH_postfix): New mode attribute.
6454 (SEZ_prefix): New code attribute.
6455 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
6456 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
6457 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
6459 (noncommutative_operator): Check the availability of barrel
6462 2018-04-09 Richard Biener <rguenther@suse.de>
6464 PR tree-optimization/85284
6465 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
6466 Only use the niter constraining form of simple_iv when the exit
6469 2018-04-09 Tom de Vries <tom@codesourcery.com>
6472 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
6473 (define_expand "*memory_barrier"): New define_expand.
6474 (define_insn "memory_barrier"): New insn.
6476 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6478 PR rtl-optimization/80463
6479 PR rtl-optimization/83972
6480 PR rtl-optimization/83480
6482 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
6483 correct producer for the insn.
6484 (tidy_control_flow): Fixup seqnos in case of debug insns.
6486 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6488 PR rtl-optimization/83913
6490 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
6491 different sched-times when merging exprs.
6493 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6495 PR rtl-optimization/83962
6497 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
6498 tidy_fallthru_edge and tidy_control_flow.
6500 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6502 PR rtl-optimization/83530
6504 * sel-sched.c (force_next_insn): New global variable.
6505 (remove_insn_for_debug): When force_next_insn is true, also leave only
6506 next insn in the ready list.
6507 (sel_sched_region): When the region wasn't scheduled, make another pass
6508 over it with force_next_insn set to 1.
6510 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
6512 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
6514 * config/nds32/constants.md (unspec_volatile_element): Add enum values
6515 for interrupt control.
6516 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
6517 functions for interrupt control.
6518 * config/nds32/nds32-intrinsic.md: Likewise.
6519 * config/nds32/nds32_intrinsic.h: Likewise.
6520 * config/nds32/nds32.h (nds32_builtins): Likewise.
6522 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
6524 * config/nds32/nds32.c (nds32_init_machine_status,
6525 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
6526 strict_aligned_p field.
6527 (nds32_expand_to_rtl_hook): New function.
6528 (TARGET_EXPAND_TO_RTL_HOOK): Define.
6529 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
6531 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6532 Chung-Ju Wu <jasonwucj@gmail.com>
6534 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
6535 * config/nds32/nds32-n7.md: New file.
6536 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
6537 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
6539 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
6540 * config/nds32/nds32.md (pipeline_model): Add n7.
6541 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
6542 * config/nds32/pipelines.md: Include n7 settings.
6544 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6545 Chung-Ju Wu <jasonwucj@gmail.com>
6547 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
6548 * config/nds32/nds32-e8.md: New file.
6549 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
6550 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
6552 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
6553 * config/nds32/nds32.md (pipeline_model): Add e8.
6554 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
6555 * config/nds32/pipelines.md: Include e8 settings.
6557 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6558 Chung-Ju Wu <jasonwucj@gmail.com>
6560 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
6561 * config/nds32/nds32-n8.md: New file.
6562 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
6563 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
6565 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
6566 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
6567 * config/nds32/nds32.md (pipeline_model): Add n8.
6568 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
6569 * config/nds32/pipelines.md: Include n8 settings.
6571 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6572 Chung-Ju Wu <jasonwucj@gmail.com>
6574 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
6575 * config/nds32/nds32-n9-2r1w.md: New file.
6576 * config/nds32/nds32-n9-3r2w.md: New file.
6577 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
6578 nds32_register_ports): New or modify for cpu n9.
6579 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
6581 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
6582 * config/nds32/nds32-utils.c: New file.
6583 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
6584 TARGET_MUL_SLOW): Define.
6585 * config/nds32/nds32.md (pipeline_model): New attribute.
6586 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
6587 New options that support cpu n9.
6588 * config/nds32/pipelines.md: Include n9 settings.
6589 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
6591 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
6593 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
6594 information if necessary.
6595 (output_cond_branch_compare_zero): Likewise.
6596 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
6597 (nds32_target_alignment): Refine for alignment.
6598 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
6599 (FUNCTION_BOUNDARY): Modify.
6600 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
6602 * config/nds32/nds32.opt (malways-align, malign-functions): New.
6604 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
6606 * config/nds32/constants.md (unspec_volatile_element): Add values for
6607 TLB operation and data prefetch.
6608 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
6609 functions for TLB operation and data prefetch.
6610 * config/nds32/nds32-intrinsic.md: Likewise.
6611 * config/nds32/nds32_intrinsic.h: Likewise.
6612 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
6613 (nds32_print_operand): Likewise.
6614 * config/nds32/nds32.h (nds32_builtins): Likewise.
6616 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
6617 Andrew Pinski <pinsika@gcc.gnu.org>
6620 * match.pd: Use constant_boolean_node of correct type instead of
6621 boolean_true_node or boolean_false_node for simplifying
6622 pointer comparisons to zero.
6624 2018-04-07 Jakub Jelinek <jakub@redhat.com>
6626 PR tree-optimization/80021
6627 * tree.c (verify_type_variant): Make error call in verify_variant_match
6628 translatable and remove final full stop.
6630 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6632 * config/nds32/constants.md (unspec_volatile_element): Add
6633 UNSPEC_VOLATILE_EH_RETURN.
6634 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
6635 nds32_output_stack_pop): Support dwarf exception handling process.
6636 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
6637 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
6638 exception handling process.
6639 (nds32_compute_stack_frame): Likewise.
6640 (nds32_return_addr_rtx): Likewise.
6641 (nds32_initial_elimination_offset): Likewise.
6642 (nds32_expand_prologue): Likewise.
6643 (nds32_expand_epilogue): Likewise.
6644 (nds32_dynamic_chain_address): New function.
6645 * config/nds32/nds32.h (machine_function): Add fields for dwarf
6647 (DYNAMIC_CHAIN_ADDRESS): Define.
6648 (EH_RETURN_DATA_REGNO): Define.
6649 (EH_RETURN_STACKADJ_RTX): Define.
6650 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
6651 patterns for dwarf exception handling.
6653 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6655 * config/nds32/nds32.h: Clean up obsolete macros.
6657 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6659 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6660 Add enum values for particular instructions.
6661 * config/nds32/nds32-intrinsic.c: Implementation of expanding
6662 particular intrinsic functions.
6663 * config/nds32/nds32-intrinsic.md: Likewise.
6664 * config/nds32/nds32_intrinsic.h: Likewise.
6665 * config/nds32/nds32.h (nds32_builtins): Likewise.
6666 * config/nds32/nds32.md (type): Add pbsad and pbsada.
6667 (btst, ave): New patterns for particular instructions.
6669 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6671 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6672 Add enum values for atomic load/store and memory sync.
6673 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
6675 * config/nds32/nds32-intrinsic.md: Likewise.
6676 * config/nds32/nds32_intrinsic.h: Likewise.
6677 * config/nds32/nds32.h (nds32_builtins): Likewise.
6679 2018-04-07 Jakub Jelinek <jakub@redhat.com>
6681 PR tree-optimization/85257
6682 * fold-const.c (native_encode_vector): If not all elts could fit
6683 and off is -1, return 0 rather than offset.
6684 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
6685 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
6686 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
6687 adjust buffer in native_interpret_expr call.
6689 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6691 * config/nds32/constants.md (unspec_volatile_element): Add cache
6692 control enum values.
6693 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
6694 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
6695 * config/nds32/nds32.c (nds32_cctl_names): New.
6696 (nds32_print_operand): Handle cache control register names.
6697 * config/nds32/nds32.h (nds32_builtins): New enum values.
6698 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
6700 * config/nds32/nds32.md (type): Add mmu.
6701 * config/nds32/pipelines.md (simple_insn): Add mmu.
6703 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6705 * config/nds32/nds32.md (type): Remove call.
6706 * config/nds32/pipelines.md (simple_insn): Likewise.
6708 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6710 * config/nds32/constants.md (unspec_volatile_element): Add
6711 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
6712 UNSPEC_VOLATILE_FMFCFG.
6713 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
6714 description for fmfcfg and fmfcsr.
6715 (bdesc_1arg): Add fmtcsr.
6716 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
6717 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
6718 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
6719 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
6720 unspec_fmfcfg): New patterns.
6721 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
6722 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
6723 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
6724 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
6725 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
6726 __nds32__fmfcfg): Define.
6728 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6730 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
6731 intrinsic register names.
6732 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
6733 intrinsic register enum values and macros.
6735 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6737 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
6738 for load/store addressing form.
6739 (nds32_print_operand_address): Likewise.
6741 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
6744 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
6745 based on LABEL_REF. Remove useless assertion.
6746 (pic_address_needs_scratch): Fix formatting.
6747 (sparc_legitimize_pic_address): Minor tweaks.
6748 (sparc_delegitimize_address): Adjust assertion accordingly.
6749 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
6750 into symbolic_operand.
6751 (movsi_high_pic_label_ref): Likewise.
6752 (movsi_lo_sum_pic_label_ref): Likewise.
6753 (movdi_pic_label_ref): Likewise.
6754 (movdi_high_pic_label_ref): Likewise.
6755 (movdi_lo_sum_pic_label_ref): Likewise.
6757 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
6759 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
6760 custom LIB_SPEC setup.
6762 2018-04-06 Ruslan Bukin <br@bsdpad.com>
6763 Kito Cheng <kito.cheng@gmail.com>
6765 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
6766 * config/riscv/freebsd.h: New.
6768 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6770 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
6771 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
6774 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6775 Kito Cheng <kito.cheng@gmail.com>
6777 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
6778 nds32_output_call, nds32_symbol_binds_local_p): New functions.
6779 * config/nds32/nds32-protos.h (nds32_output_call,
6780 nds32_output_return): Declare.
6781 * config/nds32/nds32.md: Refine all the call and return patterns.
6783 2018-04-06 Jakub Jelinek <jakub@redhat.com>
6786 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
6787 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
6789 PR rtl-optimization/84872
6790 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
6791 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
6794 2018-04-06 Tamar Christina <tamar.christina@arm.com>
6796 * expr.c (copy_blkmode_to_reg): Revert 254862.
6797 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
6799 2018-04-06 Richard Biener <rguenther@suse.de>
6802 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
6803 after seeing a component reference with an adjacent field. Treat
6804 refs to arrays at struct end of external decls similar to
6805 refs to unconstrained commons.
6807 2018-04-06 Jakub Jelinek <jakub@redhat.com>
6810 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
6811 look through SAVE_EXPRs with non-side-effects argument. Adjust
6813 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
6816 2018-04-06 Richard Biener <rguenther@suse.de>
6819 * alias.c (find_base_term): New wrapper around find_base_term
6820 unwinding CSELIB_VAL_PTR changes.
6821 (find_base_term): Do not restore CSELIB_VAL_PTR during the
6824 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6826 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
6828 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
6829 constant definitions.
6830 ("nop"): lr 0,0 -> nopr r0
6831 ("nop_lr0", "nop_lr1"): New insn definitions.
6833 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6835 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
6836 NDS32_V3PUSH_AVAILABLE_P macro.
6838 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
6839 Chung-Ju Wu <jasonwucj@gmail.com>
6841 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
6842 (nds32*-*-*): Add float and fpu_config into supported_defaults.
6843 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
6844 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
6845 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
6846 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
6847 * config/nds32/constraints.md: New constraints and checking for hard
6848 float configuration.
6849 * config/nds32/iterators.md: New mode iterator and attribute for hard
6850 float configuration.
6851 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
6853 * config/nds32/nds32-fpu.md: New file.
6854 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
6855 deal with hard float code generation.
6856 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
6858 (abi_type, float_reg_number): New enum type.
6859 * config/nds32/nds32-predicates.c: New predicates for hard float.
6860 * config/nds32/nds32-protos.h: Declare functions for hard float.
6861 * config/nds32/nds32.c: Implementation for hard float configuration.
6862 * config/nds32/nds32.h: Definitions for hard float configuration.
6863 * config/nds32/nds32.md: Include hard float machine description and
6864 modify patterns for hard float configuration.
6865 * config/nds32/nds32.opt: New options for hard float configuration.
6866 * config/nds32/predicates.md: New predicates for hard float
6869 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6871 * common/config/nds32/nds32-common.c
6872 (nds32_option_optimization_table): Enable -mreleax-hint by default.
6874 2018-04-05 Jakub Jelinek <jakub@redhat.com>
6877 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
6878 CONSTRUCTOR_ELT (ctor, ...)->value.
6880 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
6883 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
6885 2018-04-05 Tom de Vries <tom@codesourcery.com>
6888 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
6891 2018-04-05 Shiva Chen <shiva0217@gmail.com>
6892 Kito Cheng <kito.cheng@gmail.com>
6894 * config/nds32/constraints.md (U33): Fine-tune checking condition.
6895 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
6896 * config/nds32/nds32.h (nds32_16bit_address_type): Add
6897 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
6899 2018-04-05 Shiva Chen <shiva0217@gmail.com>
6900 Kito Cheng <kito.cheng@gmail.com>
6902 * config/nds32/constraints.md (Ufe): New memory constraint.
6903 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
6904 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
6905 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
6907 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
6908 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
6910 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6912 * config/nds32/nds32.md: Use optimize_size in the condition for
6913 alu-shift instructions.
6915 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6917 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
6919 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6921 * config/nds32/nds32.md (negsi2): Refine pattern.
6923 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
6924 Chung-Ju Wu <jasonwucj@gmail.com>
6926 * config/nds32/iterators.md (shift_rotate): New code iterator.
6927 (shift): New code attribute.
6928 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
6929 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
6930 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
6931 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
6932 bit-wise operations.
6933 (andsi3, *andsi3): Ditto.
6934 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
6935 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
6936 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
6937 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
6938 nds32_ior_operand, nds32_xor_operand): New predicates.
6940 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6942 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
6943 (addsi3, subsi3): ... this.
6945 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6947 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
6949 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6951 * config/nds32/nds32.md: Adjust indention.
6953 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
6955 * config/nds32/nds32.md (feature): New attribute.
6957 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
6959 * config/nds32/nds32.md (subtype): New attribute.
6961 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
6964 * config/arm/arm-builtins.c (arm_expand_builtin): Change
6965 expansion to perform a bitwise AND of the argument followed by a
6966 boolean negation of the result.
6968 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
6970 PR rtl-optimization/84878
6971 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
6972 the basic block. Assert the use reference is not artificial and that
6973 it has an associated insn.
6975 2018-04-04 Michael Matz <matz@suse.de>
6977 * builtins.c (compute_objsize): Pass correct operand
6978 to array_at_struct_end_p.
6980 2018-04-04 Richard Biener <rguenther@suse.de>
6983 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
6984 from contexts for DINFO_LEVEL_TERSE and below.
6986 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
6988 * config/nds32/nds32-doubleword.md (move_<mode>): Require
6989 resiter_operand condition.
6990 * config/nds32/nds32.md (*move<mode>): Ditto.
6992 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
6993 Monk Chiang <sh.chiang04@gmail.com>
6995 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
6997 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
6999 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
7001 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7002 Kito Cheng <kito.cheng@gmail.com>
7004 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
7005 nds32_cond_code_str, output_cond_branch,
7006 output_cond_branch_compare_zero, nds32_expand_cbranch,
7007 nds32_expand_cstore, nds32_expand_movcc,
7008 nds32_output_cbranchsi4_equality_zero,
7009 nds32_output_cbranchsi4_equality_reg,
7010 nds32_output_cbranchsi4_equality_reg_or_const_int,
7011 nds32_output_cbranchsi4_greater_less_zero: New functions.
7012 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
7013 nds32_expand_cstore, nds32_expand_movcc,
7014 nds32_output_cbranchsi4_equality_zero,
7015 nds32_output_cbranchsi4_equality_reg,
7016 nds32_output_cbranchsi4_equality_reg_or_const_int,
7017 nds32_output_cbranchsi4_greater_less_zero): Declare.
7018 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
7019 nds32_rimm11s_operand): New predicates.
7020 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
7021 * config/nds32/nds32.md: Rewrite all the branch and conditional move
7024 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7026 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
7027 * config/nds32/nds32.md: Ditto.
7028 * config/nds32/pipelines.md: Ditto.
7030 2018-04-04 Richard Biener <rguenther@suse.de>
7032 PR tree-optimization/85168
7033 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
7034 propagating abnormals.
7036 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7038 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
7040 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7041 Kito Cheng <kito.cheng@gmail.com>
7043 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
7044 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
7045 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
7046 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
7047 * config/nds32/nds32.md (sibcall_internal): New.
7048 (sibcall_register): Remove.
7049 (sibcall_immediate): Remove.
7050 (sibcall_value_internal): New.
7051 (sibcall_value_register): Remove.
7052 (sibcall_value_immediate): Remove.
7053 * config/nds32/predicates.md (nds32_general_register_operand): New.
7054 (nds32_call_address_operand): New.
7056 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7058 PR rtl-optimization/85167
7059 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
7060 bb_defs if *split_p, instead preinitialize it to NULL.
7062 PR tree-optimization/85156
7063 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
7064 evaluating the argument multiple times.
7066 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
7068 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
7070 (_mm_cvtpd_ps): Likewise.
7071 (_mm_cvttpd_epi32): Likewise.
7072 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
7073 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
7074 vector, pixel, and bool following altivec.h include.
7076 2018-04-03 Martin Sebor <msebor@redhat.com>
7078 * doc/extend.texi (Common Function Attributes): Clarify.
7079 (const attribute): Likewise.
7080 (pure attribute): Likewise.
7082 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7085 * config/i386/i386.c (ix86_expand_vector_set): Use
7086 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
7088 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
7090 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
7091 instructions when changing rounding bits to preserve precision bits
7092 in the x87 control word.
7094 2018-04-03 Martin Liska <mliska@suse.cz>
7096 PR tree-optimization/82491
7097 * rtl.h (strip_offset_and_add): Replace += suboffset with
7098 poly_uint64 () + suboffset.
7100 2018-03-29 Martin Liska <mliska@suse.cz>
7101 Martin Jambor <mjambor@suse.cz>
7104 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
7105 param_type is not an integral or pointer type.
7107 2018-04-03 Richard Biener <rguenther@suse.de>
7109 * sese.h (recompute_all_dominators): Remove.
7111 2018-04-02 Martin Sebor <msebor@redhat.com>
7113 * doc/invoke.texi (-Wrestrict): Fix typos.
7115 2018-04-02 Jim Wilson <jimw@sifive.com>
7117 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
7118 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
7119 (<optab>di3, <optab>si3_extend): Likewise.
7120 (<optab>si3_mask, <optab>si3_mask_1): New.
7121 (<optab>di3_mask, <optab>di3_mask_1): New.
7122 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
7123 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
7124 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
7126 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
7128 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
7131 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
7133 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
7134 (nds32_canonicalize_comparison): New function.
7136 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7137 Kito Cheng <kito.cheng@gmail.com>
7138 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7140 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
7141 * config/nds32/constants.md (unspec_volatile_element): Add
7142 UNSPEC_VOLATILE_RELAX_GROUP.
7143 * config/nds32/nds32-relax-opt.c: New file.
7144 * config/nds32/nds32-predicates.c
7145 (nds32_symbol_load_store_p): New function.
7146 * config/nds32/nds32-protos.h
7147 (nds32_symbol_load_store_p): Declare function.
7148 (make_pass_nds32_relax_opt): Declare new rtl pass function.
7149 * config/nds32/nds32.c
7150 (nds32_register_pass): New function to register pass.
7151 (nds32_register_passes): New function to register passes.
7152 * config/nds32/nds32.md (relax_group): New pattern.
7153 * config/nds32/nds32.opt (mrelax-hint): New option.
7154 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
7156 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
7158 * config/nds32/t-nds32: Modify files dependency.
7160 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7162 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
7163 (PROFILE_HOOK): Define its implementation.
7165 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7167 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
7168 type and 32-bit size.
7170 2018-04-01 Jakub Jelinek <jakub@redhat.com>
7173 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
7174 (V_128_256): New mode iterator.
7175 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
7176 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
7177 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
7179 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
7180 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
7182 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
7185 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
7186 NaN inputs correctly.
7188 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
7191 * config/rs6000/vsx.md (??r): New mode attribute.
7192 (*vsx_mov<mode>_64bit): Use it.
7193 (*vsx_mov<mode>_32bit): Likewise.
7195 2018-03-30 Martin Sebor <msebor@redhat.com>
7197 PR tree-optimization/84818
7198 * builtins.c (check_access): Use warning_n.
7200 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7203 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
7205 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
7208 2018-03-30 Julia Koval <julia.koval@intel.com>
7211 * x86-tune.def (movx, partial_reg_dependency): Enable for
7214 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
7217 * lra-constraints.c (process_alt_operands): Move setting
7218 this_alternative_matches below.
7220 2018-03-29 Martin Liska <mliska@suse.cz>
7223 * doc/invoke.texi: Document how LTO works with debug info.
7224 Describe auto-load support of binutils. Mention 'x86-64'
7225 as valid option value of -march option.
7227 2018-03-29 Jakub Jelinek <jakub@redhat.com>
7229 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
7232 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
7233 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
7234 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
7237 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
7240 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
7241 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
7242 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
7243 for RS6000_BTM_POWERPC64.
7244 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
7245 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
7246 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
7251 2018-03-28 Carl Love <cel@us.ibm.com>
7254 2017-09-27 Carl Love <cel@us.ibm.com>
7256 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
7257 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
7258 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
7261 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7263 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
7264 instead of __vector bool.
7265 (_mm_max_pu8): Likewise.
7266 (_mm_min_pi16): Likewise.
7268 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
7271 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
7272 (DIVWEUO): Likewise.
7274 (DIVDEUO): Likewise.
7275 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
7276 DIVWEUO and DIVDEUO.
7277 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
7278 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
7279 (div_extend): Likewise.
7280 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
7282 (__builtin_divweuo): Likewise.
7283 (__builtin_divdeo): Likewise.
7284 (__builtin_divdeuo): Likewise.
7286 2018-03-28 Jakub Jelinek <jakub@redhat.com>
7289 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
7290 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
7292 PR tree-optimization/82004
7293 * gimple-match-head.c (optimize_pow_to_exp): New function.
7294 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
7295 Don't fold to exp if optimize_pow_to_exp is false.
7297 2018-03-28 Martin Liska <mliska@suse.cz>
7300 * calls.c (initialize_argument_information): Fix trailing space.
7301 * common.opt: Fix typo and provide better explanation for
7302 -fsanitize-coverage option.
7303 * config/i386/i386.opt: Fix typo.
7305 2018-03-28 Jakub Jelinek <jakub@redhat.com>
7306 Martin Liska <mliska@suse.cz>
7309 * gimplify.c (asan_poison_variable): Don't do the check for
7310 gimplify_omp_ctxp here.
7311 (gimplify_decl_expr): Do it here.
7312 (gimplify_target_expr): Likewise.
7314 2018-03-28 Martin Liska <mliska@suse.cz>
7317 * config/i386/i386.c (ix86_function_arg_advance): Do not call
7318 chkp_type_bounds_count if MPX is not enabled.
7320 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
7322 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
7324 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
7327 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
7328 function to create the function decl for complex long double
7329 multiply and divide for -mabi=ieeelongdouble.
7330 (init_float128_ieee): Call it.
7332 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
7335 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
7336 -fcf-protection=branch -mibt.
7337 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
7339 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7342 * config/arm/arm.c (arm_valid_symbolic_address): Handle
7343 arm_word_relocations.
7345 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
7348 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
7349 extern array declarations.
7351 2018-03-27 Richard Biener <rguenther@suse.de>
7354 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
7355 explicit single_use checks.
7357 2018-03-27 Richard Biener <rguenther@suse.de>
7359 PR tree-optimization/85082
7360 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
7363 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7365 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
7366 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
7367 Turn on fasynchronous-unwind-tables and funwind-tables.
7369 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
7372 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
7373 (*bmi_blsr_<mode>_ccz): Ditto.
7375 2018-03-26 Tom de Vries <tom@codesourcery.com>
7377 PR tree-optimization/85063
7378 * omp-general.c (offloading_function_p): New function. Factor out
7380 * omp-offload.c (pass_omp_target_link::gate): ... here.
7381 * omp-general.h (offloading_function_p): Declare.
7382 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
7383 with attribute omp declare target for offloading functions.
7385 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
7387 PR tree-optimization/84005
7388 * tree-data-ref.h (get_base_for_alignment): Declare.
7389 * tree-data-ref.c (get_base_for_alignment_1): New function.
7390 (get_base_for_alignment): Likewise.
7391 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
7392 get_base_for_alignment to find a suitable base object, instead
7393 of always using drb->base_address.
7395 2018-03-23 Jakub Jelinek <jakub@redhat.com>
7398 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
7399 known size by default.
7401 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
7404 * lra-constraints.c (process_alt_operands): Don't match BLKmode
7405 and non BLKmode operands.
7407 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7410 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
7411 Clean up attributes.
7413 2018-03-23 Richard Biener <rguenther@suse.de>
7416 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
7417 we are going to emit early debug for LTO.
7419 2018-03-23 Jakub Jelinek <jakub@redhat.com>
7422 * function.c (match_asm_constraints_1): Don't optimize if input
7423 doesn't satisfy general_operand predicate for output's mode.
7426 * alias.c (write_dependence_p): Don't require for x_canonicalized
7427 non-VOIDmode if x has VOIDmode.
7430 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
7431 just don't try to optimize it rather than assert it never happens.
7433 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
7435 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
7436 macro expansions for definition of ST_INTERNAL_<mode> and
7437 LD_INTERNAL_<mode> builtins.
7438 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
7440 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
7442 (altivec_expand_st_builtin): Likewise.
7443 (altivec_expand_builtin): Remove calls to deleted functions.
7444 (rs6000_address_for_altivec): Delete this function.
7445 * config/rs6000/vector.md: Remove expands for
7446 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
7448 2018-03-22 Sudakshina Das <sudi.das@arm.com>
7451 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
7452 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
7453 re-computing once computed.
7454 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
7455 (arm_init_machine_status): Initialize
7456 machine->static_chain_stack_bytes.
7458 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
7461 * doc/extend.texi: Add four new prototypes for vec_ld.
7462 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
7463 definitions for more logical presentation.
7464 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
7465 entries for V1TI variants of __builtin_altivec_ld builtin.
7466 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
7467 handling of V1TI variant of LVX icode pattern.
7468 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
7469 (rs6000_gimple_fold_builtin): Likewise.
7470 (altivec_init_builtins): Add code to define
7471 __builtin_altivec_lvx_v1ti function.
7473 2018-03-22 Jakub Jelinek <jakub@redhat.com>
7476 * function.c (match_asm_constraints_1): Don't do the optimization
7477 if input isn't a REG, SUBREG, MEM or constant.
7479 2018-03-22 Tom de Vries <tom@codesourcery.com>
7481 PR tree-optimization/84956
7482 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
7483 bb_has_abnormal_pred.
7485 2018-03-22 Jakub Jelinek <jakub@redhat.com>
7488 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
7489 DECL_INITIAL (decl) to decl at the end.
7490 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
7493 2018-03-21 Joseph Myers <joseph@codesourcery.com>
7495 * doc/extend.texi (__builtin_tgmath): Document when complex
7496 integer types are treated as _Complex _Float64.
7498 2018-03-21 Tom de Vries <tom@codesourcery.com>
7500 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
7502 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7504 PR tree-optimization/84960
7505 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
7506 if it is ENTRY block, move them into single succ of ENTRY in that case.
7508 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
7510 PR tree-optimization/84811
7511 * poly-int.h (poly_span_traits): Remove the T3 parameter and
7512 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
7513 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
7514 (known_subrange_p): Update accordingly. Cast each value involved
7515 in the size comparison, rather than casting the result of the
7518 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7520 PR tree-optimization/84982
7521 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
7522 by flipping the least significant bit rather than all bits from
7523 bitpos to bitpos + bitsize - 1.
7525 2018-03-21 Nathan Sidwell <nathan@acm.org>
7527 * doc/extend.texi (Deprecated Features): Remove mention of
7528 long-deleted deprecations.
7530 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7533 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
7534 * configure: Regenerate.
7536 2018-03-21 Tom de Vries <tom@codesourcery.com>
7538 PR tree-optimization/83126
7539 * tree-parloops.c (num_phis): New function.
7540 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
7542 2018-03-21 Nathan Sidwell <nathan@acm.org>
7544 * doc/extend.texi (Deprecated Features): Update deprecated flags,
7545 mention anon-struct/union members and trailing attributes.
7547 2018-03-21 Bin Cheng <bin.cheng@arm.com>
7549 PR tree-optimization/84969
7550 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
7551 builtin memset partitions if they set different rhs values.
7553 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7555 PR rtl-optimization/84989
7556 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
7557 VEC_DUPLICATE with scalar result mode.
7559 2018-03-21 Martin Liska <mliska@suse.cz>
7562 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
7563 not intended return statement.
7565 2018-03-21 Martin Liska <mliska@suse.cz>
7568 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
7569 (chkp_find_bound_slots_1): Limit number of iterations.
7571 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
7574 * Minor grammar fixes for x86 options.
7576 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7579 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
7580 holding REG_CFA_RESTORE notes, instead turn them into a USE.
7582 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
7585 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
7586 (altivec_lvx_<mode>_1op): Likewise.
7587 (altivec_stvx_<mode>_2op): Likewise.
7588 (altivec_stvx_<mode>_1op): Likewise.
7589 (altivec_lvx_<VM2:mode>): New define_expand.
7590 (altivec_stvx_<VM2:mode>): Likewise.
7591 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
7592 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
7593 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
7594 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
7595 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
7596 (rs6000_gen_lvx): Likewise.
7597 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
7598 (altivec_expand_stv_builtin): Likewise.
7599 (altivec_expand_builtin): Likewise.
7600 * config/rs6000/vector.md: Likewise.
7602 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7605 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
7608 2018-03-20 Richard Biener <rguenther@suse.de>
7611 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
7612 sign-conversions as zero, fall back to standard scalar_stmt
7615 2018-03-20 Martin Liska <mliska@suse.cz>
7618 * predict.c (rebuild_frequencies): Handle case when we have
7619 PROFILE_ABSENT, but flag_guess_branch_prob is false.
7621 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7624 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
7625 flag_section_anchors.
7626 * varasm.c (use_blocks_for_decl_p): Remove hack for
7627 dw2_force_const_mem.
7630 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
7632 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
7633 be created, use lowpart_subreg of operands[0] rather than operands[0]
7635 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
7636 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
7637 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
7638 and n constraint instead of aarch64_shift_imm_di and Usd.
7639 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
7640 (*aarch64_<optab>_reg_minus<mode>3): ... this.
7642 2018-03-20 Sudakshina Das <sudi.das@arm.com>
7645 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
7646 to favor GPR over NEON registers.
7647 (<shift>di3_neon): Likewise.
7649 2018-03-20 Tom de Vries <tom@codesourcery.com>
7652 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
7653 (nvptx_process_pars): Emit bar.sync asap and alap.
7655 2018-03-20 Tom de Vries <tom@codesourcery.com>
7658 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
7659 seen_label if seen_label is already set.
7661 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7664 * config/i386/i386.c (fold_builtin_cpu): For features above 31
7665 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
7666 Use 1U instead of 1. Formatting fixes.
7669 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
7670 instead of TREE_TYPE (s1) for the return value.
7672 2018-03-19 Jakub Jelinek <jakub@redhat.com>
7674 PR tree-optimization/84946
7675 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
7676 bitsize + bitsize in poly_uint64 rather than poly_int64.
7679 * dwarf2asm.c: Include fold-const.c.
7680 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
7681 of decl rather than decl itself.
7683 PR rtl-optimization/84643
7684 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
7686 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
7689 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
7690 calling assemble_variable.
7692 2018-03-19 Sudakshina Das <sudi.das@arm.com>
7695 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
7696 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
7698 2018-03-19 Jim Wilson <jimw@sifive.com>
7701 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
7702 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
7703 (riscv_first_stack_step): Likewise.
7704 (riscv_option_override): Use STACK_BOUNDARY instead of
7706 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
7708 (BIGGEST_ALIGNMENT): Set to 128.
7709 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
7710 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
7713 2018-03-19 Richard Biener <rguenther@suse.de>
7715 PR tree-optimization/84933
7716 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
7717 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
7719 2018-03-19 Richard Biener <rguenther@suse.de>
7721 PR tree-optimization/84859
7722 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
7723 (cond_if_else_store_replacement): Perform sinking operation on
7724 single-store BBs regardless of MAX_STORES_TO_SINK setting.
7725 Generalize what a BB with a single eligible store is.
7727 2018-03-19 Richard Biener <rguenther@suse.de>
7729 PR tree-optimization/84929
7730 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
7731 chrec_is_positive against non-chrec arg.
7733 2018-03-19 Tamar Christina <tamar.christina@arm.com>
7736 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
7738 2018-03-18 Martin Liska <mliska@suse.cz>
7740 PR rtl-optimization/84635
7741 * regrename.c (build_def_use): Use matches_mode only when
7744 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
7746 PR tree-optimization/84913
7747 * tree-vect-loop.c (vectorizable_reduction): Don't try to
7748 vectorize chains of COND_EXPRs.
7750 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
7752 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
7754 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
7756 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
7758 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
7760 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
7762 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
7763 Kito Cheng <kito.cheng@gmail.com>
7765 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
7766 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
7767 (nds32_adjust_reg_alloc_order): New function.
7768 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
7770 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
7772 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
7773 nds32_print_operand, nds32_print_operand_address): Use
7774 HOST_WIDE_INT_PRINT_DEC instead.
7776 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
7778 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
7780 2018-03-17 Jakub Jelinek <jakub@redhat.com>
7783 * config/i386/i386.c (initial_ix86_tune_features,
7784 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
7786 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
7787 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
7788 rather than 1u << ix86_tune. Formatting fix.
7789 (ix86_option_override_internal): Change ix86_arch_mask from
7790 unsigned int to unsigned HOST_WIDE_INT, initialize to
7791 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
7792 (ix86_function_specific_restore): Likewise.
7794 2018-03-16 Jakub Jelinek <jakub@redhat.com>
7797 * postreload.c (reload_combine_recognize_pattern): Perform
7798 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
7799 truncate_int_for_mode the result for the destination's mode.
7802 * hsa-gen.c (mem_type_for_type): Fix comment typo.
7803 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
7805 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
7808 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
7811 * lra-assigns.c (lra_split_hard_reg_for): Don't use
7812 regno_allocno_class_array and sorted_pseudos.
7813 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
7814 insns where regno is used.
7816 2018-03-16 Martin Liska <mliska@suse.cz>
7819 * multiple_target.c (create_dispatcher_calls): Redirect
7820 reference in the symbol table.
7822 2018-03-16 Martin Liska <mliska@suse.cz>
7825 * multiple_target.c (create_dispatcher_calls): Redirect also
7828 2018-03-16 Jakub Jelinek <jakub@redhat.com>
7832 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
7833 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
7834 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
7836 2018-03-16 Julia Koval <julia.koval@intel.com>
7838 * doc/invoke.texi (Skylake Server): Add CLWB.
7839 Cannonlake): Remove CLWB.
7841 2018-03-16 Jakub Jelinek <jakub@redhat.com>
7843 PR tree-optimization/84841
7844 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
7846 (FLOAT_ONE_CONST_TYPE): Define.
7847 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
7848 (sort_by_operand_rank): Put entries with higher constant_type last
7849 rather than first to match comments.
7851 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
7853 * config/nios2/nios2.md (movsi_internal): Fix thinko in
7856 2018-03-15 Jakub Jelinek <jakub@redhat.com>
7859 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
7860 check and use address of target always.
7862 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
7865 * config/i386/i386.c (indirect_thunk_needed): Update comments.
7866 (indirect_thunk_bnd_needed): Likewise.
7867 (indirect_thunks_used): Likewise.
7868 (indirect_thunks_bnd_used): Likewise.
7869 (indirect_return_needed): New.
7870 (indirect_return_bnd_needed): Likewise.
7871 (output_indirect_thunk_function): Add a bool argument for
7873 (output_indirect_thunk_function): Don't generate alias for
7874 function return thunk.
7875 (ix86_code_end): Call output_indirect_thunk_function to generate
7876 function return thunks.
7877 (ix86_output_function_return): Set indirect_return_bnd_needed
7878 and indirect_return_needed instead of indirect_thunk_bnd_needed
7879 and indirect_thunk_needed.
7881 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
7883 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
7884 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
7885 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
7887 2018-03-15 David Malcolm <dmalcolm@redhat.com>
7888 Paul Hua <paul.hua.gm@gmail.com>
7891 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
7893 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
7895 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
7896 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
7899 2018-03-15 Tamar Christina <tamar.christina@arm.com>
7902 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
7903 instead of GET_MODE_SIZE when comparing Units.
7905 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
7908 * varasm.c (hash_section): Return an unchangeble hash value
7909 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
7910 Return !aarch64_can_use_per_function_literal_pools_p ().
7912 2018-03-15 Jakub Jelinek <jakub@redhat.com>
7915 * optabs.c (emit_conditional_move): Pass address of cmode's copy
7916 rather than address of cmode as last argument to prepare_cmp_insn.
7918 2018-03-15 Julia Koval <julia.koval@intel.com>
7920 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
7921 F_AVX512VNNI, F_AVX512BITALG): New.
7923 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
7926 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
7927 insn for floating-point loads and stores.
7929 2018-03-14 Carl Love <cel@us.ibm.com>
7931 * config/rs6000/rs6000-c.c: Add macro definitions for
7932 ALTIVEC_BUILTIN_VEC_PERMXOR.
7933 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
7934 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
7935 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
7936 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
7938 * config/doc/extend.texi: Add prototypes for vec_permxor.
7940 2018-03-14 David Malcolm <dmalcolm@redhat.com>
7943 * diagnostic-show-locus.c (class layout_point): Convert m_line
7944 from int to linenum_type.
7945 (line_span::comparator): Use linenum "compare" function when
7946 comparing line numbers.
7947 (test_line_span): New function.
7948 (layout_range::contains_point): Convert param "row" from int to
7950 (layout_range::intersects_line_p): Likewise.
7951 (layout::will_show_line_p): Likewise.
7952 (layout::print_source_line): Likewise.
7953 (layout::should_print_annotation_line_p): Likewise.
7954 (layout::print_annotation_line): Likewise.
7955 (layout::print_leading_fixits): Likewise.
7956 (layout::annotation_line_showed_range_p): Likewise.
7957 (struct line_corrections): Likewise for field m_row.
7958 (line_corrections::line_corrections): Likewise for param "row".
7959 (layout::print_trailing_fixits): Likewise.
7960 (layout::get_state_at_point): Likewise.
7961 (layout::get_x_bound_for_row): Likewise.
7962 (layout::print_line): Likewise.
7963 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
7964 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
7965 * input.c (selftest::test_linenum_comparisons): New function.
7966 (selftest::input_c_tests): Call it.
7967 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
7968 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
7969 * selftest.h (ASSERT_GT): New macro.
7970 (ASSERT_GT_AT): New macro.
7971 (ASSERT_LT): New macro.
7972 (ASSERT_LT_AT): New macro.
7974 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
7976 PR rtl-optimization/84780
7977 * combine.c (distribute_links): Don't make a link based on pc_rtx.
7979 2018-03-14 Martin Liska <mliska@suse.cz>
7981 * tree.c (record_node_allocation_statistics): Use
7982 get_stats_node_kind.
7983 (get_stats_node_kind): New function extracted from
7984 record_node_allocation_statistics.
7985 (free_node): Use get_stats_node_kind.
7987 2018-03-14 Richard Biener <rguenther@suse.de>
7989 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
7990 that the value-set of ANTIC_IN doesn't grow.
7993 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
7995 (BB_VISITED_WITH_VISITED_SUCCS): New define.
7996 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
7998 2018-03-14 Julia Koval <julia.koval@intel.com>
8000 * config.gcc (icelake-client, icelake-server): New.
8002 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
8003 (initial_ix86_arch_features): Ditto.
8004 (PTA_SKYLAKE): Add SGX.
8005 (PTA_ICELAKE): Remove.
8006 (PTA_ICELAKE_CLIENT): New.
8007 (PTA_ICELAKE_SERVER): New.
8008 (ix86_option_override_internal): Split up icelake on icelake client and
8010 (get_builtin_code_for_version): Ditto.
8011 (fold_builtin_cpu): Ditto.
8012 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
8013 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
8014 * config/i386/i386.h (processor_type): Ditto.
8015 * doc/invoke.texi: Ditto.
8017 2018-03-14 Jakub Jelinek <jakub@redhat.com>
8020 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
8021 INTEGER_CST offset, add it together with bitpos / 8 and
8022 sign extend based on POINTER_SIZE.
8026 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
8029 * config/i386/constraints.md (Yc): New register constraint.
8030 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
8031 Use Yc constraint for alternative 2 of operand 0. Remove
8032 preferred_for_speed attribute.
8034 2018-03-14 Richard Biener <rguenther@suse.de>
8036 PR tree-optimization/84830
8037 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
8038 with the old one to avoid oscillations.
8040 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
8043 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
8045 (assign_by_spills): Return a flag of reload assignment failure.
8046 Do not process the reload assignment failures. Do not spill other
8047 reload pseudos if they has the same reg class. Update n if
8049 (lra_assign): Add a return arg. Set up from the result of
8050 assign_by_spills call.
8051 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8052 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8053 usage_insns if it is not NULL.
8054 (spill_hard_reg_in_range): New function.
8055 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8056 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8057 function prototypes.
8058 (lra_assign): Change prototype.
8059 * lra.c (lra): Add code to deal with fails by splitting hard reg
8062 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
8064 * config/riscv/riscv.opt (mrelax): New option.
8065 * config/riscv/riscv.c (riscv_file_start): Emit ".option
8066 "norelax" when riscv_mrelax is disabled.
8067 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
8069 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8072 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
8073 reassociation for int modes.
8075 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8077 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
8078 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
8080 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
8081 * config/aarch64/aarch64-sve.md
8082 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
8083 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
8084 (*extend<mode><Vwide>2): Rename to...
8085 (aarch64_sve_extend<mode><Vwide>2): ...this.
8086 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
8087 renaming the old pattern to...
8088 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
8090 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
8091 define_expand, renaming the old pattern to...
8092 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
8093 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
8094 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
8095 account when deciding which SVE instruction the optab should use.
8096 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
8098 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8100 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
8101 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
8102 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
8103 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
8104 (tlsdesc_small_<mode>): Turn a define_expand and use
8105 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
8106 (tlsdesc_small_advsimd_<mode>): ...this.
8107 (tlsdesc_small_sve_<mode>): New pattern.
8109 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8111 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
8112 (UNSPEC_UMUL_HIGHPART): New constants.
8113 (MUL_HIGHPART): New int iteraor.
8114 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
8115 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
8117 (*<su>mul<mode>3_highpart): New define_insn.
8119 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
8122 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
8125 2018-03-13 Martin Liska <mliska@suse.cz>
8128 * (sem_item_optimizer::sem_item_optimizer): Initialize new
8130 (sem_item_optimizer::~sem_item_optimizer): Release it.
8131 (sem_item_optimizer::merge_classes): Register variable aliases.
8132 (sem_item_optimizer::fixup_pt_set): New function.
8133 (sem_item_optimizer::fixup_points_to_sets): Likewise.
8134 * ipa-icf.h: Declare new variables and functions.
8136 2018-03-13 Jakub Jelinek <jakub@redhat.com>
8139 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
8140 integer_pow2p@2 and test integer_pow2p in condition.
8141 (A < 0 ? C : 0): Similarly for @1.
8144 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
8145 characters starting at p contain '\0' character, don't look beyond
8149 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
8150 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
8153 * reg-stack.c (change_stack): Change update_end var from int to
8154 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
8155 also call set_block_for_insn on the newly added insns and rescan.
8158 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
8159 on the last operand.
8162 * tree.c (stabilize_reference_1): Return save_expr (e) for
8163 STATEMENT_LIST even if it doesn't have side-effects.
8165 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
8167 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
8169 2018-03-12 Renlin Li <renlin.li@arm.com>
8171 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
8172 aarch64_output_scalar_simd_mov_immediate.
8174 2018-03-12 Martin Sebor <msebor@redhat.com>
8176 PR tree-optimization/83456
8177 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
8178 for perfectly overlapping calls to memcpy.
8179 (gimple_fold_builtin_memory_chk): Same.
8180 (gimple_fold_builtin_strcpy): Handle no-warning.
8181 (gimple_fold_builtin_stxcpy_chk): Same.
8182 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
8184 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
8186 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
8187 parameter. Use it for SFmode.
8188 (rs6000_function_arg_advance_1): Adjust.
8189 (rs6000_function_arg): Adjust.
8190 (rs6000_gimplify_va_arg): Pass false for that new parameter.
8192 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
8194 PR rtl-optimization/84169
8195 PR rtl-optimization/84780
8196 * combine.c (can_combine_p): Check for a 2-insn combination whether
8197 the destination register is used between the two insns, too.
8199 2018-03-12 Richard Biener <rguenther@suse.de>
8201 PR tree-optimization/84803
8202 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
8203 for refs DR analysis didn't process.
8205 2018-03-12 Richard Biener <rguenther@suse.de>
8207 PR tree-optimization/84777
8208 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
8209 force-vectorize loops ignore whether we are optimizing for size.
8211 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
8213 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
8214 (TARGET_MD_ASM_ADJUST): Define.
8216 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
8217 Kito Cheng <kito.cheng@gmail.com>
8218 Chung-Ju Wu <jasonwucj@gmail.com>
8220 * config/nds32/nds32.c (nds32_compute_stack_frame,
8221 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
8222 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
8223 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
8224 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
8225 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
8226 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
8227 * config/nds32/nds32.md (prologue, epilogue): Use macro
8228 NDS32_V3PUSH_AVAILABLE_P to do checking.
8230 2018-03-11 Jakub Jelinek <jakub@redhat.com>
8233 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
8234 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
8235 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
8236 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
8237 addition of most attributes on !orig_type_die or the attribute not
8238 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
8240 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8241 Chung-Ju Wu <jasonwucj@gmail.com>
8243 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
8245 * config/nds32/nds32.opt (mvh): New option.
8247 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8248 Chung-Ju Wu <jasonwucj@gmail.com>
8250 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
8252 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
8253 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
8256 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8257 Chung-Ju Wu <jasonwucj@gmail.com>
8259 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
8261 * config/nds32/nds32-multiple.md (strlensi): New pattern.
8262 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
8264 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
8265 Kito Cheng <kito.cheng@gmail.com>
8266 Chung-Ju Wu <jasonwucj@gmail.com>
8268 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
8269 UNSPEC_FFMISM and UNSPEC_FLMISM.
8270 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
8271 for ffb, ffmism and flmism.
8272 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
8273 (unspec_ffmism): Ditto.
8274 (unspec_flmism): Ditto.
8275 (nds32_expand_builtin_impl): Check if string extension is available.
8276 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
8277 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
8279 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
8282 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
8285 * lra-assigns.c (assign_by_spills): Return a flag of reload
8286 assignment failure. Do not process the reload assignment
8287 failures. Do not spill other reload pseudos if they has the same
8289 (lra_assign): Add a return arg. Set up from the result of
8290 assign_by_spills call.
8291 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8292 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8293 usage_insns if it is not NULL.
8294 (spill_hard_reg_in_range): New function.
8295 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8296 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8297 function prototypes.
8298 (lra_assign): Change prototype.
8299 * lra.c (lra): Add code to deal with fails by splitting hard reg
8302 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
8305 * config/i386/i386.opt: Replace Enforcment with Enforcement.
8307 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
8310 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
8311 (dw_val_node): Add val_symbolic_view.
8312 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
8313 (symview_upper_bound): New.
8314 (new_line_info_table): Initialize symviews_since_reset.
8315 (dwarf2out_source_line): Count symviews_since_reset and set
8316 symview_upper_bound.
8317 (dw_val_equal_p): Handle symview.
8318 (add_AT_symview): New.
8319 (print_dw_val): Handle symview.
8320 (attr_checksum, attr_checksum_ordered): Likewise.
8321 (same_dw_val_p, size_of_die): Likewise.
8322 (value_format, output_die): Likewise.
8323 (add_high_low_attributes): Use add_AT_symview for entry_view.
8324 (dwarf2out_finish): Reset symview_upper_bound, clear
8327 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
8330 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
8331 Add strict argument and use it.
8332 (rs6000_split_multireg_move): Update for new strict argument.
8333 (mem_operand_gpr): Disallow all non-offsettable addresses.
8334 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
8336 2018-03-09 Jakub Jelinek <jakub@redhat.com>
8339 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
8340 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
8341 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
8344 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
8345 decl, use remap_type if we want to use the type.
8347 2018-03-09 Martin Sebor <msebor@redhat.com>
8349 PR tree-optimization/84526
8350 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
8352 (builtin_access::generic_overlap): Be prepared to handle non-array
8355 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
8357 PR rtl-optimization/84682
8358 * lra-constraints.c (process_address_1): Check is_address flag
8359 for address constraints.
8360 (process_alt_operands): Likewise.
8361 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
8362 preprocess_constraints.
8363 * recog.h (preprocess_constraints): Add oploc parameter.
8365 * recog.c (preprocess_constraints): Test address_operand for
8366 CT_ADDRESS constraints.
8368 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
8371 * lra-assigns.c (assign_by_spills): Return a flag of reload
8372 assignment failure. Do not process the reload assignment
8373 failures. Do not spill other reload pseudos if they has the same
8375 (lra_assign): Add a return arg. Set up from the result of
8376 assign_by_spills call.
8377 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8378 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8379 usage_insns if it is not NULL.
8380 (spill_hard_reg_in_range): New function.
8381 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8382 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8383 function prototypes.
8384 (lra_assign): Change prototype.
8385 * lra.c (lra): Add code to deal with fails by splitting hard reg
8388 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8391 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
8392 Accept complain bool parameter. Only emit errors if it is true.
8393 (arm_parse_cpu_option_name): Likewise.
8394 (arm_target_thumb_only): Adjust callers of the above.
8395 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
8396 prototype to take a default true bool parameter.
8397 (arm_parse_arch_option_name): Likewise.
8399 2018-03-09 David Malcolm <dmalcolm@redhat.com>
8400 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
8404 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
8405 * configure: Regenerate.
8406 * configure.ac ("linker --version-script option"): New.
8407 ("linker soname option"): New.
8409 2018-03-09 Richard Biener <rguenther@suse.de>
8411 PR tree-optimization/84775
8412 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
8413 immediate uses of predicate stmts and mark them modified.
8416 PR tree-optimization/84178
8417 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
8419 (version_loop_for_if_conversion): Delay update_ssa call.
8420 (tree_if_conversion): Delay update_ssa until after predicate
8423 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
8426 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
8427 when the function accesses prior frames.
8429 2018-03-08 Jakub Jelinek <jakub@redhat.com>
8432 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
8433 gen_llsym, otherwise call maybe_gen_llsym.
8436 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
8437 has ',' character inside of it.
8439 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8442 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
8443 as clobbering CC_REGNUM.
8445 2018-03-08 Richard Biener <rguenther@suse.de>
8448 * tree-scalar-evolution.c: Include tree-into-ssa.h.
8449 (follow_copies_to_constant): Do not follow SSA names registered
8452 2018-03-08 Richard Biener <rguenther@suse.de>
8454 PR tree-optimization/84178
8455 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
8457 (version_loop_for_if_conversion): Delay update_ssa call.
8458 (tree_if_conversion): Delay update_ssa until after predicate
8461 2018-03-08 David Malcolm <dmalcolm@redhat.com>
8463 PR tree-optimization/84178
8464 * tree-if-conv.c (release_bb_predicate): Remove the
8465 the assertion that the stmts have NULL use_ops.
8466 Discard the statements, asserting that they haven't
8467 yet been added to a BB.
8469 2018-03-08 Richard Biener <rguenther@suse.de>
8471 PR tree-optimization/84746
8472 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
8473 (phi_translate): Pass in destination ANTIC_OUT set.
8474 (phi_translate_1): Likewise. For a simplified result lookup
8475 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
8476 (phi_translate_set): Adjust.
8477 (do_pre_regular_insertion): Likewise.
8478 (do_pre_partial_partial_insertion): Likewise.
8480 2018-03-08 Martin Liska <mliska@suse.cz>
8482 PR gcov-profile/84735
8483 * doc/gcov.texi: Document usage of profile files.
8484 * gcov-io.h: Document changes in the format.
8486 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
8490 * dwarf2out.c (struct dw_line_info_table): Update comments for
8492 (FORCE_RESET_NEXT_VIEW): New.
8493 (FORCE_RESETTING_VIEW_P): New.
8494 (RESETTING_VIEW_P): Check for -1 too.
8495 (ZERO_VIEW_P): Likewise.
8496 (new_line_info_table): Force-reset next view.
8497 (dwarf2out_begin_function): Likewise.
8498 (dwarf2out_source_line): Simplify zero_view_p initialization.
8499 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
8500 view directly. Omit view when omitting .loc at line 0.
8502 2018-03-08 Jakub Jelinek <jakub@redhat.com>
8504 PR tree-optimization/84740
8505 * tree-switch-conversion.c (process_switch): Call build_constructors
8506 only if info.phi_count is non-zero.
8508 PR tree-optimization/84739
8509 * tree-tailcall.c (find_tail_calls): Check call arguments against
8510 DECL_ARGUMENTS (current_function_decl) rather than
8511 DECL_ARGUMENTS (func) when checking for tail recursion.
8513 2018-03-07 Jakub Jelinek <jakub@redhat.com>
8515 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
8516 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
8517 Volker Reichelt's entry and add entries for people that perform
8518 GCC fuzzy testing and report numerous bugs.
8520 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
8523 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
8524 readonly data in sdata, if that is disabled.
8525 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
8526 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
8527 -mreadonly-in-sdata option.
8529 2018-03-07 Martin Sebor <msebor@redhat.com>
8531 PR tree-optimization/84468
8532 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
8533 basic block when looking for nul assignment.
8535 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
8538 * except.h (output_function_exception_table): Adjust prototype.
8539 * except.c (output_function_exception_table): Remove FNNAME parameter
8540 and add SECTION parameter. Ouput one part of the table at a time.
8541 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
8542 the first part of the exception table and emit unwind directives.
8543 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
8544 (i386_pe_seh_cold_init): Likewise.
8545 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
8546 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
8547 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
8548 (ix86_output_call_insn): Emit a nop in one more case for SEH.
8549 * config/i386/winnt.c: Include except.h.
8550 (struct seh_frame_state): Add reg_offset, after_prologue and
8551 in_cold_section fields.
8552 (i386_pe_seh_end_prologue): Set seh->after_prologue.
8553 (i386_pe_seh_cold_init): New function.
8554 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
8555 to seh->in_cold_section.
8556 (seh_emit_push): Record the offset of the push.
8557 (seh_emit_save): Record the offet of the save.
8558 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
8559 Test seh->after_prologue to disregard the epilogue.
8560 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
8561 (i386_pe_end_cold_function): New function.
8563 2018-03-07 Jakub Jelinek <jakub@redhat.com>
8566 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
8567 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
8570 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
8572 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
8573 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
8576 * multiple_target.c: Include tree-inline.h and intl.h.
8577 (expand_target_clones): Diagnose and fail if node->definition and
8578 !tree_versionable_function_p (node->decl).
8580 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
8582 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
8584 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
8585 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
8586 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
8588 2018-03-06 Jakub Jelinek <jakub@redhat.com>
8591 * combine.c (try_combine): Use reg_or_subregno instead of handling
8592 just paradoxical SUBREGs and REGs.
8594 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
8596 * config/arc/arc.c (arc_finalize_pic): Remove function.
8597 (arc_must_save_register): We use single base PIC register, remove
8598 checks to save/restore the PIC register.
8599 (arc_expand_prologue): Likewise.
8600 * config/arc/arc-protos.h (arc_set_default_type_attributes):
8602 (arc_verify_short): Likewise.
8603 (arc_attr_type): Likewise.
8604 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
8605 (walk_stores): Likewise.
8606 (arc_address_cost): Make it static.
8607 (arc_verify_short): Likewise.
8608 (branch_dest): Likewise.
8609 (arc_attr_type): Likewise.
8610 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
8611 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
8612 (arc_final_prescan_insn): Remove inserting the nops due to
8613 hardware hazards. It is done in reorg step.
8614 (insn_length_variant_t): Remove.
8615 (insn_length_parameters_t): Likewise.
8616 (arc_insn_length_parameters): Likewise.
8617 (arc_get_insn_variants): Likewise.
8618 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
8620 2018-03-06 Jakub Jelinek <jakub@redhat.com>
8623 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
8626 PR tree-optimization/84687
8627 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
8629 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
8631 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8633 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
8634 Rename to ppc_speculation_barrier.
8635 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
8636 __builtin_ppc_speculation_barrier.
8638 2018-03-05 Jakub Jelinek <jakub@redhat.com>
8641 * combine.c (combine_simplify_rtx): Don't try to simplify if
8642 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
8645 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
8647 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
8648 to 32 bytes when compiling for POWER9.
8650 2018-03-05 Jakub Jelinek <jakub@redhat.com>
8653 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
8654 regparm >= 3 with no arg reg available also for calls with
8655 flag_force_indirect_call. Pass decl to ix86_function_regparm.
8658 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
8660 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
8662 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
8665 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
8667 2018-03-05 Richard Biener <rguenther@suse.de>
8669 PR tree-optimization/84486
8670 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
8671 When inserting a __builtin_assume_aligned call set the LHS
8672 SSA name alignment info accordingly.
8674 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
8676 PR tree-optimization/84114
8677 * config/aarch64/aarch64.c (aarch64_reassociation_width)
8678 Avoid reassociation of FLOAT_MODE addition.
8680 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
8682 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
8683 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
8684 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
8685 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
8686 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
8687 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
8688 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
8690 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
8691 __builtin_ia32_wbinvd): New builtins.
8692 (SPECIAL_ARGS2): New.
8693 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
8694 (SPECIAL_ARGS2): New.
8695 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
8696 (ix86_valid_target_attribute_inner_p): Ditto.
8697 (ix86_init_mmx_sse_builtins): Add special_args2.
8698 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
8699 TARGET_WBNOINVD_P): New.
8700 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
8701 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
8702 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
8703 * config/i386/immintrin.h (_wbinvd): New intrinsic.
8704 * config/i386/pconfigintrin.h: New file.
8705 * config/i386/wbnoinvdintrin.h: Ditto.
8706 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
8708 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
8710 2018-03-05 Richard Biener <rguenther@suse.de>
8712 PR tree-optimization/84670
8713 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
8715 (BB_VISITED_WITH_VISITED_SUCCS): New define.
8716 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
8717 (compute_antic_aux): Only assert the number of values in ANTIC_IN
8718 doesn't grow if all successors (recursively) were visited at least
8721 2018-03-05 Richard Biener <rguenther@suse.de>
8723 PR tree-optimization/84650
8724 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
8725 if executed in the loop pipeline.
8727 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
8729 * doc/configfiles.texi (Configuration Files): Move info about
8730 conditionalizing $target-protos.h to...
8731 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
8732 differs from $target-protos.h.
8734 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
8735 Chung-Ju Wu <jasonwucj@gmail.com>
8737 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
8738 * config/nds32/nds32-multiple.md (setmemsi): Define.
8739 * config/nds32/nds32-memory-manipulation.c
8740 (nds32_gen_dup_4_byte_to_word_value): New.
8741 (emit_setmem_word_loop): New.
8742 (emit_setmem_byte_loop): New.
8743 (nds32_expand_setmem_loop): New.
8744 (nds32_expand_setmem_loop_v3m): New.
8745 (nds32_expand_setmem_unroll): New.
8746 (nds32_expand_setmem): New.
8748 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
8749 Chung-Ju Wu <jasonwucj@gmail.com>
8751 * config/nds32/nds32-memory-manipulation.c
8752 (nds32_emit_load_store): New.
8753 (nds32_emit_post_inc_load_store): New.
8754 (nds32_emit_mem_move): New.
8755 (nds32_emit_mem_move_block): New.
8756 (nds32_expand_movmemsi_loop_unknown_size): New.
8757 (nds32_expand_movmemsi_loop_known_size): New.
8758 (nds32_expand_movmemsi_loop): New.
8759 (nds32_expand_movmemsi_unroll): New.
8760 (nds32_expand_movmemqi): Rename ...
8761 (nds32_expand_movmemsi): ... to this.
8762 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
8763 (movmemsi): ... to this.
8764 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
8765 (nds32_expand_movmemsi): ... to this.
8767 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
8768 Monk Chiang <sh.chiang04@gmail.com>
8769 Chung-Ju Wu <jasonwucj@gmail.com>
8771 * config/nds32/nds32-protos.h
8772 (nds32_expand_load_multiple): New arguments.
8773 (nds32_expand_store_multiple): Ditto.
8774 (nds32_valid_multiple_load_store): Rename ...
8775 (nds32_valid_multiple_load_store_p): ... to this.
8776 * config/nds32/nds32-memory-manipulation.c
8777 (nds32_expand_load_multiple): Refine implementation.
8778 (nds32_expand_store_multiple): Ditto.
8779 * config/nds32/nds32-multiple.md
8780 (load_multiple): Update nds32_expand_load_multiple interface.
8781 (store_multiple): Update nds32_expand_store_multiple interface.
8782 * config/nds32/nds32-predicates.c
8783 (nds32_valid_multiple_load_store): Rename ...
8784 (nds32_valid_multiple_load_store_p): ... to this and refine
8786 * config/nds32/predicates.md
8787 (nds32_load_multiple_and_update_address_operation): New predicate.
8788 (nds32_store_multiple_and_update_address_operation): New predicate.
8790 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
8791 Chung-Ju Wu <jasonwucj@gmail.com>
8793 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
8794 (combo): New attribute.
8795 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
8797 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
8799 * config/nds32/nds32.opt: Change -mcmodel= default value.
8801 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
8802 Monk Chiang <sh.chiang04@gmail.com>
8803 Chung-Ju Wu <jasonwucj@gmail.com>
8805 * config/nds32/constants.md (unspec_element): New enum.
8806 * config/nds32/constraints.md (Umw): New constraint.
8807 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
8808 * config/nds32/nds32-intrinsic.md: Likewise.
8809 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
8810 (nds32_valid_smw_lwm_base_p): New.
8811 (nds32_output_smw_single_word): New.
8812 (nds32_output_lmw_single_word): New.
8813 (nds32_expand_unaligned_load): New.
8814 (nds32_expand_unaligned_store): New.
8815 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
8816 (nds32_output_smw_single_word): Declare.
8817 (nds32_output_lmw_single_word): Declare.
8818 (nds32_expand_unaligned_load): Declare.
8819 (nds32_expand_unaligned_store): Declare.
8820 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
8821 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
8822 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
8823 NDS32_BUILTIN_UASTORE_DW.
8824 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
8827 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
8828 Kito Cheng <kito.cheng@gmail.com>
8829 Chung-Ju Wu <jasonwucj@gmail.com>
8831 * config/nds32/nds32-intrinsic.c
8832 (nds32_expand_builtin_null_ftype_reg): Delete.
8833 (nds32_expand_builtin_reg_ftype_imm): Ditto.
8834 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
8835 (nds32_read_argument): New.
8836 (nds32_legitimize_target): Ditto.
8837 (nds32_legitimize_argument): Ditto.
8838 (nds32_check_constant_argument): Ditto.
8839 (nds32_expand_unop_builtin): Ditto.
8840 (nds32_expand_unopimm_builtin): Ditto.
8841 (nds32_expand_binop_builtin): Ditto.
8842 (nds32_builtin_decl_impl): Ditto.
8843 (builtin_description): Ditto.
8844 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
8845 (nds32_init_builtins_impl): Ditto.
8846 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
8847 (nds32_builtin_decl): New.
8848 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
8849 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
8851 2018-03-02 Jeff Law <law@redhat.com>
8853 * reorg.c (stop_search_p): Handle DEBUG_INSN.
8854 (redundant_insn, fill_simple_delay_slots): Likewise.
8855 (fill_slots_from_thread): Likewise.
8856 * resource.c (mark_referenced_resources): Likewise.
8857 (mark_set_resources, find_dead_or_set_registers): Likewise.
8859 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8861 * substring-locations.h (format_warning_va): Formatting fix for
8863 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
8865 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
8866 * substring-locations.c: Include intl.h.
8867 (format_warning_va): Turned into small wrapper around
8868 format_warning_n_va, renamed to ...
8869 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
8870 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
8872 (format_warning_at_substring_n): New function.
8873 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
8874 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
8875 format_warning_at_substring with just a shorter name instead of
8876 const function pointer.
8877 (fmtwarn_n): New function.
8878 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
8879 appropriate, get rid of all the fmtstr temporaries, move conditionals
8880 with G_() wrapped string literals directly into fmtwarn arguments,
8881 cast dir.len to (int), formatting fixes.
8883 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
8885 * doc/invoke.texi: Remove "Cilk Plus" references.
8887 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8888 Richard Biener <rguenther@suse.de>
8891 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
8892 for error or warning attributes if CALL_FROM_THUNK_P is set.
8895 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8898 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
8899 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
8902 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
8903 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
8905 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
8906 instead of -1U in last predictors element's probability member.
8908 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
8911 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
8912 arguments if they are comparable.
8914 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
8916 PR tree-optimization/84634
8917 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
8918 masks and masked_loop_p with a single loop_masks, making sure it's
8919 null for bb vectorization.
8921 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
8923 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
8924 (vect_analyze_data_ref_access): Use loop->safe_len rather than
8925 loop->force_vectorize to check whether there is no alias.
8927 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8930 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
8932 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
8934 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
8935 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
8936 instead of a loop around prev_real_insn.
8937 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
8941 * config/i386/i386.c (ix86_print_operand): Use conditional
8942 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
8945 2018-03-02 Richard Biener <rguenther@suse.de>
8947 PR tree-optimization/84427
8948 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
8949 (bitmap_set_subtract_values): Rewrite to handle multiple
8952 (prune_clobbered_mems): Likewise.
8953 (phi_translate): Take edge instead of pred/phiblock.
8954 (phi_translate_1): Likewise.
8955 (phi_translate_set): Likewise. Insert all translated
8956 exprs for a value into the set, keeping possibly multiple
8957 expressions per value.
8958 (compute_antic_aux): Adjust for phi_translate changes.
8959 When intersecting union the expressions and prune those
8960 not in the final value set, keeping possibly multiple
8961 expressions per value. Do not use value-insertion
8962 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
8963 all expressions. Add verification that the value-sets
8964 only shrink during iteration.
8965 (compute_partial_antic_aux): Adjust for the phi_translate changes.
8966 (do_pre_regular_insertion): Likewise.
8967 (do_pre_partial_partial_insertion): Likewise.
8969 2018-03-02 Richard Biener <rguenther@suse.de>
8972 * config/darwin.c (saved_debug_info_level): New static global.
8973 (darwin_asm_lto_start): Disable debug info generation for LTO out.
8974 (darwin_asm_lto_end): Restore debug info generation settings.
8976 2018-03-01 Martin Liska <mliska@suse.cz>
8979 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
8982 2018-03-01 Richard Biener <rguenther@suse.de>
8985 * dwarf2out.c (gen_variable_die): Properly handle late VLA
8986 type annotation with LTO when debug was disabled at compile-time.
8988 2018-03-01 Matthew Fortune <mfortune@gmail.com>
8990 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
8992 (mips_final_postscan_insn): Likewise.
8994 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
8996 PR rtl-optimization/84528
8997 * alias.c (init_alias_target): Add commentary.
8998 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
8999 a unique base value if the frame pointer is not eliminated
9000 to the stack pointer.
9002 2018-03-01 Tom de Vries <tom@codesourcery.com>
9004 PR rtl-optimization/83327
9005 * lra-int.h (hard_regs_spilled_into): Declare.
9006 * lra.c (hard_regs_spilled_into): Define.
9007 (init_reg_info): Init hard_regs_spilled_into.
9008 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
9009 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
9010 (process_bb_lives): Handle hard_regs_spilled_into.
9011 (lra_create_live_ranges_1): Before doing liveness propagation, clear
9012 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
9014 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
9016 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
9017 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
9018 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
9019 * config/rs6000/aix72.h: New file.
9021 2018-02-28 Jakub Jelinek <jakub@redhat.com>
9023 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
9024 instead of warning_at with conditional singular and plural messages
9028 * stor-layout.c (update_alignment_for_field): For
9029 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
9030 && !DECL_PACKED (field), do the alignment update, just use
9031 only desired_align instead of MAX (type_align, desired_align)
9033 (place_field): Don't do known_align < desired_align handling
9034 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
9035 is non-NULL, instead do it after rli->prev_field handling and
9036 only if not within a bitfield word. For DECL_PACKED (field)
9037 use type_align of BITS_PER_UNIT.
9039 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
9041 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
9042 superfluous parentheses and trailing spaces.
9044 2018-02-28 Richard Biener <rguenther@suse.de>
9046 PR tree-optimization/84584
9047 * graphite-scop-detection.c (scop_detection::add_scop): Discard
9048 SCoPs with fake exit edge.
9050 2018-02-28 Martin Liska <mliska@suse.cz>
9053 * timevar.c (timer::print): Fix format to properly print 100%
9056 2018-02-28 Richard Biener <rguenther@suse.de>
9059 * genmatch.c (capture_info::walk_match): Do not mark
9060 captured expressions without operands as expr_p given
9061 they act more like predicates and should be subject to
9062 "lost tail" side-effect preserving.
9064 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
9066 PR rtl-optimization/81611
9067 * auto-inc-dec.c (attempt_change): Move dead note from
9068 mem_insn if it's the next use of regno
9069 (find_address): Take address use of reg holding
9070 non-incremented value. Add parm to limit search to the named
9072 (merge_in_block): Attempt to use a mem insn that is the next
9073 use of the original regno.
9075 2018-02-27 Martin Sebor <msebor@redhat.com>
9078 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
9079 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
9081 2018-02-27 Martin Sebor <msebor@redhat.com>
9083 PR translation/84207
9084 * diagnostic-core.h (warning_n, error_n, inform_n): Change
9085 n argument to unsigned HOST_WIDE_INT.
9086 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
9087 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
9088 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
9089 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
9091 2018-02-27 Richard Biener <rguenther@suse.de>
9093 PR tree-optimization/84512
9094 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
9095 Do not use the estimate returned from record_stmt_cost for
9096 the scalar iteration cost but sum properly using add_stmt_cost.
9098 2018-02-27 Richard Biener <rguenther@suse.de>
9100 PR tree-optimization/84466
9101 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
9102 Adjust last change to less strictly validate use operands.
9104 2018-02-27 Martin Liska <mliska@suse.cz>
9106 PR gcov-profile/84548
9107 * gcov.c (process_file): Allow partial overlap and consider it
9108 also as group functions.
9109 (output_lines): Properly calculate range of lines for a group.
9111 2018-02-27 Martin Liska <mliska@suse.cz>
9113 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
9114 'ggc' suffixes. Change first column width.
9115 (timer::print): Fix formatting of the column.
9117 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
9119 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
9120 preserve inline entry blocks for the sake of debug inline
9121 entry point markers alone.
9122 (remove_unused_locals): Suggest in comments a better place to
9123 force the preservation of inline entry blocks that are
9124 otherwise unused, but do not preserve them.
9126 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9128 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
9130 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9133 * config/i386/constraints.md (Bs): Replace
9134 ix86_indirect_branch_register with
9135 TARGET_INDIRECT_BRANCH_REGISTER.
9137 * config/i386/i386.md (indirect_jump): Likewise.
9138 (tablejump): Likewise.
9139 (*sibcall_memory): Likewise.
9140 (*sibcall_value_memory): Likewise.
9141 Peepholes of indirect call and jump via memory: Likewise.
9142 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
9143 (*sibcall_value_GOT_32): Likewise.
9144 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9145 (GOT_memory_operand): Likewise.
9146 (call_insn_operand): Likewise.
9147 (sibcall_insn_operand): Likewise.
9148 (GOT32_symbol_operand): Likewise.
9149 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
9151 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
9153 PR rtl-optimization/83496
9154 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
9155 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
9156 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
9157 redundant insn, if any.
9158 (relax_delay_slots): Likewise.
9159 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
9161 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
9163 PR tree-optimization/83965
9164 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
9165 that grouped statements are part of a reduction chain. Return
9166 true if the statement is not marked as a reduction itself but
9168 (vect_recog_dot_prod_pattern): Don't check whether the statement
9169 is part of a group here.
9170 (vect_recog_sad_pattern): Likewise.
9171 (vect_recog_widen_sum_pattern): Likewise.
9173 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
9176 * final.c (rest_of_clean_state): Also look for calls inside sequences.
9178 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9181 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
9183 (ix86_output_indirect_function_return): New prototype.
9184 (ix86_split_simple_return_pop_internal): Likewise.
9185 * config/i386/i386.c (indirect_return_via_cx): New.
9186 (indirect_return_via_cx_bnd): Likewise.
9187 (indirect_thunk_name): Handle return va CX_REG.
9188 (output_indirect_thunk_function): Create alias for
9189 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
9190 (ix86_output_indirect_jmp): Remove the bool argument.
9191 (ix86_output_indirect_function_return): New function.
9192 (ix86_split_simple_return_pop_internal): Likewise.
9193 * config/i386/i386.md (*indirect_jump): Don't pass false
9194 to ix86_output_indirect_jmp.
9195 (*tablejump_1): Likewise.
9196 (simple_return_pop_internal): Change it to define_insn_and_split.
9197 Call ix86_split_simple_return_pop_internal to split it for
9199 (simple_return_indirect_internal): Call
9200 ix86_output_indirect_function_return instead of
9201 ix86_output_indirect_jmp.
9203 2018-02-26 Jakub Jelinek <jakub@redhat.com>
9206 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
9207 memset and value initialization afterwards.
9209 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
9211 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
9213 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9216 * common/config/aarch64/aarch64-common.c
9217 (aarch_option_optimization_table[]): Switch
9218 off fomit-frame-pointer
9220 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
9221 Chung-Ju Wu <jasonwucj@gmail.com>
9223 * config/nds32/nds32-multiple.md (load_multiple): Disallow
9225 (store_multiple): Ditto.
9227 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
9229 * config.gcc: Add --with-cpu support for nds32 target.
9230 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
9231 * config/nds32/nds32.opt: Add -mcpu= option.
9233 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
9235 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
9236 isel=yes): Warn for these deprecated options.
9238 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
9240 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
9241 ISA_2_5_MASKS_EMBEDDED.
9243 2018-02-23 Jakub Jelinek <jakub@redhat.com>
9245 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
9246 p->max as pointers rather than using iterative_hash_expr.
9248 2018-02-23 Carl Love <cel@us.ibm.com>
9250 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
9251 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
9253 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
9254 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
9255 P8V_BUILTIN_VEC_VUNSIGNED2.
9257 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
9260 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
9261 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
9263 * lra-constraints.c (curr_insn_transform): Set up
9264 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
9266 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
9267 * lra-eliminations.c (spill_pseudos): Ditto.
9268 (process_insn_for_elimination): Ditto.
9269 * lra-lives.c (reg_early_clobber_p): Use the new macros.
9270 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
9271 LRA_NON_CLOBBERED_ALT.
9273 2018-02-22 Martin Sebor <msebor@redhat.com>
9275 PR tree-optimization/84480
9276 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
9277 to maybe_diag_stxncpy_trunc. Call it.
9278 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
9279 from gimple_fold_builtin_strcpy. Print inlining stack.
9280 (handle_builtin_stxncpy): Print inlining stack.
9281 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
9283 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
9286 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
9287 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
9288 and -fcheck-pointer-bounds are used together.
9289 (indirect_thunk_prefix): New enum.
9290 (indirect_thunk_need_prefix): New function.
9291 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
9292 "_nt" instead of "_bnd" for NOTRACK prefix.
9293 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
9294 (output_indirect_thunk_function): Likewise.
9296 (ix86_code_end): Update output_indirect_thunk_function calls.
9297 (ix86_output_indirect_branch_via_reg): Replace
9298 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
9299 (ix86_output_indirect_branch_via_push): Likewise.
9300 (ix86_output_function_return): Likewise.
9301 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
9302 incompatible with -fcf-protection=branch and
9303 -fcheck-pointer-bounds.
9305 2018-02-22 Steve Ellcey <sellcey@cavium.com>
9308 * config/aarch64/aarch64.c (aarch64_print_address_internal):
9309 Change gcc_assert call to output_operand_lossage.
9311 2018-02-22 Steve Ellcey <sellcey@cavium.com>
9313 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
9315 2018-02-22 DJ Delorie <dj@redhat.com>
9316 Sebastian Perta <sebastian.perta@renesas.com>
9317 Oleg Endo <olegendo@gcc.gnu.org>
9319 * config/rx/rx.c (rx_rtx_costs): New function.
9320 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
9322 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
9324 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
9326 2018-02-22 Martin Liska <mliska@suse.cz>
9329 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
9330 Add "native" as a possible value.
9332 2018-02-22 Martin Liska <mliska@suse.cz>
9335 * config/i386/i386.c (ix86_option_override_internal):
9336 Add "native" as a possible value for -march and -mtune.
9338 2018-02-22 Jakub Jelinek <jakub@redhat.com>
9341 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
9342 to all type variants.
9344 PR tree-optimization/84503
9345 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
9346 width as info->bitpos + info->bitsize - start.
9347 (merged_store_group::merge_overlapping): Simplify width computation.
9348 (check_no_overlap): New function.
9349 (imm_store_chain_info::try_coalesce_bswap): Compute expected
9350 start + width and last_order of the group, fail if check_no_overlap
9352 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
9353 to group if check_no_overlap fails.
9355 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
9357 * config/rs6000/altivec.md: Delete contraint arguments to
9358 define_expand, define_split, and define_peephole2, and in
9359 define_insn_and_split if always unused.
9360 * config/rs6000/darwin.md: Ditto.
9361 * config/rs6000/dfp.md: Ditto.
9362 * config/rs6000/rs6000.md: Ditto.
9363 * config/rs6000/sync.md: Ditto.
9364 * config/rs6000/vector.md: Ditto.
9365 * config/rs6000/vsx.md: Ditto.
9367 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
9369 * config/rs6000/altivec.md: Write output control strings as braced
9370 blocks instead of double-quoted strings.
9371 * config/rs6000/darwin.md: Ditto.
9372 * config/rs6000/rs6000.md: Ditto.
9373 * config/rs6000/vector.md: Ditto.
9374 * config/rs6000/vsx.md: Ditto.
9376 2018-02-21 Jason Merrill <jason@redhat.com>
9378 PR c++/84314 - ICE with templates and fastcall attribute.
9379 * attribs.c (build_type_attribute_qual_variant): Remove assert.
9381 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
9383 * ipa-cp.c (determine_versionability): Fix comment typos.
9385 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
9388 * ipa-cp.c (determine_versionability): Do not version functions caling
9391 2018-02-21 Martin Liska <mliska@suse.cz>
9394 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
9395 Add "native" as a possible value.
9396 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
9397 the macro when native cpu detection is available.
9399 2018-02-21 Martin Liska <mliska@suse.cz>
9402 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
9403 Add "native" as a possible value.
9404 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
9405 when native cpu detection is available.
9407 2018-02-21 Jakub Jelinek <jakub@redhat.com>
9408 Martin Sebor <msebor@redhat.com>
9410 PR tree-optimization/84478
9411 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
9413 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
9414 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
9415 support which is conservatively correct, for 2 only stay conservative
9416 for maxlen. Formatting and comment capitalization fixes. Add STRICT
9417 argument to the 2 argument get_range_strlen, adjust 6 arg
9418 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
9420 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
9421 (gimple_fold_builtin_strlen): Pass true as last argument to
9424 2018-02-20 Martin Sebor <msebor@redhat.com>
9427 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
9428 (builtin_memref::set_base_and_offset): Same. Handle inner references.
9429 (builtin_memref::builtin_memref): Factor out parts into
9430 set_base_and_offset and call it.
9432 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
9435 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
9436 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
9437 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
9438 search at the associated MODE_INT.
9440 2018-02-20 Jeff Law <law@redhat.com>
9443 PR tree-optimization/81592
9445 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
9446 for range data rather than using global data.
9447 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
9448 range data rather than using global data.
9449 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
9450 pass it to children as needed.
9451 (struct directive::fmtresult): Similarly.
9452 (struct directive::set_width): Similarly.
9453 (struct directive::set_precision): Similarly.
9454 (format_integer, format_directive, parse_directive): Similarly.
9455 (format_none): Accept unnamed vr_values parameter.
9456 (format_percent, format_floating, format_character): Similarly.
9457 (format_string, format_plain): Similarly.
9458 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
9459 the EVRP range analyzer for range data rather than using global data.
9460 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
9461 gimple-ssa-evrp-analyze.h
9462 (class sprintf_dom_walker): Add after_dom_children member function.
9463 Add evrp_range_analyzer member.
9464 (sprintf_dom_walker::before_dom_children): Call into the EVRP
9465 range analyzer as needed.
9466 (sprintf_dom_walker::after_dom_children): New member function.
9467 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
9469 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
9470 (evrp_range_analyzer::pop_to_marker): Likewise.
9472 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
9474 PR tree-optimization/84419
9475 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
9476 with the required type if its current type is compatible but
9479 2018-02-20 Jakub Jelinek <jakub@redhat.com>
9482 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
9483 after vectorization.
9485 2018-02-20 Martin Liska <mliska@suse.cz>
9488 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
9489 possible values if we don't have a hint.
9491 2018-02-20 Martin Liska <mliska@suse.cz>
9495 * final.c (shorten_branches): Build align_tab array with one
9497 * opts.c (finish_options): Add alignment option limit check.
9498 (MAX_CODE_ALIGN): Likewise.
9499 (MAX_CODE_ALIGN_VALUE): Likewise.
9500 * doc/invoke.texi: Document maximum allowed option value for
9501 all -falign-* options.
9503 2018-02-19 Jakub Jelinek <jakub@redhat.com>
9506 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
9507 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
9508 * var-tracking.c (emit_note_insn_var_location): Remove all references
9509 to NOTE_INSN_CALL_ARG_LOCATION.
9510 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
9511 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
9512 Use copy_rtx_if_shared.
9513 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
9514 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
9515 (dwarf2out_var_location): Remove handling of
9516 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
9518 * final.c (final_scan_insn): Remove all references to
9519 NOTE_INSN_CALL_ARG_LOCATION.
9520 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
9521 before dumping final insns.
9522 * except.c (emit_note_eh_region_end): Remove all references to
9523 NOTE_INSN_CALL_ARG_LOCATION.
9524 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
9525 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
9526 * config/arc/arc.c (hwloop_optimize): Likewise.
9527 * config/arm/arm.c (create_fix_barrier): Likewise.
9528 * config/s390/s390.c (s390_chunkify_start): Likewise.
9529 * config/sh/sh.c (find_barrier): Likewise.
9530 * config/i386/i386.c (rest_of_insert_endbranch,
9531 ix86_seh_fixup_eh_fallthru): Likewise.
9532 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
9533 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
9534 * config/frv/frv.c (frv_function_prologue): Likewise.
9535 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
9537 (note_outside_basic_block_p): Remove all references to
9538 NOTE_INSN_CALL_ARG_LOCATION.
9539 * gengtype.c (adjust_field_rtx_def): Likewise.
9540 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
9542 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
9543 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
9546 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
9549 PR tree-optimization/84452
9550 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
9551 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
9554 2018-02-19 Martin Liska <mliska@suse.cz>
9557 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
9559 2018-02-19 Martin Liska <mliska@suse.cz>
9560 Richard Sandiford <richard.sandiford@linaro.org>
9562 PR tree-optimization/82491
9563 * gimple-fold.c (get_base_constructor): Make earlier bail out
9566 2018-02-19 Carl Love <cel@us.ibm.com>
9568 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
9569 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
9571 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
9572 P8V_BUILTIN_VEC_NEG.
9574 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
9576 * config/rl78/rl78.md (movdf): New define expand.
9578 2018-02-19 Martin Liska <mliska@suse.cz>
9581 * doc/invoke.texi: Fix typo.
9582 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
9584 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
9586 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
9587 handle rs6000_single_float and rs6000_double_float specially for
9590 2018-02-16 Jeff Law <law@redhat.com>
9592 * config/rx/rx.c (add_pop_cfi_notes): New function.;
9595 2018-02-16 Jakub Jelinek <jakub@redhat.com>
9598 * ipa-inline.c (inline_small_functions): Fix a typo.
9600 2018-02-16 Nathan Sidwell <nathan@acm.org>
9602 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
9604 2018-02-16 Carl Love <cel@us.ibm.com>
9606 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
9607 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
9608 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
9609 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
9610 expansion to P8V_BUILTIN_VEC_FLOAT2.
9612 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
9614 PR rtl-optimization/70023
9615 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
9616 src_regno into account.
9618 2018-02-16 Carl Love <cel@us.ibm.com>
9620 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
9621 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
9622 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
9623 * config/rs6000/rs6000.c: Remove case statements for
9624 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
9625 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
9626 and P9V_BUILTIN_VEC_VINSERT4B.
9627 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
9628 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
9629 * config/rs6000/vsx.md:
9630 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
9633 2018-02-16 Carl Love <cel@us.ibm.com>
9635 * config/rs6000/altivec.h: Add builtin names vec_extract4b
9637 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
9639 * config/rs6000/rs6000-c.c: Add the definitions for
9640 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
9641 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
9642 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
9643 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
9644 definition for insert4b and define insn *insert3b_internal.
9645 * doc/extend.texi: Add documentation for vec_extract4b.
9647 2018-02-16 Nathan Sidwell <nathan@acm.org>
9649 * doc/extend.texi (Backwards Compatibility): Mention friend
9650 injection. Note for-scope is deprecated.
9651 * doc/invoke.texi (-ffriend-injection): Deprecate.
9653 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
9655 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
9656 that moved to I2, also allow destinations that are a paradoxical
9657 subreg (instead of a normal reg).
9659 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
9662 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
9665 2018-02-16 Richard Biener <rguenther@suse.de>
9667 PR tree-optimization/84037
9668 PR tree-optimization/84016
9670 * config/i386/i386.c (ix86_builtin_vectorization_cost):
9671 Adjust vec_construct for the fact we need additional higher latency
9672 128bit inserts for AVX256 and AVX512 vector builds.
9673 (ix86_add_stmt_cost): Scale vector construction cost for
9676 2018-02-16 Richard Biener <rguenther@suse.de>
9678 PR tree-optimization/84417
9679 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
9680 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
9681 (non_rewritable_lvalue_p): Likewise, use poly-ints.
9683 2018-02-16 Martin Liska <mliska@suse.cz>
9686 * internal-fn.def (ASAN_CHECK): Set proper flags.
9687 (ASAN_MARK): Likewise.
9689 2018-02-16 Julia Koval <julia.koval@intel.com>
9691 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
9692 from PTA_CANNONLAKE.
9694 2018-02-16 Jakub Jelinek <jakub@redhat.com>
9697 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
9698 Use ++iter rather than iter++ for std::list iterators.
9699 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
9700 defer deleting them until all nodes in the forest are processed. Do
9701 free even leaf nodes. Change to_process into auto_vec.
9704 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
9705 * vec.h (vec_default_construct): Use memset instead of placement new
9706 if BROKEN_VALUE_INITIALIZATION is defined.
9707 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
9708 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
9711 PR rtl-optimization/83723
9712 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
9713 * lra.c (lra_substitute_pseudo): Likewise. If true, use
9714 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
9716 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
9718 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
9720 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
9722 PR rtl-optimization/81443
9723 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
9724 from inner REGs to paradoxical SUBREGs.
9726 2018-02-16 Richard Biener <rguenther@suse.de>
9728 PR tree-optimization/84399
9729 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
9730 For operands we can analyze at their definition make sure we can
9731 analyze them at each use as well.
9733 2018-02-16 Richard Biener <rguenther@suse.de>
9735 PR tree-optimization/84190
9736 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
9737 volatile accesses if the decl isn't volatile.
9739 2018-02-15 Jason Merrill <jason@redhat.com>
9741 PR c++/84314 - ICE with templates and fastcall attribute.
9742 * attribs.c (build_type_attribute_qual_variant): Don't clobber
9743 TYPE_CANONICAL on an existing type.
9745 2018-02-15 Jakub Jelinek <jakub@redhat.com>
9747 PR tree-optimization/84383
9748 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
9749 dstoff nor call operand_equal_p if dstbase is NULL.
9751 PR tree-optimization/84334
9752 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
9753 also a CONSTANT_CLASS_P, punt.
9755 2018-02-14 Jim Wilson <jimw@sifive.com>
9757 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
9758 first SMALL_OPERAND check. New local min_second_step. Move assert
9759 to where locals are set. Add TARGET_RVC support.
9760 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
9762 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
9764 * doc/invoke.texi: Correct -Wformat-overflow code sample.
9766 2018-02-14 Martin Sebor <msebor@redhat.com>
9768 PR tree-optimization/83698
9769 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
9770 arrays constrain the offset range to their bounds.
9771 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
9772 (builtin_access::overlap): Avoid setting the size of overlap if it's
9774 (maybe_diag_overlap): Also consider arrays when deciding what values
9775 of offsets to include in diagnostics.
9777 2018-02-14 Martin Sebor <msebor@redhat.com>
9780 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
9781 that correspond to the kind of a declaration.
9783 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
9786 * config/pa/pa.md: Load address of PIC label using the linkage table
9787 if the label is nonlocal.
9789 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
9791 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
9792 warning message if user requests -maltivec=be.
9793 * doc/invoke.texi: Document deprecation of -maltivec=be.
9795 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
9798 * config/rs6000/rs6000-c.c: Update definitions for
9799 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
9800 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
9802 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9805 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
9806 add _get_ssp intrinsics. Remove argument from
9807 __builtin_ia32_rdssp[d|q].
9808 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
9809 * config/i386/i386-builtin.def: Remove argument from
9810 __builtin_ia32_rdssp[d|q].
9811 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
9812 ix86_expand_special_args_builtin for _rdssp[d|q].
9813 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
9814 Clear register before usage.
9815 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
9816 Add documentation for new _get_ssp and _inc_ssp intrinsics.
9818 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
9820 PR tree-optimization/84357
9821 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
9822 operand 1 of an ARRAY_REF too.
9824 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
9827 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
9828 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
9830 (set_of_reg): New struct.
9831 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
9832 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
9833 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
9835 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
9836 Split into bitclr, bitset, bitinvert patterns if appropriate.
9837 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
9838 use rx_fuse_in_memory_bitop.
9839 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
9840 to named insn, correct maximum insn length.
9842 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
9845 * machmode.def: Define a complex mode for PARTIAL_INT.
9846 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
9848 * doc/rtl.texi: Document CSPImode.
9849 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
9851 (msp430_hard_regno_nregs_with_padding): Likewise.
9853 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
9856 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
9858 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
9860 PR rtl-optimization/84169
9861 * combine.c (try_combine): New variable split_i2i3. Set it to true if
9862 we generated a parallel as new i3 and we split that to new i2 and i3
9863 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
9864 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
9865 those to i2, not i1. Partially rewrite this scan code.
9867 2018-02-13 Jakub Jelinek <jakub@redhat.com>
9870 * stor-layout.c (place_field): For variable length fields, adjust
9871 offset_align afterwards not just based on the field's alignment,
9872 but also on the size.
9875 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
9876 of exps and logs in the use_exp2 case.
9878 2018-02-13 Jeff Law <law@redhat.com>
9880 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
9883 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
9886 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
9890 * common.opt (gas-loc-support, gas-locview-support): New.
9891 (ginline-points, ginternal-reset-location-views): New.
9892 * doc/invoke.texi: Document them. Use @itemx where intended.
9893 (gvariable-location-views): Adjust.
9894 * target.def (reset_location_view): New.
9895 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
9896 (TARGET_RESET_LOCATION_VIEW): New.
9897 * doc/tm.texi: Rebuilt.
9898 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
9899 (dwarf2out_default_as_locview_support): New.
9900 (output_asm_line_debug_info): Use option variables.
9901 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
9902 (output_loc_list): Likewise.
9903 (add_high_low_attributes): Check option variables.
9904 Don't output entry view attribute in strict mode.
9905 (gen_inlined_subroutine_die): Check option variables.
9906 (dwarf2out_inline_entry): Likewise.
9907 (init_sections_and_labels): Likewise.
9908 (dwarf2out_early_finish): Likewise.
9909 (maybe_reset_location_view): New, from...
9910 (dwarf2out_var_location): ... here. Call it.
9911 * debug.h (dwarf2out_default_as_loc_support): Declare.
9912 (dwarf2out_default_as_locview_support): Declare.
9913 * hooks.c (hook_int_rtx_insn_0): New.
9914 * hooks.h (hook_int_rtx_insn_0): Declare.
9915 * toplev.c (process_options): Take -gas-loc-support and
9916 -gas-locview-support from dwarf2out. Enable
9917 -gvariable-location-views by default only with locview
9918 assembler support. Enable -ginternal-reset-location-views by
9919 default only if the target defines the corresponding hook.
9920 Enable -ginline-points by default if location views are
9921 enabled; force it disabled if statement frontiers are
9923 * tree-inline.c (expand_call_inline): Check option variables.
9924 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
9926 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
9928 PR tree-optimization/84321
9929 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
9930 handling. Also check whether the anti-range contains any values
9931 that satisfy the mask; switch to a VR_RANGE if not.
9933 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
9936 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
9938 2018-02-13 Martin Jambor <mjambor@suse.cz>
9941 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
9942 of call statements, also set location of a load to a temporary.
9944 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
9946 * config/rl78/rl78.c (add_vector_labels): New function.
9947 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
9948 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
9949 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
9950 which checks that no arguments are passed.
9951 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
9952 * doc/extend.texi: Documentation for the new attribute.
9954 2018-02-13 Andreas Schwab <schwab@suse.de>
9956 * config/riscv/linux.h (CPP_SPEC): Define.
9958 2018-02-13 Jakub Jelinek <jakub@redhat.com>
9961 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
9962 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
9963 OPTION_MASK_ISA_AES as first argument to def_builtin_const
9964 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
9965 instead of OPTION_MASK_ISA_PCLMUL as first argument to
9966 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
9967 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
9968 temporarily for AES and PCLMUL builtins.
9970 PR tree-optimization/84339
9971 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
9972 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
9976 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
9977 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
9978 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
9980 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
9982 * omp-simd-clone.h: New file.
9983 * omp-simd-clone.c: Include omp-simd-clone.h.
9984 (expand_simd_clones): No longer static.
9985 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
9986 cgraph.h and omp-simd-clone.h.
9987 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
9988 (vect_recog_widen_shift_pattern): Formatting fix.
9989 (vect_pattern_recog_1): Don't check optab for calls.
9992 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
9993 operands[2] into a REG before using gen_lowpart on it.
9995 2018-02-12 Jeff Law <law@redhat.com>
9998 * config/sh/sh.c (find_barrier): Consider a sibling call
10001 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
10002 successfully back substituting a reg.
10004 2018-02-12 Richard Biener <rguenther@suse.de>
10006 PR tree-optimization/84037
10007 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
10008 parameter, move visited init to caller.
10009 (vect_slp_analyze_operations): Separate cost from validity
10010 check, initialize visited once for all instances.
10011 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
10013 * tree-vect-stmts.c (vect_model_simple_cost): Make early
10015 (vect_model_promotion_demotion_cost): Likewise.
10016 (vectorizable_bswap): Guard cost modeling with !slp_node
10017 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
10019 (vectorizable_call): Likewise.
10020 (vectorizable_conversion): Likewise.
10021 (vectorizable_assignment): Likewise.
10022 (vectorizable_shift): Likewise.
10023 (vectorizable_operation): Likewise.
10024 (vectorizable_store): Likewise.
10025 (vectorizable_load): Likewise.
10026 (vectorizable_condition): Likewise.
10027 (vectorizable_comparison): Likewise.
10029 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
10032 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
10033 (ASAN_MARK): Fix fnspec to account for return value, change pointer
10034 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
10036 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
10038 PR middle-end/83665
10039 * params.def (inline-min-speedup): Increase from 8 to 15.
10040 (max-inline-insns-auto): Decrease from 40 to 30.
10041 * ipa-split.c (consider_split): Add some buffer for function to
10042 be considered inlining candidate.
10043 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
10046 2018-02-12 Richard Biener <rguenther@suse.de>
10048 PR tree-optimization/84037
10049 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
10050 matched stmts if we cannot swap the non-matched ones.
10052 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10054 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
10055 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
10056 _mm_maskz_scalef_round_ss): New intrinsics.
10057 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
10058 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
10059 __builtin_ia32_scalefss_round): Remove.
10060 (__builtin_ia32_scalefsd_mask_round,
10061 __builtin_ia32_scalefss_mask_round): New intrinsics.
10062 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
10063 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
10064 ((match_operand:VF_128 2 "<round_nimm_predicate>"
10065 "<round_constraint>")): Changed to ...
10066 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
10067 "<round_scalar_constraint>")): ... this.
10068 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
10069 %0, %1, %2<round_op3>}"): Changed to ...
10070 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
10071 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
10072 %2<round_scalar_mask_op3>}"): ... this.
10073 * config/i386/subst.md (round_scalar_nimm_predicate): New.
10075 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10077 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
10078 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
10079 (_mm_maskz_sqrt_round_ss): New intrinsics.
10080 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
10081 (__builtin_ia32_sqrtsd_mask_round)
10082 (__builtin_ia32_sqrtss_mask_round): New builtins.
10083 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
10084 (__builtin_ia32_sqrtss_round): Remove.
10085 (__builtin_ia32_sqrtsd_mask_round)
10086 (__builtin_ia32_sqrtss_mask_round): New builtins.
10087 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
10088 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
10089 ((match_operand:VF_128 1 "vector_operand"
10090 "xBm,<round_constraint>")): Changed to ...
10091 ((match_operand:VF_128 1 "vector_operand"
10092 "xBm,<round_scalar_constraint>")): ... this.
10093 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
10094 %0, %2, %<iptr>1<round_op3>}): Changed to ...
10095 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
10096 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
10097 %<iptr>1<round_scalar_mask_op3>}): ... this.
10098 ((set_attr "prefix" "<round_prefix>")): Changed to ...
10099 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
10101 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
10104 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
10105 Cast vec_cmpeq result to correct type.
10106 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
10107 Cast vec_cmpgt result to correct type.
10109 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
10111 * final.c (final_scan_insn_1): Renamed from...
10112 (final_scan_insn): ... this. New wrapper, to recover
10113 seen from the outermost call in recursive ones.
10114 * config/sparc/sparc.c (output_return): Drop seen from call.
10115 (output_sibcall): Likewise.
10116 * config/visium/visium.c (output_branch): Likewise.
10118 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
10120 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
10123 2018-02-10 Alan Modra <amodra@gmail.com>
10126 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
10127 Specify LR as an input.
10129 2018-02-10 Jakub Jelinek <jakub@redhat.com>
10132 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
10133 remove_member_access_dummy_vars): New functions.
10134 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
10135 lower_omp_1, execute_lower_omp): Use them.
10137 PR rtl-optimization/84308
10138 * shrink-wrap.c (spread_components): Release todo vector.
10140 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
10142 PR rtl-optimization/57193
10143 * ira-color.c (struct allocno_color_data): Add member
10144 conflict_allocno_hard_prefs.
10145 (update_conflict_allocno_hard_prefs): New.
10146 (bucket_allocno_compare_func): Add a preference based on
10147 conflict_allocno_hard_prefs.
10148 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
10149 (color_allocnos): Remove a dead code. Initiate
10150 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
10152 2018-02-09 Jakub Jelinek <jakub@redhat.com>
10155 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
10156 constraint from =wa to wa. Avoid a subreg on the output operand,
10157 instead use a pseudo and subreg it in a move.
10158 (p9_xxbrd_<mode>): Changed to ...
10159 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
10160 (p9_xxbrd_v2df): New expander.
10161 (p9_xxbrw_<mode>): Changed to ...
10162 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
10163 (p9_xxbrw_v4sf): New expander.
10165 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
10167 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
10169 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
10172 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
10173 multiply in 32-bit mode.
10174 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
10175 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
10178 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
10180 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
10181 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
10182 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
10183 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
10185 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
10188 * dwarf2out.c (is_trivial_indirect_ref): New function.
10189 (dwarf2out_late_global_decl): Do not generate a location
10190 attribute for variables that have a non-trivial DECL_VALUE_EXPR
10191 and that are not defined in the current unit.
10193 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
10195 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
10196 instead of a libcall for UNORDERED.
10198 2018-02-09 Tamar Christina <tamar.christina@arm.com>
10201 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
10202 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
10204 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10207 * config/s390/s390.c (s390_set_current_function): Invoke
10208 s390_indirect_branch_settings also if fndecl didn't change.
10210 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
10212 * config/rs6000/rs6000.md (blockage): Set length to zero.
10214 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
10216 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
10218 2018-02-09 Jakub Jelinek <jakub@redhat.com>
10221 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
10222 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
10226 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
10227 PARALLEL incoming that failed vt_get_decl_and_offset check.
10229 PR middle-end/84237
10230 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
10231 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
10233 (get_variable_section): For decls in named .bss* sections pass true as
10234 second argument to bss_initializer_p.
10236 2018-02-09 Marek Polacek <polacek@redhat.com>
10237 Jakub Jelinek <jakub@redhat.com>
10240 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
10241 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
10242 Sync some changes from cxx_fold_indirect_ref.
10244 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
10246 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
10248 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
10249 (BLOCK_INLINE_ENTRY_LABEL): New.
10250 (dwarf2out_var_location): Disregard inline entry markers.
10251 (inline_entry_data): New struct.
10252 (inline_entry_data_hasher): New hashtable type.
10253 (inline_entry_data_hasher::hash): New.
10254 (inline_entry_data_hasher::equal): New.
10255 (inline_entry_data_table): New variable.
10256 (add_high_low_attributes): Add DW_AT_entry_pc and
10257 DW_AT_GNU_entry_view attributes if a pending entry is found
10258 in inline_entry_data_table. Add old entry_pc attribute only
10259 if debug nonbinding markers are disabled.
10260 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
10261 markers are enabled.
10262 (block_within_block_p, dwarf2out_inline_entry): New.
10263 (dwarf2out_finish): Check that no entries remained in
10264 inline_entry_data_table.
10265 * final.c (reemit_insn_block_notes): Handle inline entry notes.
10266 (final_scan_insn, notice_source_line): Likewise.
10267 (rest_of_clean_state): Skip inline entry markers.
10268 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
10270 * gimple.c (gimple_build_debug_inline_entry): New.
10271 * gimple.h (enum gimple_debug_subcode): Add
10272 GIMPLE_DEBUG_INLINE_ENTRY.
10273 (gimple_build_debug_inline_entry): Declare.
10274 (gimple_debug_inline_entry_p): New.
10275 (gimple_debug_nonbind_marker_p): Adjust.
10276 * insn-notes.def (INLINE_ENTRY): New.
10277 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
10278 inline entry marker notes.
10279 (print_insn): Likewise.
10280 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
10281 (INSN_DEBUG_MARKER_KIND): Likewise.
10282 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
10283 * tree-inline.c (expand_call_inline): Build and insert
10284 debug_inline_entry stmt.
10285 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
10286 inline entry blocks early, if nonbind markers are enabled.
10287 (dump_scope_block): Dump fragment info.
10288 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
10289 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
10290 (gimple_build_debug_inline_entry): New.
10291 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
10292 Enable/disable inline entry points too.
10293 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
10294 (DEBUG_INSN): Describe inline entry markers.
10296 * common.opt (gvariable-location-views): New.
10297 (gvariable-location-views=incompat5): New.
10298 * config.in: Rebuilt.
10299 * configure: Rebuilt.
10300 * configure.ac: Test assembler for view support.
10301 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
10302 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
10303 * dwarf2out.c (var_loc_view): New typedef.
10304 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
10305 (dwarf2out_locviews_in_attribute): New.
10306 (dwarf2out_locviews_in_loclist): New.
10307 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
10308 (enum dw_line_info_opcode): Add LI_adv_address.
10309 (struct dw_line_info_table): Add view.
10310 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
10311 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
10312 (zero_view_p): New variable.
10313 (ZERO_VIEW_P): New macro.
10314 (output_asm_line_debug_info): New.
10315 (struct var_loc_node): Add view.
10316 (add_AT_view_list, AT_loc_list): New.
10317 (add_var_loc_to_decl): Add view param. Test it against last.
10318 (new_loc_list): Add view params. Record them.
10319 (AT_loc_list_ptr): Handle loc and view lists.
10320 (view_list_to_loc_list_val_node): New.
10321 (print_dw_val): Handle dw_val_class_view_list.
10322 (size_of_die): Likewise.
10323 (value_format): Likewise.
10324 (loc_list_has_views): New.
10325 (gen_llsym): Set vl_symbol too.
10326 (maybe_gen_llsym, skip_loc_list_entry): New.
10327 (dwarf2out_maybe_output_loclist_view_pair): New.
10328 (output_loc_list): Output view list or entries too.
10329 (output_view_list_offset): New.
10330 (output_die): Handle dw_val_class_view_list.
10331 (output_dwarf_version): New.
10332 (output_compilation_unit_header): Use it.
10333 (output_skeleton_debug_sections): Likewise.
10334 (output_rnglists, output_line_info): Likewise.
10335 (output_pubnames, output_aranges): Update version comments.
10336 (output_one_line_info_table): Output view numbers in asm comments.
10337 (dw_loc_list): Determine current endview, pass it to new_loc_list.
10338 Call maybe_gen_llsym.
10339 (loc_list_from_tree_1): Adjust.
10340 (add_AT_location_description): Create view list attribute if
10341 needed, check it's absent otherwise.
10342 (convert_cfa_to_fb_loc_list): Adjust.
10343 (maybe_emit_file): Call output_asm_line_debug_info for test.
10344 (dwarf2out_var_location): Reset views as needed. Precompute
10345 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
10346 attribute. Set view.
10347 (new_line_info_table): Reset next view.
10348 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
10349 (dwarf2out_source_line): Likewise. Output view resets and labels to
10350 the assembler, or select appropriate line info opcodes.
10351 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
10352 (optimize_string_length): Catch it. Adjust.
10353 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
10354 dw_val_class_view_list, and remove it if no longer needed.
10355 (hash_loc_list): Hash view numbers.
10356 (loc_list_hasher::equal): Compare them.
10357 (optimize_location_lists): Check whether a view list symbol is
10358 needed, and whether the locview attribute is present, and
10359 whether they match. Remove the locview attribute if no longer
10361 (index_location_lists): Call skip_loc_list_entry for test.
10362 (dwarf2out_finish): Call output_asm_line_debug_info for test.
10363 Use output_dwarf_version.
10364 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
10365 (struct dw_val_node): Add val_view_list.
10366 * final.c (SEEN_NEXT_VIEW): New.
10367 (set_next_view_needed): New.
10368 (clear_next_view_needed): New.
10369 (maybe_output_next_view): New.
10370 (final_start_function): Rename to...
10371 (final_start_function_1): ... this. Take pointer to FIRST,
10372 add SEEN parameter. Emit param bindings in the initial view.
10373 (final_start_function): Reintroduce SEEN-less interface.
10374 (final): Rename to...
10375 (final_1): ... this. Take SEEN parameter. Output final pending
10376 next view at the end.
10377 (final): Reintroduce seen-less interface.
10378 (final_scan_insn): Output pending next view before switching
10379 sections or ending a block. Mark the next view as needed when
10380 outputting variable locations. Notify debug backend of section
10381 changes, and of location view changes.
10382 (rest_of_handle_final): Adjust.
10383 * toplev.c (process_options): Autodetect value for debug variable
10384 location views option. Warn on incompat5 without -gdwarf-5.
10385 * doc/invoke.texi (gvariable-location-views): New.
10386 (gvariable-location-views=incompat5): New.
10387 (gno-variable-location-views): New.
10389 2018-02-08 David Malcolm <dmalcolm@redhat.com>
10391 PR tree-optimization/84136
10392 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
10393 that the result of find_edge is non-NULL.
10395 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
10398 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
10399 storing integer register in SImode. Fix cost of 256 and 512
10400 byte aligned SSE register store.
10402 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
10404 * config/i386/i386.c (ix86_multiplication_cost): Fix
10405 multiplication cost for TARGET_AVX512DQ.
10407 2018-02-08 Marek Polacek <polacek@redhat.com>
10409 PR tree-optimization/84238
10410 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
10413 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10415 PR tree-optimization/84265
10416 * tree-vect-stmts.c (vectorizable_store): Don't treat
10417 VMAT_CONTIGUOUS accesses as grouped.
10418 (vectorizable_load): Likewise.
10420 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10422 PR tree-optimization/81635
10423 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
10424 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
10425 (test_round_for_mask): New functions.
10426 (wide_int_cc_tests): Call test_round_for_mask.
10427 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
10428 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
10429 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
10430 range returned by get_range_info.
10432 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
10435 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
10436 * symtab.c: Include builtins.h
10437 (symtab_node::output_to_lto_symbol_table_p): Move here
10438 from lto-streamer-out.c:output_symbol_p.
10439 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
10440 (output_symbol_p): Move all logic to symtab.c
10441 (produce_symtab): Update.
10443 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10445 * config/s390/s390-opts.h (enum indirect_branch): Define.
10446 * config/s390/s390-protos.h (s390_return_addr_from_memory)
10447 (s390_indirect_branch_via_thunk)
10448 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
10449 (enum s390_indirect_branch_type): Define.
10450 * config/s390/s390.c (struct s390_frame_layout, struct
10451 machine_function): Remove.
10452 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
10453 (indirect_branch_table_label_no, indirect_branch_table_name):
10455 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
10456 (enum s390_indirect_branch_option): Define.
10457 (s390_return_addr_from_memory): New function.
10458 (s390_handle_string_attribute): New function.
10459 (s390_attribute_table): Add new attribute handler.
10460 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
10461 (s390_indirect_branch_via_thunk): New function.
10462 (s390_indirect_branch_via_inline_thunk): New function.
10463 (s390_function_ok_for_sibcall): When jumping via thunk disallow
10464 sibling call optimization for non z10 compiles.
10465 (s390_emit_call): Force indirect branch target to be a single
10466 register. Add r1 clobber for non-z10 compiles.
10467 (s390_emit_epilogue): Emit return jump via return_use expander.
10468 (s390_reorg): Handle JUMP_INSNs as execute targets.
10469 (s390_option_override_internal): Perform validity checks for the
10470 new command line options.
10471 (s390_indirect_branch_attrvalue): New function.
10472 (s390_indirect_branch_settings): New function.
10473 (s390_set_current_function): Invoke s390_indirect_branch_settings.
10474 (s390_output_indirect_thunk_function): New function.
10475 (s390_code_end): Implement target hook.
10476 (s390_case_values_threshold): Implement target hook.
10477 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
10479 * config/s390/s390.h (struct s390_frame_layout)
10480 (struct machine_function): Move here from s390.c.
10481 (TARGET_INDIRECT_BRANCH_NOBP_RET)
10482 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
10483 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
10484 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
10485 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
10486 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
10487 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
10488 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
10489 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
10490 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
10491 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
10492 (mnemonic attribute): Add values which aren't recognized
10494 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
10495 pattern for branch conversion. Fix mnemonic attribute.
10496 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
10497 indirect branch via thunk if requested.
10498 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
10499 ("*indirect_jump"): Disable for branch conversion using out of
10501 ("indirect_jump_via_thunk<mode>_z10")
10502 ("indirect_jump_via_thunk<mode>")
10503 ("indirect_jump_via_inlinethunk<mode>_z10")
10504 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
10505 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
10506 ("casesi_jump_via_inlinethunk<mode>_z10")
10507 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
10508 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
10509 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
10510 ("*indirect2_jump"): Disable for branch conversion.
10511 ("casesi_jump"): Turn into expander and expand patterns for branch
10513 ("return_use"): New expander.
10514 ("*return"): Emit return via thunk and rename it to ...
10515 ("*return<mode>"): ... this one.
10516 * config/s390/s390.opt: Add new options and and enum for the
10519 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10521 * lra-constraints.c (match_reload): Unconditionally use
10522 gen_lowpart_SUBREG, rather than selecting between that
10523 and equivalent gen_rtx_SUBREG code.
10525 2018-02-08 Richard Biener <rguenther@suse.de>
10527 PR tree-optimization/84233
10528 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
10529 changed flag instead of boguously re-using phi_inserted.
10531 2018-02-08 Martin Jambor <mjambor@suse.cz>
10533 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
10534 static local variables.
10536 2018-02-08 Richard Biener <rguenther@suse.de>
10538 PR tree-optimization/84278
10539 * tree-vect-stmts.c (vectorizable_store): When looking for
10540 smaller vector types to perform grouped strided loads/stores
10541 make sure the mode is supported by the target.
10542 (vectorizable_load): Likewise.
10544 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
10546 * config/aarch64/aarch64.c (aarch64_components_for_bb):
10547 Increase LDP/STP opportunities by adding adjacent callee-saves.
10549 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
10551 PR rtl-optimization/84068
10552 PR rtl-optimization/83459
10553 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
10555 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
10557 PR tree-optimization/84224
10558 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
10559 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
10560 non-zero arguments.
10562 2018-02-07 Iain Sandoe <iain@codesourcery.com>
10565 * config/rs6000/altivec.md (*restore_world): Remove LR use.
10566 * config/rs6000/predicates.md (restore_world_operation): Adjust op
10567 count, remove one USE.
10569 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
10571 * doc/install.texi (Configuration): Document the
10572 --with-long-double-format={ibm,ieee} PowerPC configuration
10576 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
10577 Convert from define_expand to be define_insn_and_split. Rework
10578 float/double/_Float128 conversions to QI/HI/SImode to work with
10579 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
10580 conversions to QI/HImode types did a store and then a load to
10581 truncate the value. For conversions to VSX registers, don't split
10582 the insn, instead emit the code directly. Use the code iterator
10583 any_fix to combine signed and unsigned conversions.
10584 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
10585 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
10586 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
10587 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
10588 (fix_<mode>di2_hw): Likewise.
10589 (fixuns_<mode>di2_hw): Likewise.
10590 (fix_<mode>si2_hw): Likewise.
10591 (fixuns_<mode>si2_hw): Likewise.
10592 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
10593 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
10594 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
10595 fix<uns>_trunc<SFDF:mode>si2_p8.
10596 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
10597 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
10598 (fix<uns>_<mode>_mem): Likewise.
10599 (fctiw<u>z_<mode>_mem): Likewise.
10600 (fix<uns>_<mode>_mem): Likewise.
10601 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
10602 the register allocator from doing a direct move to the GPRs to do
10603 a store, and instead use the ISA 3.0 store byte/half-word from
10604 vector register instruction. For IEEE 128-bit floating point,
10605 also optimize stores of 32-bit ints.
10606 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
10608 2018-02-07 Alan Hayward <alan.hayward@arm.com>
10610 * genextract.c (push_pathstr_operand): New function to support
10612 (walk_rtx): Call push_pathstr_operand.
10613 (print_path): Support [a-zA-Z].
10615 2018-02-07 Richard Biener <rguenther@suse.de>
10617 PR tree-optimization/84037
10618 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
10619 (cse_and_gimplify_to_preheader): Declare.
10620 (vect_get_place_in_interleaving_chain): Likewise.
10621 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10623 (_loop_vec_info::~_loop_vec_info): Delete it.
10624 (cse_and_gimplify_to_preheader): New function.
10625 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
10626 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
10627 (vectorizable_load): Likewise. For grouped stores always base
10628 the IV on the first element.
10629 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
10630 condition before gimplifying.
10632 2018-02-07 Jakub Jelinek <jakub@redhat.com>
10634 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
10635 *DIV_EXPR and *MOD_EXPR.
10637 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
10640 * config/i386/i386.c (ix86_option_override_internal): Mask out
10641 the CF_SET bit when checking -fcf-protection.
10643 2018-02-07 Tom de Vries <tom@codesourcery.com>
10646 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
10649 2018-02-07 Richard Biener <rguenther@suse.de>
10651 PR tree-optimization/84204
10652 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
10655 PR tree-optimization/84205
10656 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
10657 special-case isl_ast_op_zdiv_r.
10659 PR tree-optimization/84223
10660 * graphite-scop-detection.c (gather_bbs::before_dom_children):
10661 Only add conditions from within the region.
10662 (gather_bbs::after_dom_children): Adjust.
10664 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
10667 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
10668 * config/avr/avr.md: Only post-reload split REG-REG moves if
10669 either register is GENERAL_REG_P.
10671 2018-02-07 Jakub Jelinek <jakub@redhat.com>
10673 PR tree-optimization/84235
10674 * tree-ssa-scopedtables.c
10675 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
10676 if the subtraction is performed in floating point type where NaNs are
10677 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
10678 build 1. Formatting fix.
10680 2018-02-06 Jakub Jelinek <jakub@redhat.com>
10683 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
10684 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
10685 and skip it regardless of bb boundaries. Use CALL_P macro,
10686 don't test INSN_P (insn) together with CALL_P or JUMP_P check
10687 unnecessarily, formatting fix.
10689 2018-02-06 Michael Collison <michael.collison@arm.com>
10691 * config/arm/thumb2.md:
10692 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
10693 (*thumb_mov_notscc): Ditto.
10695 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
10698 * config/rs6000/rs6000.md (su code attribute): Use "u" for
10699 unsigned_fix, not "s".
10701 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10703 * configure.ac (gcc_fn_eh_frame_ro): New function.
10704 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
10705 correct .eh_frame permissions.
10706 * configure: Regenerate.
10708 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
10710 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
10711 irrelevant options.
10713 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10715 * config/rs6000/rs6000.c (rs6000_option_override_internal):
10716 Display warning message for -mno-speculate-indirect-jumps.
10718 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
10720 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
10722 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
10724 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
10726 PR tree-optimization/84225
10727 * tree-eh.c (find_trapping_overflow): Only call
10728 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
10730 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10733 * config/i386/i386.c: Reimplement the check of possible options
10734 -mibt/-mshstk conbination. Change error messages.
10735 * doc/invoke.texi: Fix a typo: remove extra '='.
10737 2018-02-06 Marek Polacek <polacek@redhat.com>
10739 PR tree-optimization/84228
10740 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
10742 2018-02-06 Tamar Christina <tamar.christina@arm.com>
10745 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
10746 emitted arch directives.
10747 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
10748 __ARM_FEATURE_COPROC before changing architectures.
10750 2018-02-06 Richard Biener <rguenther@suse.de>
10752 * config/i386/i386.c (print_reg): Fix typo.
10753 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
10755 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
10757 * configure: Regenerate.
10759 2018-02-05 Martin Sebor <msebor@redhat.com>
10761 PR tree-optimization/83369
10762 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
10765 2018-02-05 Martin Liska <mliska@suse.cz>
10767 * doc/invoke.texi: Cherry-pick upstream r323995.
10769 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
10771 * ira.c (ira_init_register_move_cost): Adjust comment.
10773 2018-02-05 Martin Liska <mliska@suse.cz>
10775 PR gcov-profile/84137
10776 * doc/gcov.texi: Fix typo in documentation.
10778 2018-02-05 Martin Liska <mliska@suse.cz>
10780 PR gcov-profile/83879
10781 * doc/gcov.texi: Document necessity of --dynamic-list-data when
10782 using dlopen functionality.
10784 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
10786 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
10787 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
10788 _mm_maskz_range_ss, _mm_mask_range_round_ss,
10789 _mm_maskz_range_round_ss): New intrinsics.
10790 (__builtin_ia32_rangesd128_round)
10791 (__builtin_ia32_rangess128_round): Remove.
10792 (__builtin_ia32_rangesd128_mask_round,
10793 __builtin_ia32_rangess128_mask_round): New builtins.
10794 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
10795 __builtin_ia32_rangess128_round): Remove.
10796 (__builtin_ia32_rangesd128_mask_round,
10797 __builtin_ia32_rangess128_mask_round): New builtins.
10798 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
10799 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
10800 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
10801 "<round_saeonly_constraint>")): Changed to ...
10802 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
10803 "<round_saeonly_scalar_constraint>")): ... this.
10804 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
10805 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
10806 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
10807 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
10808 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
10810 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
10812 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
10814 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
10815 Remove all values except native, 8540 and 8548.
10817 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
10819 * config/i386/i386.c (ix86_output_function_return): Pass
10820 INVALID_REGNUM, instead of -1, as invalid register number to
10821 indirect_thunk_name and output_indirect_thunk.
10823 2018-02-02 Julia Koval <julia.koval@intel.com>
10825 * config.gcc: Add -march=icelake.
10826 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
10827 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
10828 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
10829 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
10830 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
10831 (processor_target_table): Add icelake.
10832 (ix86_option_override_internal): Handle new PTAs.
10833 (get_builtin_code_for_version): Handle icelake.
10834 (M_INTEL_COREI7_ICELAKE): New.
10835 (fold_builtin_cpu): Handle icelake.
10836 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
10837 * doc/invoke.texi: Add -march=icelake.
10839 2018-02-02 Julia Koval <julia.koval@intel.com>
10841 * config/i386/i386.c (ix86_option_override_internal): Change flags type
10842 to wide_int_bitmask.
10843 * wide-int-bitmask.h: New.
10845 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10848 * config/i386/i386.md: Replace Pmode with word_mode in
10849 builtin_setjmp_setup and builtin_longjmp to support x32.
10851 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
10855 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
10857 (rs6000_supported_cpu_names): New static variable.
10858 (linux_cpu_translation_table): Likewise.
10859 (elf_platform) <cpu>: Define new static variable and use it.
10860 Translate kernel AT_PLATFORM name to canonical name if needed.
10861 Error if platform name is unknown.
10863 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
10866 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
10868 2018-02-01 Jeff Law <law@redhat.com>
10871 * config/i386/i386.c (release_scratch_register_on_entry): Add new
10872 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
10873 the scratch if RELEASE_VIA_POP is false.
10874 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
10875 If we have to save a temporary register, decrement SIZE appropriately.
10876 Pass new arguments to release_scratch_register_on_entry.
10877 (ix86_adjust_stack_and_probe): Likewise.
10878 (ix86_emit_probe_stack_range): Pass new arguments to
10879 release_scratch_register_on_entry.
10881 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
10883 PR rtl-optimization/84157
10884 * combine.c (change_zero_ext): Use REG_P predicate in
10885 front of HARD_REGISTER_P predicate.
10887 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
10889 * config/avr/avr.c (avr_option_override): Move disabling of
10890 -fdelete-null-pointer-checks to...
10891 * common/config/avr/avr-common.c (avr_option_optimization_table):
10894 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10896 PR tree-optimization/81635
10897 * tree-data-ref.c (split_constant_offset_1): For types that
10898 wrap on overflow, try to use range info to prove that wrapping
10901 2018-02-01 Renlin Li <renlin.li@arm.com>
10904 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
10905 TAILCALL_ADDR_REGS.
10906 (aarch64_register_move_cost): Likewise.
10907 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
10908 TAILCALL_ADDR_REGS.
10909 (REG_CLASS_NAMES): Likewise.
10910 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
10911 TAILCALL_ADDR_REGS. Remove IP registers.
10912 * config/aarch64/aarch64.md (Ucs): Update register constraint.
10914 2018-02-01 Richard Biener <rguenther@suse.de>
10916 * domwalk.h (dom_walker::dom_walker): Add additional constructor
10917 for specifying RPO order and allow NULL for that.
10918 * domwalk.c (dom_walker::dom_walker): Likewise.
10919 (dom_walker::walk): Handle NULL RPO order.
10920 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
10922 (rewrite_update_dom_walker): Likewise.
10923 (mark_def_dom_walker): Likewise.
10925 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10927 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
10928 (aarch64_maybe_expand_sve_subreg_move): Declare.
10929 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
10930 * config/aarch64/predicates.md (aarch64_any_register_operand): New
10932 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
10933 that are semantically a reverse operation.
10934 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
10935 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
10936 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
10938 (aarch64_can_change_mode_class): For big-endian, forbid changes
10939 between two SVE modes if they have different element sizes.
10941 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10943 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
10944 the TImode handling for big-endian targets.
10946 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10948 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
10949 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
10951 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
10952 Remove BSWAP handing for big-endian targets and use the form of
10953 LD1RQ appropariate for the mode.
10955 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10957 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
10958 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
10959 duplicated element.
10961 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
10964 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
10965 check for operands that need to go through aarch64_sve_reload_be.
10967 2018-02-01 Jakub Jelinek <jakub@redhat.com>
10969 PR tree-optimization/81661
10970 PR tree-optimization/84117
10971 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
10972 * tree-eh.c: Include gimplify.h.
10973 (find_trapping_overflow, replace_trapping_overflow,
10974 rewrite_to_non_trapping_overflow): New functions.
10975 * tree-vect-loop.c: Include tree-eh.h.
10976 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
10977 * tree-data-ref.c: Include tree-eh.h.
10978 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
10980 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
10982 PR rtl-optimization/84123
10983 * combine.c (change_zero_ext): Check if hard register satisfies
10984 can_change_dest_mode before calling gen_lowpart_SUBREG.
10986 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
10989 * ira.c (ira_init_register_move_cost): Remove assert.
10991 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
10993 PR rtl-optimization/84071
10994 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
10995 * doc/tm.texi: Regenerate.
10997 2018-01-31 Richard Biener <rguenther@suse.de>
10999 PR tree-optimization/84132
11000 * tree-data-ref.c (analyze_miv_subscript): Properly
11001 check whether evolution_function_is_affine_multivariate_p
11002 before calling gcd_of_steps_may_divide_p.
11004 2018-01-31 Julia Koval <julia.koval@intel.com>
11007 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
11008 * config/i386/i386.md (rdpid_rex64) New.
11009 (rdpid): Make 32bit only.
11011 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
11014 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
11015 an IDENTIFIER_NODE for FUNCTION_TYPE's.
11017 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11020 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11022 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11024 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11026 PR rtl-optimization/84071
11027 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
11028 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
11030 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11032 * config/arc/arc.c (arc_handle_aux_attribute): New function.
11033 (arc_attribute_table): Add 'aux' attribute.
11034 (arc_in_small_data_p): Consider aux like variables.
11035 (arc_is_aux_reg_p): New function.
11036 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
11037 (arc_get_aux_arg): New function.
11038 (prepare_move_operands): Handle aux-register access.
11039 (arc_handle_aux_attribute): New function.
11040 * doc/extend.texi (ARC Variable attributes): Add subsection.
11042 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11044 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
11045 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
11046 (arc_attribute_table): Add 'uncached' attribute.
11047 (arc_print_operand): Print '.di' flag for uncached memory
11049 (arc_in_small_data_p): Do not consider for small data the uncached
11051 (arc_is_uncached_mem_p): New function.
11052 * config/arc/predicates.md (compact_store_memory_operand): Check
11053 for uncached memory accesses.
11054 (nonvol_nonimm_operand): Likewise.
11055 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
11057 2018-01-31 Jakub Jelinek <jakub@redhat.com>
11060 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
11061 falign-loops=): Add Optimization flag.
11063 2018-01-30 Jeff Law <law@redhat.com>
11066 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
11067 INT_REGISTERS_SAVED. Check it prior to calling
11068 get_scratch_register_on_entry.
11069 (ix86_adjust_stack_and_probe): Similarly.
11070 (ix86_emit_probe_stack_range): Similarly.
11071 (ix86_expand_prologue): Corresponding changes.
11073 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11076 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
11077 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
11079 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
11082 * lra-constraints.c (curr_insn_transform): Process AND in the
11085 2018-01-30 Jakub Jelinek <jakub@redhat.com>
11087 PR rtl-optimization/83986
11088 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
11089 dependence against last_pending_memory_flush in addition to
11090 pending_jump_insns.
11092 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
11094 PR tree-optimization/81611
11095 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
11098 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11101 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
11104 2018-01-30 Richard Biener <rguenther@suse.de>
11105 Jakub Jelinek <jakub@redhat.com>
11107 PR tree-optimization/84111
11108 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
11109 inner loops added during recursion, as they don't have up-to-date
11112 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
11115 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
11116 (can_inline_edge_by_limits_p): ... here.
11117 (can_early_inline_edge_p, check_callers,
11118 update_caller_keys, update_callee_keys, recursive_inlining,
11119 add_new_edges_to_heap, speculation_useful_p,
11120 inline_small_functions,
11121 inline_small_functions, flatten_function,
11122 inline_to_all_callers_1): Update.
11124 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
11126 * profile-count.c (profile_count::combine_with_ipa_count): Handle
11129 2018-01-30 Richard Biener <rguenther@suse.de>
11131 PR tree-optimization/83008
11132 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
11133 invariant and constant vector uses in stmts when they need
11134 more than one stmt.
11136 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11139 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
11140 * configure: Regenerate.
11142 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11144 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
11146 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
11147 Use gen_rtx_REG rather than gen_lowpart.
11149 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11151 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
11152 rather than 0 when creating partial subregs.
11154 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11156 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
11159 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
11162 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
11163 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
11164 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
11165 flags. This restores the settings used before the 2017-07-24.
11166 Turning off pre increment/decrement/modify allows IVOPTS to
11167 optimize DF/SF loops where the index is an int.
11169 2018-01-29 Richard Biener <rguenther@suse.de>
11170 Kelvin Nilsen <kelvin@gcc.gnu.org>
11173 * tree-vect-stmts.c (vectorizable_call): Don't call
11174 targetm.vectorize_builtin_md_vectorized_function if callee is
11177 2018-01-22 Carl Love <cel@us.ibm.com>
11179 * doc/extend.tex: Fix typo in second arg in
11180 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
11182 2018-01-29 Richard Biener <rguenther@suse.de>
11184 PR tree-optimization/84086
11185 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
11186 (flush_ssaname_freelist): When SSA names were released reset
11187 the SCEV hash table.
11189 2018-01-29 Richard Biener <rguenther@suse.de>
11191 PR tree-optimization/84057
11192 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
11193 removed paths when removing edges.
11195 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
11197 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
11198 -mfunction-return=@var{choice}.
11200 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
11202 PR diagnostic/84034
11203 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
11204 Handle CR like TAB.
11205 (layout::print_source_line): Likewise.
11206 (test_get_line_width_without_trailing_whitespace): Add test cases.
11208 2018-01-27 Jakub Jelinek <jakub@redhat.com>
11210 PR middle-end/84040
11211 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
11214 2018-01-26 Jim Wilson <jimw@sifive.com>
11216 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
11218 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
11221 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11223 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
11224 and CMP + SUB-immediate -> SUBS.
11226 2018-01-26 Martin Sebor <msebor@redhat.com>
11228 PR tree-optimization/83896
11229 * tree-ssa-strlen.c (get_string_len): Rename...
11230 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
11231 Avoid assuming length is constant.
11232 (handle_char_store): Use HOST_WIDE_INT for string length.
11234 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
11237 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
11238 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
11240 2018-01-26 Richard Biener <rguenther@suse.de>
11242 PR rtl-optimization/84003
11243 * dse.c (record_store): Only record redundant stores when
11244 the earlier store aliases at least all accesses the later one does.
11246 2018-01-26 Jakub Jelinek <jakub@redhat.com>
11248 PR rtl-optimization/83985
11249 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
11250 REG_CFA_RESTORE insns.
11251 (delete_unmarked_insns): Don't ignore separate shrink wrapping
11252 REG_CFA_RESTORE insns here.
11255 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
11256 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
11258 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11260 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
11261 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
11262 (arc_init): Likewise.
11263 (arc_override_options): Likewise.
11264 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
11266 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
11268 * config/arc/arc.h (TARGET_DBNZ): Define.
11269 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
11270 properly set the tune attribute.
11271 (dbnz): Use TARGET_DBNZ guard.
11272 * config/arc/arc.opt (mtune): Add core3 option.
11274 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11276 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
11277 recognize new pic like addresses.
11278 (arc_delegitimize_address): Clean up.
11280 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11282 * config/arc/arc-arches.def: Option mrf16 valid for all
11284 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
11285 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
11286 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
11287 * config/arc/arc-tables.opt: Regenerate.
11288 * config/arc/arc.c (arc_conditional_register_usage): Handle
11289 reduced register file case.
11290 (arc_file_start): Set must have build attributes.
11291 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
11292 mrf16 option value.
11293 * config/arc/arc.opt (mrf16): Add new option.
11294 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
11295 * config/arc/genmultilib.awk: Handle new mrf16 option.
11296 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
11297 * config/arc/t-multilib: Regenerate.
11298 * doc/invoke.texi (ARC Options): Document mrf16 option.
11300 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11302 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
11303 * config/arc/arc.c (arc_handle_secure_attribute): New function.
11304 (arc_attribute_table): Add 'secure_call' attribute.
11305 (arc_print_operand): Print secure call operand.
11306 (arc_function_ok_for_sibcall): Don't optimize tail calls when
11308 (arc_is_secure_call_p): New function. * config/arc/arc.md
11309 (call_i): Add support for sjli instruction.
11310 (call_value_i): Likewise.
11311 * config/arc/constraints.md (Csc): New constraint.
11313 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11314 John Eric Martin <John.Martin@emmicro-us.com>
11316 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
11317 * config/arc/arc.c (_arc_jli_section): New struct.
11318 (arc_jli_section): New type.
11319 (rc_jli_sections): New static variable.
11320 (arc_handle_jli_attribute): New function.
11321 (arc_attribute_table): Add jli_always and jli_fixed attribute.
11322 (arc_file_end): New function.
11323 (TARGET_ASM_FILE_END): Define.
11324 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
11325 (arc_add_jli_section): New function.
11326 (jli_call_scan): Likewise.
11327 (arc_reorg): Call jli_call_scan.
11328 (arc_output_addsi): Remove 'S' from printing asm operand.
11329 (arc_is_jli_call_p): New function.
11330 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
11332 (movhi_insn): Likewise.
11333 (movsi_insn): Likewise.
11334 (movsi_set_cc_insn): Likewise.
11335 (loadqi_update): Likewise.
11336 (load_zeroextendqisi_update): Likewise.
11337 (load_signextendqisi_update): Likewise.
11338 (loadhi_update): Likewise.
11339 (load_zeroextendhisi_update): Likewise.
11340 (load_signextendhisi_update): Likewise.
11341 (loadsi_update): Likewise.
11342 (loadsf_update): Likewise.
11343 (movsicc_insn): Likewise.
11344 (bset_insn): Likewise.
11345 (bxor_insn): Likewise.
11346 (bclr_insn): Likewise.
11347 (bmsk_insn): Likewise.
11348 (bicsi3_insn): Likewise.
11349 (cmpsi_cc_c_insn): Likewise.
11350 (movsi_ne): Likewise.
11351 (movsi_cond_exec): Likewise.
11352 (clrsbsi2): Likewise.
11353 (norm_f): Likewise.
11363 (call_i): Remove 'S' asm letter, add jli instruction.
11364 (call_value_i): Likewise.
11365 * config/arc/arc.op (mjli-always): New option.
11366 * config/arc/constraints.md (Cji): New constraint.
11367 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
11369 (subsf3_fpx): Likewise.
11370 (mulsf3_fpx): Likewise.
11371 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
11373 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
11374 function attrbutes.
11375 * doc/invoke.texi (ARC): Document mjli-always option.
11377 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
11379 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
11380 avoid addition with 0 and use incw and decw where possible.
11382 2018-01-26 Richard Biener <rguenther@suse.de>
11384 PR tree-optimization/81082
11385 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
11386 association if it requires casting to unsigned.
11387 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
11388 from fold_plusminus_mult_expr to catch important cases late when
11389 range info is available.
11391 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11393 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
11394 * configure.ac (hidden_linkonce): New test.
11395 * configure: Regenerate.
11396 * config.in: Regenerate.
11398 2018-01-26 Julia Koval <julia.koval@intel.com>
11400 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
11401 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
11402 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
11403 _mm_mask_bitshuffle_epi64_mask): Fix type.
11404 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
11405 USI_FTYPE_V4DI_V4DI_USI): Remove.
11406 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
11407 __builtin_ia32_vpshufbitqmb256_mask,
11408 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
11409 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
11410 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
11412 2018-01-26 Alan Modra <amodra@gmail.com>
11415 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
11416 UNSPEC_VBPERMQ. Sort other unspecs.
11418 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
11420 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
11422 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
11424 PR middle-end/83055
11425 * predict.c (drop_profile): Do not push/pop cfun; update also
11427 (handle_missing_profiles): Fix logic looking for zero profiles.
11429 2018-01-25 Jakub Jelinek <jakub@redhat.com>
11431 PR middle-end/83977
11432 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
11433 on functions with #pragma omp declare simd or functions with simd
11435 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
11436 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
11437 Remove trailing \n from warning_at calls.
11439 2018-01-25 Tom de Vries <tom@codesourcery.com>
11442 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
11443 for neutered workers.
11445 2018-01-24 Joseph Myers <joseph@codesourcery.com>
11448 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
11449 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
11451 2018-01-24 Jeff Law <law@redhat.com>
11454 * i386.c (get_probe_interval): Move to earlier point.
11455 (ix86_compute_frame_layout): If -fstack-clash-protection and
11456 the frame is larger than the probe interval, then use pushes
11457 to save registers rather than reg->mem moves.
11458 (ix86_expand_prologue): Remove conditional for int_registers_saved
11461 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
11464 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
11465 min/max for never referenced object.
11467 2018-01-24 Jakub Jelinek <jakub@redhat.com>
11469 PR middle-end/83977
11470 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
11472 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
11473 attributes from DECL_ATTRIBUTES (decl) without affecting
11474 DECL_ATTRIBUTES (current_function_decl).
11475 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
11476 functions with non-NULL DECL_ABSTRACT_ORIGIN.
11478 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
11480 PR tree-optimization/83979
11481 * fold-const.c (fold_comparison): Use constant_boolean_node
11482 instead of boolean_{true,false}_node.
11484 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
11486 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
11489 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11491 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11492 Simplify the clause that sets the length attribute.
11493 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11494 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
11495 clause that sets the length attribute.
11496 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11498 2018-01-24 Tom de Vries <tom@codesourcery.com>
11501 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
11502 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
11503 Add strict parameter.
11504 (prevent_branch_around_nothing): Insert dummy insn between branch to
11505 label and label with no ptx insn inbetween.
11506 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
11508 2018-01-24 Tom de Vries <tom@codesourcery.com>
11511 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
11512 for neutered threads in warp.
11513 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
11515 2018-01-24 Richard Biener <rguenther@suse.de>
11517 PR tree-optimization/83176
11518 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
11521 2018-01-24 Richard Biener <rguenther@suse.de>
11523 PR tree-optimization/82819
11524 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
11525 code generating pluses that are no-ops in the target precision.
11527 2018-01-24 Richard Biener <rguenther@suse.de>
11529 PR middle-end/84000
11530 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
11532 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11534 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
11535 to merge probabilities.
11536 * predict.c (probably_never_executed): Also mark as cold functions
11537 with global 0 profile and guessed local profile.
11538 * profile-count.c (profile_probability::combine_with_count): New
11540 * profile-count.h (profile_probability::operator*,
11541 profile_probability::operator*=, profile_probability::operator/,
11542 profile_probability::operator/=): Reduce precision to adjusted
11543 and set value to guessed on contradictory divisions.
11544 (profile_probability::combine_with_freq): Remove.
11545 (profile_probability::combine_wiht_count): Declare.
11546 (profile_count::force_nonzero):: Set to adjusted.
11547 (profile_count::probability_in):: Set quality to adjusted.
11548 * tree-ssa-tail-merge.c (replace_block_by): Use
11549 combine_with_count.
11551 2018-01-23 Andrew Waterman <andrew@sifive.com>
11552 Jim Wilson <jimw@sifive.com>
11554 * config/riscv/riscv.c (riscv_stack_boundary): New.
11555 (riscv_option_override): Set riscv_stack_boundary. Handle
11556 riscv_preferred_stack_boundary_arg.
11557 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
11558 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
11559 (STACK_BOUNDARY): Set to riscv_stack_boundary.
11560 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
11561 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
11562 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
11564 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
11567 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
11568 of struct ix86_frame.
11569 (ix86_expand_epilogue): Likewise. Add a local variable for
11570 the reg_save_offset field in struct ix86_frame.
11572 2018-01-23 Bin Cheng <bin.cheng@arm.com>
11574 PR tree-optimization/82604
11575 * tree-loop-distribution.c (enum partition_kind): New enum item
11576 PKIND_PARTIAL_MEMSET.
11577 (partition_builtin_p): Support above new enum item.
11578 (generate_code_for_partition): Ditto.
11579 (compute_access_range): Differentiate cases that equality can be
11580 proven at all loops, the innermost loops or no loops.
11581 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
11582 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
11583 (finalize_partitions, distribute_loop): Don't fuse partition of
11584 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
11585 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
11586 parloop is enabled.
11588 2018-01-23 Martin Liska <mliska@suse.cz>
11590 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
11591 order to ignore the predictor.
11592 (PRED_POLYMORPHIC_CALL): Likewise.
11593 (PRED_RECURSIVE_CALL): Likewise.
11595 2018-01-23 Martin Liska <mliska@suse.cz>
11597 * tree-profile.c (tree_profiling): Print function header to
11598 aware reader which function we are working on.
11599 * value-prof.c (gimple_find_values_to_profile): Do not print
11600 not interesting value histograms.
11602 2018-01-23 Martin Liska <mliska@suse.cz>
11604 * profile-count.h (enum profile_quality): Add
11605 profile_uninitialized as the first value. Do not number values
11606 as they are zero based.
11607 (profile_count::verify): Update sanity check.
11608 (profile_probability::verify): Likewise.
11610 2018-01-23 Nathan Sidwell <nathan@acm.org>
11612 * doc/invoke.texi (ffor-scope): Deprecate.
11614 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11616 PR tree-optimization/83510
11617 * domwalk.c (set_all_edges_as_executable): New function.
11618 (dom_walker::dom_walker): Convert bool param
11619 "skip_unreachable_blocks" to enum reachability. Move setup of
11620 edge flags to set_all_edges_as_executable and only do it when
11621 reachability is REACHABLE_BLOCKS.
11622 * domwalk.h (enum dom_walker::reachability): New enum.
11623 (dom_walker::dom_walker): Convert bool param
11624 "skip_unreachable_blocks" to enum reachability.
11625 (set_all_edges_as_executable): New decl.
11626 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
11627 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
11629 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
11630 but converting true to REACHABLE_BLOCKS.
11631 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
11633 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
11634 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
11635 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
11637 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
11638 if check_all_array_refs will be called.
11640 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11642 * tree.c (selftest::test_location_wrappers): Add more test
11645 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11647 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
11648 (selftest::test_bit_in_range): Likewise.
11650 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
11653 * doc/sourcebuild.texi (vect_float): Say that the selector
11654 only describes the situation when -funsafe-math-optimizations is on.
11655 (vect_float_strict): Document.
11657 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
11659 PR tree-optimization/83965
11660 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
11661 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
11662 instead of checking only for a reduction.
11663 (vect_recog_widen_sum_pattern): Likewise.
11665 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11667 * predict.c (probably_never_executed): Only use precise profile info.
11668 (compute_function_frequency): Skip after inlining hack since we now
11669 have quality checking.
11671 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11673 * profile-count.h (profile_probability::very_unlikely,
11674 profile_probability::unlikely, profile_probability::even): Set
11675 precision to guessed.
11677 2018-01-23 Richard Biener <rguenther@suse.de>
11679 PR tree-optimization/83963
11680 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
11681 Properly terminate dominator walk when crossing the exit edge not
11682 when visiting its source block.
11684 2018-01-23 Jakub Jelinek <jakub@redhat.com>
11687 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
11688 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
11690 2018-01-22 Jakub Jelinek <jakub@redhat.com>
11692 PR tree-optimization/83957
11693 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
11694 semicolon after for body surrounded by braces.
11696 PR tree-optimization/83081
11697 * profile-count.h (profile_probability::split): New method.
11698 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
11699 Use profile_probability::split.
11700 (do_compare_rtx_and_jump): Fix adjustment of probabilities
11701 when splitting a single conditional jump into 2.
11703 2018-01-22 David Malcolm <dmalcolm@redhat.com>
11705 PR tree-optimization/69452
11706 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
11709 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11711 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
11712 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
11713 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
11715 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11717 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
11719 * config/rl78/rl78.md (movdi): New define_expand.
11720 * config/rl78/rl78.c (rl78_split_movdi): New function.
11722 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
11725 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
11727 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
11728 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
11729 128-bit to produce an UNSPEC move to get the double word with the
11730 signbit and then a shift directly to do signbit.
11731 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
11732 implementation with a new version that just does either a direct
11733 move or a regular move. Move memory interface to separate insns.
11734 Move insns so they are next to the expander.
11735 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
11736 with signbit move. Split big and little endian case.
11737 (signbit<mode>2_dm_mem_le): Likewise.
11738 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
11739 (signbit<mode>2_dm2): Likewise.
11741 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11743 * config/rl78/rl78.md (anddi3): New define_expand.
11745 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11747 * config/rl78/rl78.md (umindi3): New define_expand.
11749 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11751 * config/rl78/rl78.md (smindi3): New define_expand.
11753 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11755 * config/rl78/rl78.md (smaxdi3): New define_expand.
11757 2018-01-22 Carl Love <cel@us.ibm.com>
11759 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
11760 LVX_V1TI): Add macro expansion.
11761 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
11762 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
11763 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
11764 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
11765 Change check to determine if the instruction is a byte reversing
11766 entry. Fix typo in comment.
11767 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
11768 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
11769 Add def_builtin calls for new builtins.
11770 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
11771 Add define_insn expansion.
11773 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11775 * config/rl78/rl78.md (umaxdi3): New define_expand.
11777 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11779 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
11780 for non-QImode registers.
11782 2018-01-22 Richard Biener <rguenther@suse.de>
11784 PR tree-optimization/83963
11785 * graphite-scop-detection.c (scop_detection::get_sese): Delay
11786 including the loop exit block.
11787 (scop_detection::merge_sese): Likewise.
11788 (scop_detection::add_scop): Do it here instead.
11790 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11792 * doc/sourcebuild.texi (arm_softfloat): Document.
11794 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
11797 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
11798 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
11799 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
11801 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11802 David Edelsohn <dje.gcc@gmail.com>
11805 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11806 Change "crset eq" to "crset 2".
11807 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11808 (*call_indirect_aix<mode>_nospec): Likewise.
11809 (*call_value_indirect_aix<mode>_nospec): Likewise.
11810 (*call_indirect_elfv2<mode>_nospec): Likewise.
11811 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
11812 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
11813 change assembly output from . to $.
11814 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11815 (indirect_jump<mode>_nospec): Change assembly output from . to $.
11816 (*tablejump<mode>_internal1_nospec): Likewise.
11818 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
11821 * config/sh/sh_optimize_sett_clrt.cc:
11822 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
11824 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
11826 PR tree-optimization/83940
11827 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
11828 offset_dt to vect_constant_def rather than vect_unknown_def_type.
11829 (vect_check_load_store_mask): Add a mask_dt_out parameter and
11830 use it to pass back the definition type.
11831 (vect_check_store_rhs): Likewise rhs_dt_out.
11832 (vect_build_gather_load_calls): Add a mask_dt argument and use
11833 it instead of a call to vect_is_simple_use.
11834 (vectorizable_store): Update calls to vect_check_load_store_mask
11835 and vect_check_store_rhs. Use the dt returned by the latter instead
11836 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
11837 instead of calls to vect_is_simple_use. Pass the scalar rather
11838 than the vector operand to vect_is_simple_use when handling
11839 second and subsequent copies of an rhs value.
11840 (vectorizable_load): Update calls to vect_check_load_store_mask
11841 and vect_build_gather_load_calls. Use the cached mask_dt and
11842 gs_info.offset_dt instead of calls to vect_is_simple_use.
11844 2018-01-20 Jakub Jelinek <jakub@redhat.com>
11846 PR middle-end/83945
11847 * tree-emutls.c: Include gimplify.h.
11848 (lower_emutls_2): New function.
11849 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
11850 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
11851 it before further processing.
11854 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
11855 UINTVAL (trueop1) instead of INTVAL (op1).
11857 2018-01-19 Jakub Jelinek <jakub@redhat.com>
11861 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
11862 INCOMING_FRAME_SP_OFFSET if not defined.
11863 (scan_trace): Add ENTRY argument. If true and
11864 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
11865 emit a note to adjust the CFA offset.
11866 (create_cfi_notes): Adjust scan_trace callers.
11867 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
11868 INCOMING_FRAME_SP_OFFSET in the CIE.
11869 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
11870 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
11872 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
11873 * doc/tm.texi: Regenerated.
11875 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
11877 PR rtl-optimization/83147
11878 * lra-constraints.c (remove_inheritance_pseudos): Use
11879 lra_substitute_pseudo_within_insn.
11881 2018-01-19 Tom de Vries <tom@codesourcery.com>
11882 Cesar Philippidis <cesar@codesourcery.com>
11885 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
11887 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
11890 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
11891 spaces for function labels.
11893 2018-01-19 Martin Liska <mliska@suse.cz>
11895 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
11896 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
11897 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
11898 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
11899 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
11900 (PRED_CONST_RETURN): Change from 69 to 65.
11901 (PRED_NULL_RETURN): Change from 91 to 71.
11902 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
11903 (PRED_LOOP_GUARD): Change from 66 to 73.
11905 2018-01-19 Martin Liska <mliska@suse.cz>
11907 * predict.c (predict_insn_def): Add new assert.
11908 (struct branch_predictor): Change type to signed integer.
11909 (test_prediction_value_range): Amend test to cover
11910 PROB_UNINITIALIZED.
11911 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
11912 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
11913 (PRED_LOOP_ITERATIONS_MAX): Likewise.
11914 (PRED_LOOP_IV_COMPARE): Likewise.
11915 * predict.h (PROB_UNINITIALIZED): Define new constant.
11917 2018-01-19 Martin Liska <mliska@suse.cz>
11919 * predict.c (dump_prediction): Add new format for
11920 analyze_brprob.py script which is enabled with -details
11922 * profile-count.h (precise_p): New function.
11924 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
11926 PR tree-optimization/83922
11927 * tree-vect-loop.c (vect_verify_full_masking): Return false if
11928 there are no statements that need masking.
11929 (vect_active_double_reduction_p): New function.
11930 (vect_analyze_loop_operations): Use it when handling phis that
11931 are not in the loop header.
11933 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
11935 PR tree-optimization/83914
11936 * tree-vect-loop.c (vectorizable_induction): Don't convert
11937 init_expr or apply the peeling adjustment for inductions
11938 that are nested within the vectorized loop.
11940 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11942 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
11945 2018-01-18 Jakub Jelinek <jakub@redhat.com>
11949 * function.h (gimplify_parameters): Add gimple_seq * argument.
11950 * function.c: Include gimple.h and options.h.
11951 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
11952 for the added local temporaries if needed.
11953 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
11954 if there are any parameter cleanups, wrap whole body into a
11955 try/finally with the cleanups.
11957 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
11960 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
11961 Use GET_MODE_CLASS for scalar floating point.
11963 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
11967 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
11968 Fix call of call_cgraph_insertion_hooks.
11970 2018-01-18 Martin Sebor <msebor@redhat.com>
11972 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
11974 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
11977 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
11980 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
11983 * common.opt: (-ffile-prefix-map): New option.
11984 * opts.c (common_handle_option): Defer it.
11985 * opts-global.c (handle_common_deferred_options): Handle it.
11986 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
11987 * file-prefix-map.h: New file.
11988 (remap_debug_filename, add_debug_prefix_map): ...here.
11989 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
11990 * final.c (debug_prefix_map, add_debug_prefix_map
11991 remap_debug_filename): Move to...
11992 * file-prefix-map.c: New file.
11993 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
11994 generalize, get rid of alloca(), use strrchr() instead of strchr().
11995 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
11996 Implement in terms of add_prefix_map().
11997 (remap_macro_filename, remap_debug_filename): Implement in term of
11999 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
12000 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
12001 * dbxout.c: Include file-prefix-map.h.
12002 * varasm.c: Likewise.
12003 * vmsdbgout.c: Likewise.
12004 * xcoffout.c: Likewise.
12005 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
12006 * doc/cppopts.texi (-fmacro-prefix-map): Document.
12007 * doc/invoke.texi (-ffile-prefix-map): Document.
12008 (-fdebug-prefix-map): Update description.
12010 2018-01-18 Martin Liska <mliska@suse.cz>
12012 * config/i386/i386.c (indirect_thunk_name): Document that also
12014 (output_indirect_thunk): Document why both instructions
12015 (pause and lfence) are generated.
12017 2018-01-18 Richard Biener <rguenther@suse.de>
12019 PR tree-optimization/83887
12020 * graphite-scop-detection.c
12021 (scop_detection::get_nearest_dom_with_single_entry): Remove.
12022 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
12023 (scop_detection::merge_sese): Re-implement with a flood-fill
12024 algorithm that properly finds a SESE region if it exists.
12026 2018-01-18 Jakub Jelinek <jakub@redhat.com>
12029 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
12030 pointer_diff optimizations use view_convert instead of convert.
12032 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12034 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12035 Generate different code for -mno-speculate-indirect-jumps.
12036 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12037 (*call_indirect_aix<mode>): Disable for
12038 -mno-speculate-indirect-jumps.
12039 (*call_indirect_aix<mode>_nospec): New define_insn.
12040 (*call_value_indirect_aix<mode>): Disable for
12041 -mno-speculate-indirect-jumps.
12042 (*call_value_indirect_aix<mode>_nospec): New define_insn.
12043 (*sibcall_nonlocal_sysv<mode>): Generate different code for
12044 -mno-speculate-indirect-jumps.
12045 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12047 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
12049 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
12050 long double type, set the flags for noting the default long double
12051 type, even if we don't pass or return a long double type.
12053 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
12056 * ipa-inline.c (flatten_function): Do not overwrite final inlining
12059 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
12061 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
12062 support for merge[hl].
12063 (fold_mergehl_helper): New helper function.
12064 (tree-vector-builder.h): New #include for tree_vector_builder usage.
12065 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
12066 (altivec_vmrglw_direct): Add xxmrglw insn.
12068 2018-01-17 Andrew Waterman <andrew@sifive.com>
12070 * config/riscv/riscv.c (riscv_conditional_register_usage): If
12071 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
12073 2018-01-17 David Malcolm <dmalcolm@redhat.com>
12076 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
12077 call the lto_location_cache before reading the
12078 DECL_SOURCE_LOCATION of the types.
12080 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
12081 Richard Sandiford <richard.sandiford@linaro.org>
12083 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
12084 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
12085 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
12086 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
12087 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
12089 * config/aarch64/constraints.md (aarch64_movti_operand):
12091 * config/aarch64/predicates.md (Uti): Add new constraint.
12093 2018-01-17 Carl Love <cel@us.ibm.com>
12095 * config/rs6000/vsx.md (define_expand xl_len_r,
12096 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
12097 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
12099 (define_expand, define_insn): Move the shift left from the
12100 define_insn to the define_expand for lxvl and stxvl instructions.
12101 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
12102 and XL_LEN_R definitions to PURE.
12104 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
12106 * config/i386/i386.c (indirect_thunk_name): Declare regno
12107 as unsigned int. Compare regno with INVALID_REGNUM.
12108 (output_indirect_thunk): Ditto.
12109 (output_indirect_thunk_function): Ditto.
12110 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
12111 in the call to output_indirect_thunk_function.
12113 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
12115 PR middle-end/83884
12116 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
12117 rather than the size of inner_type to determine the stack slot size
12118 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
12120 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
12123 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
12126 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
12128 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
12129 endian Linux systems to optionally enable multilibs for selecting
12130 the long double type if the user configured an explicit type.
12131 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
12132 have no long double multilibs if not defined.
12133 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
12134 warn if the user used -mabi={ieee,ibm}longdouble and we built
12135 multilibs for long double.
12136 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
12137 appropriate multilib option.
12138 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
12140 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
12141 for building long double multilibs.
12142 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
12144 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
12146 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
12149 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
12151 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
12154 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
12157 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
12160 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
12162 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
12163 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
12165 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
12167 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
12168 different rtl trees depending on TARGET_64BIT.
12169 (rs6000_gen_lvx): Likewise.
12171 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
12173 * config/visium/visium.md (nop): Tweak comment.
12174 (hazard_nop): Likewise.
12176 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12178 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
12179 -mspeculate-indirect-jumps.
12180 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
12181 for -mno-speculate-indirect-jumps.
12182 (*call_indirect_elfv2<mode>_nospec): New define_insn.
12183 (*call_value_indirect_elfv2<mode>): Disable for
12184 -mno-speculate-indirect-jumps.
12185 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
12186 (indirect_jump): Emit different RTL for
12187 -mno-speculate-indirect-jumps.
12188 (*indirect_jump<mode>): Disable for
12189 -mno-speculate-indirect-jumps.
12190 (*indirect_jump<mode>_nospec): New define_insn.
12191 (tablejump): Emit different RTL for
12192 -mno-speculate-indirect-jumps.
12193 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
12194 (tablejumpsi_nospec): New define_expand.
12195 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
12196 (tablejumpdi_nospec): New define_expand.
12197 (*tablejump<mode>_internal1): Disable for
12198 -mno-speculate-indirect-jumps.
12199 (*tablejump<mode>_internal1_nospec): New define_insn.
12200 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
12203 2018-01-16 Artyom Skrobov tyomitch@gmail.com
12205 * caller-save.c (insert_save): Drop unnecessary parameter. All
12208 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12209 Richard Biener <rguenth@suse.de>
12212 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
12213 return early, inline manually is_gimple_sizepos. Make sure if we
12214 call gimplify_expr we don't end up with a gimple constant.
12215 * tree.c (variably_modified_type_p): Don't return true for
12216 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
12217 * gimplify.h (is_gimple_sizepos): Remove.
12219 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12221 PR tree-optimization/83857
12222 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
12223 vectorizable_live_operation for pure SLP statements.
12224 (vectorizable_live_operation): Handle PHIs.
12226 2018-01-16 Richard Biener <rguenther@suse.de>
12228 PR tree-optimization/83867
12229 * tree-vect-stmts.c (vect_transform_stmt): Precompute
12230 nested_in_vect_loop_p since the scalar stmt may get invalidated.
12232 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12235 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
12236 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
12237 If off is not INTEGER_CST, issue a may not be aligned warning
12238 rather than isn't aligned. Use isn%'t rather than isn't.
12239 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
12241 <case MULT_EXPR>: Improve the case when bottom and one of the
12242 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
12243 operand, in that case check if the other operand is multiple of
12244 bottom divided by the INTEGER_CST operand.
12246 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12249 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
12250 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
12251 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
12252 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
12253 * config/pa/pa.c (pa_function_arg_advance): Likewise.
12254 (pa_function_arg, pa_arg_partial_bytes): Likewise.
12255 (pa_function_arg_size): New function.
12257 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12259 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
12260 in a separate statement.
12262 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12264 PR tree-optimization/83847
12265 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
12266 group gathers and scatters.
12268 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12270 PR rtl-optimization/86620
12271 * params.def (max-sched-ready-insns): Bump minimum value to 1.
12273 PR rtl-optimization/83213
12274 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
12275 to last if both are JUMP_INSNs.
12277 PR tree-optimization/83843
12278 * gimple-ssa-store-merging.c
12279 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
12280 store_immediate_info for bswap/nop orig_stores.
12282 2018-01-15 Andrew Waterman <andrew@sifive.com>
12284 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
12286 <UDIV>: Increase cost if !TARGET_DIV.
12288 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
12290 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
12291 (define_attr "cr_logical_3op"): New.
12292 (cceq_ior_compare): Adjust.
12293 (cceq_ior_compare_complement): Adjust.
12294 (*cceq_rev_compare): Adjust.
12295 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
12296 (is_cracked_insn): Adjust.
12297 (insn_must_be_first_in_group): Adjust.
12298 * config/rs6000/40x.md: Adjust.
12299 * config/rs6000/440.md: Adjust.
12300 * config/rs6000/476.md: Adjust.
12301 * config/rs6000/601.md: Adjust.
12302 * config/rs6000/603.md: Adjust.
12303 * config/rs6000/6xx.md: Adjust.
12304 * config/rs6000/7450.md: Adjust.
12305 * config/rs6000/7xx.md: Adjust.
12306 * config/rs6000/8540.md: Adjust.
12307 * config/rs6000/cell.md: Adjust.
12308 * config/rs6000/e300c2c3.md: Adjust.
12309 * config/rs6000/e500mc.md: Adjust.
12310 * config/rs6000/e500mc64.md: Adjust.
12311 * config/rs6000/e5500.md: Adjust.
12312 * config/rs6000/e6500.md: Adjust.
12313 * config/rs6000/mpc.md: Adjust.
12314 * config/rs6000/power4.md: Adjust.
12315 * config/rs6000/power5.md: Adjust.
12316 * config/rs6000/power6.md: Adjust.
12317 * config/rs6000/power7.md: Adjust.
12318 * config/rs6000/power8.md: Adjust.
12319 * config/rs6000/power9.md: Adjust.
12320 * config/rs6000/rs64.md: Adjust.
12321 * config/rs6000/titan.md: Adjust.
12323 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12325 * config/i386/predicates.md (indirect_branch_operand): Rewrite
12326 ix86_indirect_branch_register logic.
12328 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12330 * config/i386/constraints.md (Bs): Update
12331 ix86_indirect_branch_register check. Don't check
12332 ix86_indirect_branch_register with GOT_memory_operand.
12334 * config/i386/predicates.md (GOT_memory_operand): Don't check
12335 ix86_indirect_branch_register here.
12336 (GOT32_symbol_operand): Likewise.
12338 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12340 * config/i386/predicates.md (constant_call_address_operand):
12341 Rewrite ix86_indirect_branch_register logic.
12342 (sibcall_insn_operand): Likewise.
12344 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12346 * config/i386/constraints.md (Bs): Replace
12347 ix86_indirect_branch_thunk_register with
12348 ix86_indirect_branch_register.
12350 * config/i386/i386.md (indirect_jump): Likewise.
12351 (tablejump): Likewise.
12352 (*sibcall_memory): Likewise.
12353 (*sibcall_value_memory): Likewise.
12354 Peepholes of indirect call and jump via memory: Likewise.
12355 * config/i386/i386.opt: Likewise.
12356 * config/i386/predicates.md (indirect_branch_operand): Likewise.
12357 (GOT_memory_operand): Likewise.
12358 (call_insn_operand): Likewise.
12359 (sibcall_insn_operand): Likewise.
12360 (GOT32_symbol_operand): Likewise.
12362 2018-01-15 Jakub Jelinek <jakub@redhat.com>
12364 PR middle-end/83837
12365 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
12366 type rather than type addr's type points to.
12367 (expand_omp_atomic_mutex): Likewise.
12368 (expand_omp_atomic): Likewise.
12370 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12373 * config/i386/i386.c (output_indirect_thunk_function): Use
12374 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
12375 for __x86_return_thunk.
12377 2018-01-15 Richard Biener <rguenther@suse.de>
12379 PR middle-end/83850
12380 * expmed.c (extract_bit_field_1): Fix typo.
12382 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12385 * config/arm/iterators.md (VF): New mode iterator.
12386 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
12387 Remove integer-related logic from pattern.
12388 (neon_vabd<mode>_3): Likewise.
12390 2018-01-15 Jakub Jelinek <jakub@redhat.com>
12392 PR middle-end/82694
12393 * common.opt (fstrict-overflow): No longer an alias.
12394 (fwrapv-pointer): New option.
12395 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
12396 also for pointer types based on flag_wrapv_pointer.
12397 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
12398 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
12399 opts->x_flag_wrapv got set.
12400 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
12401 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
12402 POINTER_TYPE_OVERFLOW_UNDEFINED.
12403 * match.pd: Likewise in address comparison pattern.
12404 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
12406 2018-01-15 Richard Biener <rguenther@suse.de>
12409 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
12410 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
12411 Reset type names to their identifier if their TYPE_DECL doesn't
12412 have linkage (and thus is used for ODR and devirt).
12413 (save_debug_info_for_decl): Remove.
12414 (save_debug_info_for_type): Likewise.
12415 (add_tree_to_fld_list): Adjust.
12416 * tree-pretty-print.c (dump_generic_node): Make dumping of
12417 type names more robust.
12419 2018-01-15 Richard Biener <rguenther@suse.de>
12421 * BASE-VER: Bump to 8.0.1.
12423 2018-01-14 Martin Sebor <msebor@redhat.com>
12426 * builtins.c (check_access): Avoid warning when the no-warning bit
12429 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
12431 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
12432 * ira-color (allocno_hard_regs_compare): Likewise.
12434 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
12437 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
12438 Use .pushsection/.popsection.
12440 2018-01-14 Martin Sebor <msebor@redhat.com>
12443 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
12445 2018-01-14 Jakub Jelinek <jakub@redhat.com>
12447 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
12448 entry from extra_headers.
12449 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
12450 extra_headers, make the list bitwise identical to the i?86-*-* one.
12452 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12454 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
12455 -mcmodel=large with -mindirect-branch=thunk,
12456 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
12457 -mfunction-return=thunk-extern.
12458 * doc/invoke.texi: Document -mcmodel=large is incompatible with
12459 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
12460 -mfunction-return=thunk and -mfunction-return=thunk-extern.
12462 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12464 * config/i386/i386.c (print_reg): Print the name of the full
12465 integer register without '%'.
12466 (ix86_print_operand): Handle 'V'.
12467 * doc/extend.texi: Document 'V' modifier.
12469 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12471 * config/i386/constraints.md (Bs): Disallow memory operand for
12472 -mindirect-branch-register.
12474 * config/i386/predicates.md (indirect_branch_operand): Likewise.
12475 (GOT_memory_operand): Likewise.
12476 (call_insn_operand): Likewise.
12477 (sibcall_insn_operand): Likewise.
12478 (GOT32_symbol_operand): Likewise.
12479 * config/i386/i386.md (indirect_jump): Call convert_memory_address
12480 for -mindirect-branch-register.
12481 (tablejump): Likewise.
12482 (*sibcall_memory): Likewise.
12483 (*sibcall_value_memory): Likewise.
12484 Disallow peepholes of indirect call and jump via memory for
12485 -mindirect-branch-register.
12486 (*call_pop): Replace m with Bw.
12487 (*call_value_pop): Likewise.
12488 (*sibcall_pop_memory): Replace m with Bs.
12489 * config/i386/i386.opt (mindirect-branch-register): New option.
12490 * doc/invoke.texi: Document -mindirect-branch-register option.
12492 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12494 * config/i386/i386-protos.h (ix86_output_function_return): New.
12495 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
12496 set function_return_type.
12497 (indirect_thunk_name): Add ret_p to indicate thunk for function
12499 (output_indirect_thunk_function): Pass false to
12500 indirect_thunk_name.
12501 (ix86_output_indirect_branch_via_reg): Likewise.
12502 (ix86_output_indirect_branch_via_push): Likewise.
12503 (output_indirect_thunk_function): Create alias for function
12504 return thunk if regno < 0.
12505 (ix86_output_function_return): New function.
12506 (ix86_handle_fndecl_attribute): Handle function_return.
12507 (ix86_attribute_table): Add function_return.
12508 * config/i386/i386.h (machine_function): Add
12509 function_return_type.
12510 * config/i386/i386.md (simple_return_internal): Use
12511 ix86_output_function_return.
12512 (simple_return_internal_long): Likewise.
12513 * config/i386/i386.opt (mfunction-return=): New option.
12514 (indirect_branch): Mention -mfunction-return=.
12515 * doc/extend.texi: Document function_return function attribute.
12516 * doc/invoke.texi: Document -mfunction-return= option.
12518 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12520 * config/i386/i386-opts.h (indirect_branch): New.
12521 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
12522 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
12523 with local indirect jump when converting indirect call and jump.
12524 (ix86_set_indirect_branch_type): New.
12525 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
12526 (indirectlabelno): New.
12527 (indirect_thunk_needed): Likewise.
12528 (indirect_thunk_bnd_needed): Likewise.
12529 (indirect_thunks_used): Likewise.
12530 (indirect_thunks_bnd_used): Likewise.
12531 (INDIRECT_LABEL): Likewise.
12532 (indirect_thunk_name): Likewise.
12533 (output_indirect_thunk): Likewise.
12534 (output_indirect_thunk_function): Likewise.
12535 (ix86_output_indirect_branch_via_reg): Likewise.
12536 (ix86_output_indirect_branch_via_push): Likewise.
12537 (ix86_output_indirect_branch): Likewise.
12538 (ix86_output_indirect_jmp): Likewise.
12539 (ix86_code_end): Call output_indirect_thunk_function if needed.
12540 (ix86_output_call_insn): Call ix86_output_indirect_branch if
12542 (ix86_handle_fndecl_attribute): Handle indirect_branch.
12543 (ix86_attribute_table): Add indirect_branch.
12544 * config/i386/i386.h (machine_function): Add indirect_branch_type
12545 and has_local_indirect_jump.
12546 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
12548 (tablejump): Likewise.
12549 (*indirect_jump): Use ix86_output_indirect_jmp.
12550 (*tablejump_1): Likewise.
12551 (simple_return_indirect_internal): Likewise.
12552 * config/i386/i386.opt (mindirect-branch=): New option.
12553 (indirect_branch): New.
12556 (thunk-inline): Likewise.
12557 (thunk-extern): Likewise.
12558 * doc/extend.texi: Document indirect_branch function attribute.
12559 * doc/invoke.texi: Document -mindirect-branch= option.
12561 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
12564 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
12566 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
12568 * ipa-inline.c (want_inline_small_function_p): Return false if
12569 inlining has already failed with CIF_FINAL_ERROR.
12570 (update_caller_keys): Call want_inline_small_function_p before
12572 (update_callee_keys): Likewise.
12574 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
12576 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
12578 (rs6000_quadword_masked_address_p): Likewise.
12579 (quad_aligned_load_p): Likewise.
12580 (quad_aligned_store_p): Likewise.
12581 (const_load_sequence_p): Add comment to describe the outer-most loop.
12582 (mimic_memory_attributes_and_flags): New function.
12583 (rs6000_gen_stvx): Likewise.
12584 (replace_swapped_aligned_store): Likewise.
12585 (rs6000_gen_lvx): Likewise.
12586 (replace_swapped_aligned_load): Likewise.
12587 (replace_swapped_load_constant): Capitalize argument name in
12588 comment describing this function.
12589 (rs6000_analyze_swaps): Add a third pass to search for vector loads
12590 and stores that access quad-word aligned addresses and replace
12591 with stvx or lvx instructions when appropriate.
12592 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
12593 New function prototype.
12594 (rs6000_quadword_masked_address_p): Likewise.
12595 (rs6000_gen_lvx): Likewise.
12596 (rs6000_gen_stvx): Likewise.
12597 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
12598 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
12599 when memory address is aligned.
12600 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
12601 this split to select lvx instruction when memory address is aligned.
12602 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
12603 instruction when memory address is aligned.
12604 (*vsx_le_perm_load_v16qi): Likewise.
12605 (four unnamed splitters): Modify to select the stvx instruction
12606 when memory is aligned.
12608 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
12610 * predict.c (determine_unlikely_bbs): Handle correctly BBs
12611 which appears in the queue multiple times.
12613 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12614 Alan Hayward <alan.hayward@arm.com>
12615 David Sherwood <david.sherwood@arm.com>
12617 * tree-vectorizer.h (vec_lower_bound): New structure.
12618 (_loop_vec_info): Add check_nonzero and lower_bounds.
12619 (LOOP_VINFO_CHECK_NONZERO): New macro.
12620 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
12621 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
12622 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
12623 fields. Make seg_len the distance travelled, not including the
12625 (dr_direction_indicator): Declare.
12626 (dr_zero_step_indicator): Likewise.
12627 (dr_known_forward_stride_p): Likewise.
12628 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
12630 (runtime_alias_check_p): Allow runtime alias checks with
12632 (operator ==): Compare access_size and align.
12633 (prune_runtime_alias_test_list): Rework for new distinction between
12634 the access_size and seg_len.
12635 (create_intersect_range_checks_index): Likewise. Cope with polynomial
12637 (get_segment_min_max): New function.
12638 (create_intersect_range_checks): Use it.
12639 (dr_step_indicator): New function.
12640 (dr_direction_indicator): Likewise.
12641 (dr_zero_step_indicator): Likewise.
12642 (dr_known_forward_stride_p): Likewise.
12643 * tree-loop-distribution.c (data_ref_segment_size): Return
12644 DR_STEP * (niters - 1).
12645 (compute_alias_check_pairs): Update call to the dr_with_seg_len
12647 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
12648 (vect_preserves_scalar_order_p): New function, split out from...
12649 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
12650 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
12651 (vect_vfa_access_size): New function.
12652 (vect_vfa_align): Likewise.
12653 (vect_compile_time_alias): Take access_size_a and access_b arguments.
12654 (dump_lower_bound): New function.
12655 (vect_check_lower_bound): Likewise.
12656 (vect_small_gap_p): Likewise.
12657 (vectorizable_with_step_bound_p): Likewise.
12658 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
12659 depencies if the vectorization factor is 1. Convert the checks
12660 for nonzero steps into checks on the bounds of DR_STEP. Try using
12661 a bunds check for variable steps if the minimum required step is
12662 relatively small. Update calls to the dr_with_seg_len
12663 constructor and to vect_compile_time_alias.
12664 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
12666 (vect_loop_versioning): Call it.
12667 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
12669 (vect_estimate_min_profitable_iters): Account for any bounds checks.
12671 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12672 Alan Hayward <alan.hayward@arm.com>
12673 David Sherwood <david.sherwood@arm.com>
12675 * doc/sourcebuild.texi (vect_scatter_store): Document.
12676 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
12678 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
12680 * genopinit.c (main): Add supports_vec_scatter_store and
12681 supports_vec_scatter_store_cached to target_optabs.
12682 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
12683 IFN_MASK_SCATTER_STORE.
12684 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
12686 * internal-fn.h (internal_store_fn_p): Declare.
12687 (internal_fn_stored_value_index): Likewise.
12688 * internal-fn.c (scatter_store_direct): New macro.
12689 (expand_scatter_store_optab_fn): New function.
12690 (direct_scatter_store_optab_supported_p): New macro.
12691 (internal_store_fn_p): New function.
12692 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
12693 IFN_MASK_SCATTER_STORE.
12694 (internal_fn_mask_index): Likewise.
12695 (internal_fn_stored_value_index): New function.
12696 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
12697 for scatter stores.
12698 * optabs-query.h (supports_vec_scatter_store_p): Declare.
12699 * optabs-query.c (supports_vec_scatter_store_p): New function.
12700 * tree-vectorizer.h (vect_get_store_rhs): Declare.
12701 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
12702 true for scatter stores.
12703 (vect_gather_scatter_fn_p): Handle scatter stores too.
12704 (vect_check_gather_scatter): Consider using scatter stores if
12705 supports_vec_scatter_store_p.
12706 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
12707 scatter stores too.
12708 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
12709 internal_fn_stored_value_index.
12710 (check_load_store_masking): Handle scatter stores too.
12711 (vect_get_store_rhs): Make public.
12712 (vectorizable_call): Use internal_store_fn_p.
12713 (vectorizable_store): Handle scatter store internal functions.
12714 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
12715 when deciding whether the end of the group has been reached.
12716 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
12717 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
12718 (mask_scatter_store<mode>): New insns.
12720 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12721 Alan Hayward <alan.hayward@arm.com>
12722 David Sherwood <david.sherwood@arm.com>
12724 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
12725 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
12726 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
12728 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
12729 Use vect_truncate_gather_scatter_offset if we can't treat the
12730 operation as a normal gather load or scatter store.
12731 (get_group_load_store_type): Take the gather_scatter_info
12732 as argument. Try using a gather load or scatter store for
12733 single-element groups.
12734 (get_load_store_type): Update calls to get_group_load_store_type
12735 and vect_use_strided_gather_scatters_p.
12737 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12738 Alan Hayward <alan.hayward@arm.com>
12739 David Sherwood <david.sherwood@arm.com>
12741 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
12742 optional tree argument.
12743 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
12745 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
12746 but continue to use the current value as a fallback.
12747 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
12748 to compare the updates.
12749 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
12750 (get_load_store_type): Use it when handling a strided access.
12751 (vect_get_strided_load_store_ops): New function.
12752 (vect_get_data_ptr_increment): Likewise.
12753 (vectorizable_load): Handle strided gather loads. Always pass
12754 a step to vect_create_data_ref_ptr and bump_vector_ptr.
12756 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12757 Alan Hayward <alan.hayward@arm.com>
12758 David Sherwood <david.sherwood@arm.com>
12760 * doc/md.texi (gather_load@var{m}): Document.
12761 (mask_gather_load@var{m}): Likewise.
12762 * genopinit.c (main): Add supports_vec_gather_load and
12763 supports_vec_gather_load_cached to target_optabs.
12764 * optabs-tree.c (init_tree_optimization_optabs): Use
12765 ggc_cleared_alloc to allocate target_optabs.
12766 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
12767 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
12769 * internal-fn.h (internal_load_fn_p): Declare.
12770 (internal_gather_scatter_fn_p): Likewise.
12771 (internal_fn_mask_index): Likewise.
12772 (internal_gather_scatter_fn_supported_p): Likewise.
12773 * internal-fn.c (gather_load_direct): New macro.
12774 (expand_gather_load_optab_fn): New function.
12775 (direct_gather_load_optab_supported_p): New macro.
12776 (direct_internal_fn_optab): New function.
12777 (internal_load_fn_p): Likewise.
12778 (internal_gather_scatter_fn_p): Likewise.
12779 (internal_fn_mask_index): Likewise.
12780 (internal_gather_scatter_fn_supported_p): Likewise.
12781 * optabs-query.c (supports_at_least_one_mode_p): New function.
12782 (supports_vec_gather_load_p): Likewise.
12783 * optabs-query.h (supports_vec_gather_load_p): Declare.
12784 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
12785 and memory_type field.
12786 (NUM_PATTERNS): Bump to 15.
12787 * tree-vect-data-refs.c: Include internal-fn.h.
12788 (vect_gather_scatter_fn_p): New function.
12789 (vect_describe_gather_scatter_call): Likewise.
12790 (vect_check_gather_scatter): Try using internal functions for
12791 gather loads. Recognize existing calls to a gather load function.
12792 (vect_analyze_data_refs): Consider using gather loads if
12793 supports_vec_gather_load_p.
12794 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
12795 (vect_get_gather_scatter_offset_type): Likewise.
12796 (vect_convert_mask_for_vectype): Likewise.
12797 (vect_add_conversion_to_patterm): Likewise.
12798 (vect_try_gather_scatter_pattern): Likewise.
12799 (vect_recog_gather_scatter_pattern): New pattern recognizer.
12800 (vect_vect_recog_func_ptrs): Add it.
12801 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
12802 internal_fn_mask_index and internal_gather_scatter_fn_p.
12803 (check_load_store_masking): Take the gather_scatter_info as an
12804 argument and handle gather loads.
12805 (vect_get_gather_scatter_ops): New function.
12806 (vectorizable_call): Check internal_load_fn_p.
12807 (vectorizable_load): Likewise. Handle gather load internal
12809 (vectorizable_store): Update call to check_load_store_masking.
12810 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
12811 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
12812 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
12813 (aarch64_gather_scale_operand_d): New predicates.
12814 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
12815 (mask_gather_load<mode>): New insns.
12817 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12818 Alan Hayward <alan.hayward@arm.com>
12819 David Sherwood <david.sherwood@arm.com>
12821 * optabs.def (fold_left_plus_optab): New optab.
12822 * doc/md.texi (fold_left_plus_@var{m}): Document.
12823 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
12824 * internal-fn.c (fold_left_direct): Define.
12825 (expand_fold_left_optab_fn): Likewise.
12826 (direct_fold_left_optab_supported_p): Likewise.
12827 * fold-const-call.c (fold_const_fold_left): New function.
12828 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
12829 * tree-parloops.c (valid_reduction_p): New function.
12830 (gather_scalar_reductions): Use it.
12831 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
12832 (vect_finish_replace_stmt): Declare.
12833 * tree-vect-loop.c (fold_left_reduction_fn): New function.
12834 (needs_fold_left_reduction_p): New function, split out from...
12835 (vect_is_simple_reduction): ...here. Accept reductions that
12836 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
12837 (vect_force_simple_reduction): Also store the reduction type in
12838 the assignment's STMT_VINFO_REDUC_TYPE.
12839 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
12840 (merge_with_identity): New function.
12841 (vect_expand_fold_left): Likewise.
12842 (vectorize_fold_left_reduction): Likewise.
12843 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
12844 scalar phi in place for it. Check for target support and reject
12845 cases that would reassociate the operation. Defer the transform
12846 phase to vectorize_fold_left_reduction.
12847 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
12848 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
12849 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
12851 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12853 * tree-if-conv.c (predicate_mem_writes): Remove redundant
12854 call to ifc_temp_var.
12856 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12857 Alan Hayward <alan.hayward@arm.com>
12858 David Sherwood <david.sherwood@arm.com>
12860 * target.def (legitimize_address_displacement): Take the original
12861 offset as a poly_int.
12862 * targhooks.h (default_legitimize_address_displacement): Update
12864 * targhooks.c (default_legitimize_address_displacement): Likewise.
12865 * doc/tm.texi: Regenerate.
12866 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
12867 as an argument, moving assert of ad->disp == ad->disp_term to...
12868 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
12869 Try calling targetm.legitimize_address_displacement before expanding
12870 the address rather than afterwards, and adjust for the new interface.
12871 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
12872 Match the new hook interface. Handle SVE addresses.
12873 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
12874 new hook interface.
12876 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12878 * Makefile.in (OBJS): Add early-remat.o.
12879 * target.def (select_early_remat_modes): New hook.
12880 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
12881 * doc/tm.texi: Regenerate.
12882 * targhooks.h (default_select_early_remat_modes): Declare.
12883 * targhooks.c (default_select_early_remat_modes): New function.
12884 * timevar.def (TV_EARLY_REMAT): New timevar.
12885 * passes.def (pass_early_remat): New pass.
12886 * tree-pass.h (make_pass_early_remat): Declare.
12887 * early-remat.c: New file.
12888 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
12890 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
12892 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12893 Alan Hayward <alan.hayward@arm.com>
12894 David Sherwood <david.sherwood@arm.com>
12896 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
12897 vfm1 with a bound_epilog parameter.
12898 (vect_do_peeling): Update calls accordingly, and move the prologue
12899 call earlier in the function. Treat the base bound_epilog as 0 for
12900 fully-masked loops and retain vf - 1 for other loops. Add 1 to
12901 this base when peeling for gaps.
12902 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
12903 with fully-masked loops.
12904 (vect_estimate_min_profitable_iters): Handle the single peeled
12905 iteration in that case.
12907 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12908 Alan Hayward <alan.hayward@arm.com>
12909 David Sherwood <david.sherwood@arm.com>
12911 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
12912 single-element interleaving even if the size is not a power of 2.
12913 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
12914 accesses for single-element interleaving if the group size is
12917 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12918 Alan Hayward <alan.hayward@arm.com>
12919 David Sherwood <david.sherwood@arm.com>
12921 * doc/md.texi (fold_extract_last_@var{m}): Document.
12922 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
12923 * optabs.def (fold_extract_last_optab): New optab.
12924 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
12925 * internal-fn.c (fold_extract_direct): New macro.
12926 (expand_fold_extract_optab_fn): Likewise.
12927 (direct_fold_extract_optab_supported_p): Likewise.
12928 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
12929 * tree-vect-loop.c (vect_model_reduction_cost): Handle
12930 EXTRACT_LAST_REDUCTION.
12931 (get_initial_def_for_reduction): Do not create an initial vector
12932 for EXTRACT_LAST_REDUCTION reductions.
12933 (vectorizable_reduction): Leave the scalar phi in place for
12934 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
12935 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
12936 epilogue code for EXTRACT_LAST_REDUCTION and defer the
12937 transform phase to vectorizable_condition.
12938 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
12940 (vect_finish_stmt_generation): ...here.
12941 (vect_finish_replace_stmt): New function.
12942 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
12943 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
12945 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
12947 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12948 Alan Hayward <alan.hayward@arm.com>
12949 David Sherwood <david.sherwood@arm.com>
12951 * doc/md.texi (extract_last_@var{m}): Document.
12952 * optabs.def (extract_last_optab): New optab.
12953 * internal-fn.def (EXTRACT_LAST): New internal function.
12954 * internal-fn.c (cond_unary_direct): New macro.
12955 (expand_cond_unary_optab_fn): Likewise.
12956 (direct_cond_unary_optab_supported_p): Likewise.
12957 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
12958 loops using EXTRACT_LAST.
12959 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
12960 (extract_last_<mode>): ...this optab.
12961 (vec_extract<mode><Vel>): Update accordingly.
12963 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12964 Alan Hayward <alan.hayward@arm.com>
12965 David Sherwood <david.sherwood@arm.com>
12967 * target.def (empty_mask_is_expensive): New hook.
12968 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
12969 * doc/tm.texi: Regenerate.
12970 * targhooks.h (default_empty_mask_is_expensive): Declare.
12971 * targhooks.c (default_empty_mask_is_expensive): New function.
12972 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
12973 if the target says that empty masks are expensive.
12974 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
12976 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
12978 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12979 Alan Hayward <alan.hayward@arm.com>
12980 David Sherwood <david.sherwood@arm.com>
12982 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
12983 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
12984 (vect_use_loop_mask_for_alignment_p): New function.
12985 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
12986 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
12987 niters_skip argument. Make sure that the first niters_skip elements
12988 of the first iteration are inactive.
12989 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
12990 Update call to vect_set_loop_masks_directly.
12991 (get_misalign_in_elems): New function, split out from...
12992 (vect_gen_prolog_loop_niters): ...here.
12993 (vect_update_init_of_dr): Take a code argument that specifies whether
12994 the adjustment should be added or subtracted.
12995 (vect_update_init_of_drs): Likewise.
12996 (vect_prepare_for_masked_peels): New function.
12997 (vect_do_peeling): Skip prologue peeling if we're using a mask
12998 instead. Update call to vect_update_inits_of_drs.
12999 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13001 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
13002 alignment. Do not include the number of peeled iterations in
13003 the minimum threshold in that case.
13004 (vectorizable_induction): Adjust the start value down by
13005 LOOP_VINFO_MASK_SKIP_NITERS iterations.
13006 (vect_transform_loop): Call vect_prepare_for_masked_peels.
13007 Take the number of skipped iterations into account when calculating
13009 * tree-vect-stmts.c (vect_gen_while_not): New function.
13011 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13012 Alan Hayward <alan.hayward@arm.com>
13013 David Sherwood <david.sherwood@arm.com>
13015 * doc/sourcebuild.texi (vect_fully_masked): Document.
13016 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
13017 default value to 0.
13018 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
13020 (vect_analyze_loop_2): ...here. Don't check the vectorization
13021 factor against the number of loop iterations if the loop is
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 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
13029 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
13030 (dump_groups): Update accordingly.
13031 (iv_use::mem_type): New member variable.
13032 (address_p): New function.
13033 (record_use): Add a mem_type argument and initialize the new
13035 (record_group_use): Add a mem_type argument. Use address_p.
13036 Remove obsolete null checks of base_object. Update call to record_use.
13037 (find_interesting_uses_op): Update call to record_group_use.
13038 (find_interesting_uses_cond): Likewise.
13039 (find_interesting_uses_address): Likewise.
13040 (get_mem_type_for_internal_fn): New function.
13041 (find_address_like_use): Likewise.
13042 (find_interesting_uses_stmt): Try find_address_like_use before
13043 calling find_interesting_uses_op.
13044 (addr_offset_valid_p): Use the iv mem_type field as the type
13045 of the addressed memory.
13046 (add_autoinc_candidates): Likewise.
13047 (get_address_cost): Likewise.
13048 (split_small_address_groups_p): Use address_p.
13049 (split_address_groups): Likewise.
13050 (add_iv_candidate_for_use): Likewise.
13051 (autoinc_possible_for_pair): Likewise.
13052 (rewrite_groups): Likewise.
13053 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
13054 (determine_group_iv_cost): Update after split of USE_ADDRESS.
13055 (get_alias_ptr_type_for_ptr_address): New function.
13056 (rewrite_use_address): Rewrite address uses in calls that were
13057 identified by find_address_like_use.
13059 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13060 Alan Hayward <alan.hayward@arm.com>
13061 David Sherwood <david.sherwood@arm.com>
13063 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
13065 * gimple-expr.h (is_gimple_addressable: Likewise.
13066 * gimple-expr.c (is_gimple_address): Likewise.
13067 * internal-fn.c (expand_call_mem_ref): New function.
13068 (expand_mask_load_optab_fn): Use it.
13069 (expand_mask_store_optab_fn): Likewise.
13071 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13072 Alan Hayward <alan.hayward@arm.com>
13073 David Sherwood <david.sherwood@arm.com>
13075 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
13076 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
13077 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
13078 (cond_umax@var{mode}): Document.
13079 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
13080 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
13081 (cond_umin_optab, cond_umax_optab): New optabs.
13082 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
13083 (COND_IOR, COND_XOR): New internal functions.
13084 * internal-fn.h (get_conditional_internal_fn): Declare.
13085 * internal-fn.c (cond_binary_direct): New macro.
13086 (expand_cond_binary_optab_fn): Likewise.
13087 (direct_cond_binary_optab_supported_p): Likewise.
13088 (get_conditional_internal_fn): New function.
13089 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
13090 Cope with reduction statements that are vectorized as calls rather
13092 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
13093 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
13094 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
13095 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
13096 (UNSPEC_COND_EOR): New unspecs.
13097 (optab): Add mappings for them.
13098 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
13099 (sve_int_op, sve_fp_op): New int attributes.
13101 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13102 Alan Hayward <alan.hayward@arm.com>
13103 David Sherwood <david.sherwood@arm.com>
13105 * optabs.def (while_ult_optab): New optab.
13106 * doc/md.texi (while_ult@var{m}@var{n}): Document.
13107 * internal-fn.def (WHILE_ULT): New internal function.
13108 * internal-fn.h (direct_internal_fn_supported_p): New override
13109 that takes two types as argument.
13110 * internal-fn.c (while_direct): New macro.
13111 (expand_while_optab_fn): New function.
13112 (convert_optab_supported_p): Likewise.
13113 (direct_while_optab_supported_p): New macro.
13114 * wide-int.h (wi::udiv_ceil): New function.
13115 * tree-vectorizer.h (rgroup_masks): New structure.
13116 (vec_loop_masks): New typedef.
13117 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
13118 and fully_masked_p.
13119 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
13120 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
13121 (vect_max_vf): New function.
13122 (slpeel_make_loop_iterate_ntimes): Delete.
13123 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
13124 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
13125 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
13126 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
13127 internal-fn.h, stor-layout.h and optabs-query.h.
13128 (vect_set_loop_mask): New function.
13129 (add_preheader_seq): Likewise.
13130 (add_header_seq): Likewise.
13131 (interleave_supported_p): Likewise.
13132 (vect_maybe_permute_loop_masks): Likewise.
13133 (vect_set_loop_masks_directly): Likewise.
13134 (vect_set_loop_condition_masked): Likewise.
13135 (vect_set_loop_condition_unmasked): New function, split out from
13136 slpeel_make_loop_iterate_ntimes.
13137 (slpeel_make_loop_iterate_ntimes): Rename to..
13138 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
13139 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
13140 (vect_do_peeling): Update call accordingly.
13141 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
13143 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13144 mask_compare_type, can_fully_mask_p and fully_masked_p.
13145 (release_vec_loop_masks): New function.
13146 (_loop_vec_info): Use it to free the loop masks.
13147 (can_produce_all_loop_masks_p): New function.
13148 (vect_get_max_nscalars_per_iter): Likewise.
13149 (vect_verify_full_masking): Likewise.
13150 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
13151 retries, and free the mask rgroups before retrying. Check loop-wide
13152 reasons for disallowing fully-masked loops. Make the final decision
13153 about whether use a fully-masked loop or not.
13154 (vect_estimate_min_profitable_iters): Do not assume that peeling
13155 for the number of iterations will be needed for fully-masked loops.
13156 (vectorizable_reduction): Disable fully-masked loops.
13157 (vectorizable_live_operation): Likewise.
13158 (vect_halve_mask_nunits): New function.
13159 (vect_double_mask_nunits): Likewise.
13160 (vect_record_loop_mask): Likewise.
13161 (vect_get_loop_mask): Likewise.
13162 (vect_transform_loop): Handle the case in which the final loop
13163 iteration might handle a partial vector. Call vect_set_loop_condition
13164 instead of slpeel_make_loop_iterate_ntimes.
13165 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
13166 (check_load_store_masking): New function.
13167 (prepare_load_store_mask): Likewise.
13168 (vectorizable_store): Handle fully-masked loops.
13169 (vectorizable_load): Likewise.
13170 (supportable_widening_operation): Use vect_halve_mask_nunits for
13172 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
13173 (vect_gen_while): New function.
13174 * config/aarch64/aarch64.md (umax<mode>3): New expander.
13175 (aarch64_uqdec<mode>): New insn.
13177 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13178 Alan Hayward <alan.hayward@arm.com>
13179 David Sherwood <david.sherwood@arm.com>
13181 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
13182 (reduc_xor_scal_optab): New optabs.
13183 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
13184 (reduc_xor_scal_@var{m}): Document.
13185 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
13186 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
13187 internal functions.
13188 * fold-const-call.c (fold_const_call): Handle them.
13189 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
13190 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
13191 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
13192 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
13193 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
13194 (UNSPEC_XORV): New unspecs.
13195 (optab): Add entries for them.
13196 (BITWISEV): New int iterator.
13197 (bit_reduc_op): New int attributes.
13199 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13200 Alan Hayward <alan.hayward@arm.com>
13201 David Sherwood <david.sherwood@arm.com>
13203 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
13204 * internal-fn.def (VEC_SHL_INSERT): New internal function.
13205 * optabs.def (vec_shl_insert_optab): New optab.
13206 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
13207 (duplicate_and_interleave): Likewise.
13208 * tree-vect-loop.c: Include internal-fn.h.
13209 (neutral_op_for_slp_reduction): New function, split out from
13210 get_initial_defs_for_reduction.
13211 (get_initial_def_for_reduction): Handle option 2 for variable-length
13212 vectors by loading the neutral value into a vector and then shifting
13213 the initial value into element 0.
13214 (get_initial_defs_for_reduction): Replace the code argument with
13215 the neutral value calculated by neutral_op_for_slp_reduction.
13216 Use gimple_build_vector for constant-length vectors.
13217 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
13218 but the first group_size elements have a neutral value.
13219 Use duplicate_and_interleave otherwise.
13220 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
13221 Update call to get_initial_defs_for_reduction. Handle SLP
13222 reductions for variable-length vectors by creating one vector
13223 result for each scalar result, with the elements associated
13224 with other scalar results stubbed out with the neutral value.
13225 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
13226 Require IFN_VEC_SHL_INSERT for double reductions on
13227 variable-length vectors, or SLP reductions that have
13228 a neutral value. Require can_duplicate_and_interleave_p
13229 support for variable-length unchained SLP reductions if there
13230 is no neutral value, such as for MIN/MAX reductions. Also require
13231 the number of vector elements to be a multiple of the number of
13232 SLP statements when doing variable-length unchained SLP reductions.
13233 Update call to vect_create_epilog_for_reduction.
13234 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
13235 and remove initial values.
13236 (duplicate_and_interleave): Make public.
13237 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
13238 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
13240 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13241 Alan Hayward <alan.hayward@arm.com>
13242 David Sherwood <david.sherwood@arm.com>
13244 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
13245 (can_duplicate_and_interleave_p): New function.
13246 (vect_get_and_check_slp_defs): Take the vector of statements
13247 rather than just the current one. Remove excess parentheses.
13248 Restriction rejectinon of vect_constant_def and vect_external_def
13249 for variable-length vectors to boolean types, or types for which
13250 can_duplicate_and_interleave_p is false.
13251 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
13252 (duplicate_and_interleave): New function.
13253 (vect_get_constant_vectors): Use gimple_build_vector for
13254 constant-length vectors and suitable variable-length constant
13255 vectors. Use duplicate_and_interleave for other variable-length
13256 vectors. Don't defer the update when inserting new statements.
13258 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13259 Alan Hayward <alan.hayward@arm.com>
13260 David Sherwood <david.sherwood@arm.com>
13262 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
13263 min_profitable_iters doesn't go negative.
13265 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13266 Alan Hayward <alan.hayward@arm.com>
13267 David Sherwood <david.sherwood@arm.com>
13269 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
13270 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
13271 * optabs.def (vec_mask_load_lanes_optab): New optab.
13272 (vec_mask_store_lanes_optab): Likewise.
13273 * internal-fn.def (MASK_LOAD_LANES): New internal function.
13274 (MASK_STORE_LANES): Likewise.
13275 * internal-fn.c (mask_load_lanes_direct): New macro.
13276 (mask_store_lanes_direct): Likewise.
13277 (expand_mask_load_optab_fn): Handle masked operations.
13278 (expand_mask_load_lanes_optab_fn): New macro.
13279 (expand_mask_store_optab_fn): Handle masked operations.
13280 (expand_mask_store_lanes_optab_fn): New macro.
13281 (direct_mask_load_lanes_optab_supported_p): Likewise.
13282 (direct_mask_store_lanes_optab_supported_p): Likewise.
13283 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
13285 (vect_load_lanes_supported): Likewise.
13286 * tree-vect-data-refs.c (strip_conversion): New function.
13287 (can_group_stmts_p): Likewise.
13288 (vect_analyze_data_ref_accesses): Use it instead of checking
13289 for a pair of assignments.
13290 (vect_store_lanes_supported): Take a masked_p parameter.
13291 (vect_load_lanes_supported): Likewise.
13292 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
13293 vect_store_lanes_supported and vect_load_lanes_supported.
13294 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
13295 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
13296 parameter. Don't allow gaps for masked accesses.
13297 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
13298 and vect_load_lanes_supported.
13299 (get_load_store_type): Take a masked_p parameter and update
13300 call to get_group_load_store_type.
13301 (vectorizable_store): Update call to get_load_store_type.
13302 Handle IFN_MASK_STORE_LANES.
13303 (vectorizable_load): Update call to get_load_store_type.
13304 Handle IFN_MASK_LOAD_LANES.
13306 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13307 Alan Hayward <alan.hayward@arm.com>
13308 David Sherwood <david.sherwood@arm.com>
13310 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
13312 * config/aarch64/aarch64-protos.h
13313 (aarch64_sve_struct_memory_operand_p): Declare.
13314 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
13315 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
13316 (VPRED, vpred): Handle SVE structure modes.
13317 * config/aarch64/constraints.md (Utx): New constraint.
13318 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
13319 (aarch64_sve_struct_nonimmediate_operand): New predicates.
13320 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
13321 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
13322 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
13323 structure modes. Split into pieces after RA.
13324 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
13325 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
13327 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
13328 SVE structure modes.
13329 (aarch64_classify_address): Likewise.
13330 (sizetochar): Move earlier in file.
13331 (aarch64_print_operand): Handle SVE register lists.
13332 (aarch64_array_mode): New function.
13333 (aarch64_sve_struct_memory_operand_p): Likewise.
13334 (TARGET_ARRAY_MODE): Redefine.
13336 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13337 Alan Hayward <alan.hayward@arm.com>
13338 David Sherwood <david.sherwood@arm.com>
13340 * target.def (array_mode): New target hook.
13341 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
13342 * doc/tm.texi: Regenerate.
13343 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
13344 * hooks.c (hook_optmode_mode_uhwi_none): New function.
13345 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
13346 targetm.array_mode.
13347 * stor-layout.c (mode_for_array): Likewise. Support polynomial
13350 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13351 Alan Hayward <alan.hayward@arm.com>
13352 David Sherwood <david.sherwood@arm.com>
13354 * fold-const.c (fold_binary_loc): Check the argument types
13355 rather than the result type when testing for a vector operation.
13357 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13359 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
13360 * doc/tm.texi: Regenerate.
13362 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13363 Alan Hayward <alan.hayward@arm.com>
13364 David Sherwood <david.sherwood@arm.com>
13366 * doc/invoke.texi (-msve-vector-bits=): Document new option.
13367 (sve): Document new AArch64 extension.
13368 * doc/md.texi (w): Extend the description of the AArch64
13369 constraint to include SVE vectors.
13370 (Upl, Upa): Document new AArch64 predicate constraints.
13371 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
13373 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
13374 (msve-vector-bits=): New option.
13375 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
13376 SVE when these are disabled.
13377 (sve): New extension.
13378 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
13379 modes. Adjust their number of units based on aarch64_sve_vg.
13380 (MAX_BITSIZE_MODE_ANY_MODE): Define.
13381 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
13382 aarch64_addr_query_type.
13383 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
13384 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
13385 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
13386 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
13387 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
13388 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
13389 (aarch64_simd_imm_zero_p): Delete.
13390 (aarch64_check_zero_based_sve_index_immediate): Declare.
13391 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
13392 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
13393 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
13394 (aarch64_sve_float_mul_immediate_p): Likewise.
13395 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
13396 rather than an rtx.
13397 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
13398 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
13399 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
13400 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
13401 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
13402 (aarch64_regmode_natural_size): Likewise.
13403 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
13404 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
13406 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
13407 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
13408 for VG and the SVE predicate registers.
13409 (V_ALIASES): Add a "z"-prefixed alias.
13410 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
13411 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
13412 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
13413 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
13414 (REG_CLASS_NAMES): Add entries for them.
13415 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
13416 and the predicate registers.
13417 (aarch64_sve_vg): Declare.
13418 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
13419 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
13420 (REGMODE_NATURAL_SIZE): Define.
13421 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
13423 * config/aarch64/aarch64.c: Include cfgrtl.h.
13424 (simd_immediate_info): Add a constructor for series vectors,
13425 and an associated step field.
13426 (aarch64_sve_vg): New variable.
13427 (aarch64_dbx_register_number): Handle VG and the predicate registers.
13428 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
13429 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
13430 (VEC_ANY_DATA, VEC_STRUCT): New constants.
13431 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
13432 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
13433 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
13434 (aarch64_get_mask_mode): New functions.
13435 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
13436 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
13437 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
13438 predicate modes and predicate registers. Explicitly restrict
13439 GPRs to modes of 16 bytes or smaller. Only allow FP registers
13440 to store a vector mode if it is recognized by
13441 aarch64_classify_vector_mode.
13442 (aarch64_regmode_natural_size): New function.
13443 (aarch64_hard_regno_caller_save_mode): Return the original mode
13445 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
13446 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
13447 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
13448 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
13450 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
13451 does not overlap dest if the function is frame-related. Handle
13453 (aarch64_split_add_offset): New function.
13454 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
13455 them aarch64_add_offset.
13456 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
13457 and update call to aarch64_sub_sp.
13458 (aarch64_add_cfa_expression): New function.
13459 (aarch64_expand_prologue): Pass extra temporary registers to the
13460 functions above. Handle the case in which we need to emit new
13461 DW_CFA_expressions for registers that were originally saved
13462 relative to the stack pointer, but now have to be expressed
13463 relative to the frame pointer.
13464 (aarch64_output_mi_thunk): Pass extra temporary registers to the
13466 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
13467 IP0 and IP1 values for SVE frames.
13468 (aarch64_expand_vec_series): New function.
13469 (aarch64_expand_sve_widened_duplicate): Likewise.
13470 (aarch64_expand_sve_const_vector): Likewise.
13471 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
13472 Handle SVE constants. Use emit_move_insn to move a force_const_mem
13473 into the register, rather than emitting a SET directly.
13474 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
13475 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
13476 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
13477 (offset_9bit_signed_scaled_p): New functions.
13478 (aarch64_replicate_bitmask_imm): New function.
13479 (aarch64_bitmask_imm): Use it.
13480 (aarch64_cannot_force_const_mem): Reject expressions involving
13481 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
13482 (aarch64_classify_index): Handle SVE indices, by requiring
13483 a plain register index with a scale that matches the element size.
13484 (aarch64_classify_address): Handle SVE addresses. Assert that
13485 the mode of the address is VOIDmode or an integer mode.
13486 Update call to aarch64_classify_symbol.
13487 (aarch64_classify_symbolic_expression): Update call to
13488 aarch64_classify_symbol.
13489 (aarch64_const_vec_all_in_range_p): New function.
13490 (aarch64_print_vector_float_operand): Likewise.
13491 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
13492 "vN" for FP registers with SVE modes. Handle (const ...) vectors
13493 and the FP immediates 1.0 and 0.5.
13494 (aarch64_print_address_internal): Handle SVE addresses.
13495 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
13496 (aarch64_regno_regclass): Handle predicate registers.
13497 (aarch64_secondary_reload): Handle big-endian reloads of SVE
13499 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
13500 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
13501 (aarch64_convert_sve_vector_bits): New function.
13502 (aarch64_override_options): Use it to handle -msve-vector-bits=.
13503 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
13504 rather than an rtx.
13505 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
13506 Handle SVE vector and predicate modes. Accept VL-based constants
13507 that need only one temporary register, and VL offsets that require
13508 no temporary registers.
13509 (aarch64_conditional_register_usage): Mark the predicate registers
13510 as fixed if SVE isn't available.
13511 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
13512 Return true for SVE vector and predicate modes.
13513 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
13514 rather than an unsigned int. Handle SVE modes.
13515 (aarch64_preferred_simd_mode): Update call accordingly. Handle
13517 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
13519 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
13520 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
13521 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
13522 (aarch64_sve_float_mul_immediate_p): New functions.
13523 (aarch64_sve_valid_immediate): New function.
13524 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
13525 Explicitly reject structure modes. Check for INDEX constants.
13526 Handle PTRUE and PFALSE constants.
13527 (aarch64_check_zero_based_sve_index_immediate): New function.
13528 (aarch64_simd_imm_zero_p): Delete.
13529 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
13530 vector modes. Accept constants in the range of CNT[BHWD].
13531 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
13532 ask for an Advanced SIMD mode.
13533 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
13534 (aarch64_simd_vector_alignment): Handle SVE predicates.
13535 (aarch64_vectorize_preferred_vector_alignment): New function.
13536 (aarch64_simd_vector_alignment_reachable): Use it instead of
13538 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
13539 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
13541 (MAX_VECT_LEN): Delete.
13542 (expand_vec_perm_d): Add a vec_flags field.
13543 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
13544 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
13545 (aarch64_evpc_ext): Don't apply a big-endian lane correction
13547 (aarch64_evpc_rev): Rename to...
13548 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
13549 (aarch64_evpc_rev_global): New function.
13550 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
13551 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
13553 (aarch64_evpc_sve_tbl): New function.
13554 (aarch64_expand_vec_perm_const_1): Update after rename of
13555 aarch64_evpc_rev. Handle SVE permutes too, trying
13556 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
13557 than aarch64_evpc_tbl.
13558 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
13559 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
13560 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
13561 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
13562 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
13563 (aarch64_expand_sve_vcond): New functions.
13564 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
13565 of aarch64_vector_mode_p.
13566 (aarch64_dwarf_poly_indeterminate_value): New function.
13567 (aarch64_compute_pressure_classes): Likewise.
13568 (aarch64_can_change_mode_class): Likewise.
13569 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
13570 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
13571 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
13572 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
13573 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
13574 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
13575 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
13576 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
13578 (Dn, Dl, Dr): Accept const as well as const_vector.
13579 (Dz): Likewise. Compare against CONST0_RTX.
13580 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
13581 of "vector" where appropriate.
13582 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
13583 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
13584 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
13585 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
13586 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
13587 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
13588 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
13589 (v_int_equiv): Extend to SVE modes.
13590 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
13592 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
13593 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
13594 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
13595 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
13596 (SVE_COND_FP_CMP): New int iterators.
13597 (perm_hilo): Handle the new unpack unspecs.
13598 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
13600 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
13601 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
13602 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
13603 (aarch64_equality_operator, aarch64_constant_vector_operand)
13604 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
13605 (aarch64_sve_nonimmediate_operand): Likewise.
13606 (aarch64_sve_general_operand): Likewise.
13607 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
13608 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
13609 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
13610 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
13611 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
13612 (aarch64_sve_float_arith_immediate): Likewise.
13613 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
13614 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
13615 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
13616 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
13617 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
13618 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
13619 (aarch64_sve_float_arith_operand): Likewise.
13620 (aarch64_sve_float_arith_with_sub_operand): Likewise.
13621 (aarch64_sve_float_mul_operand): Likewise.
13622 (aarch64_sve_vec_perm_operand): Likewise.
13623 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
13624 (aarch64_mov_operand): Accept const_poly_int and const_vector.
13625 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
13626 as well as const_vector.
13627 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
13628 in file. Use CONST0_RTX and CONSTM1_RTX.
13629 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
13630 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
13631 Use aarch64_simd_imm_zero.
13632 * config/aarch64/aarch64-sve.md: New file.
13633 * config/aarch64/aarch64.md: Include it.
13634 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
13635 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
13636 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
13637 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
13638 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
13639 (sve): New attribute.
13640 (enabled): Disable instructions with the sve attribute unless
13642 (movqi, movhi): Pass CONST_POLY_INT operaneds through
13643 aarch64_expand_mov_immediate.
13644 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
13645 CNT[BHSD] immediates.
13646 (movti): Split CONST_POLY_INT moves into two halves.
13647 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
13648 Split additions that need a temporary here if the destination
13649 is the stack pointer.
13650 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
13651 (*add<mode>3_poly_1): New instruction.
13652 (set_clobber_cc): New expander.
13654 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13656 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
13657 parameter and use it instead of GET_MODE_SIZE (innermode). Use
13658 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
13659 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
13660 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
13661 Change innermode from fixed_mode_size to machine_mode.
13662 (simplify_subreg): Update call accordingly. Handle a constant-sized
13663 subreg of a variable-length CONST_VECTOR.
13665 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13666 Alan Hayward <alan.hayward@arm.com>
13667 David Sherwood <david.sherwood@arm.com>
13669 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
13670 (add_offset_to_base): New function, split out from...
13671 (create_mem_ref): ...here. When handling a scale other than 1,
13672 check first whether the address is valid without the offset.
13673 Add it into the base if so, leaving the index and scale as-is.
13675 2018-01-12 Jakub Jelinek <jakub@redhat.com>
13678 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
13679 fold_for_warn before checking if arg2 is INTEGER_CST.
13681 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
13683 * config/rs6000/predicates.md (load_multiple_operation): Delete.
13684 (store_multiple_operation): Delete.
13685 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
13686 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
13687 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
13688 guarded by TARGET_STRING.
13689 (rs6000_output_load_multiple): Delete.
13690 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
13691 OPTION_MASK_STRING / TARGET_STRING handling.
13692 (print_operand) <'N', 'O'>: Add comment that these are unused now.
13693 (const rs6000_opt_masks) <"string">: Change mask to 0.
13694 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
13695 (MASK_STRING): Delete.
13696 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
13698 (load_multiple): Delete.
13705 (store_multiple): Delete.
13712 (movmemsi_8reg): Delete.
13713 (corresponding unnamed define_insn): Delete.
13714 (movmemsi_6reg): Delete.
13715 (corresponding unnamed define_insn): Delete.
13716 (movmemsi_4reg): Delete.
13717 (corresponding unnamed define_insn): Delete.
13718 (movmemsi_2reg): Delete.
13719 (corresponding unnamed define_insn): Delete.
13720 (movmemsi_1reg): Delete.
13721 (corresponding unnamed define_insn): Delete.
13722 * config/rs6000/rs6000.opt (mno-string): New.
13723 (mstring): Replace by deprecation warning stub.
13724 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
13726 2018-01-12 Jakub Jelinek <jakub@redhat.com>
13728 * regrename.c (regrename_do_replace): If replacing the same
13729 reg multiple times, try to reuse last created gen_raw_REG.
13732 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
13733 main to workaround a bug in GDB.
13735 2018-01-12 Tom de Vries <tom@codesourcery.com>
13738 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
13740 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
13742 PR rtl-optimization/80481
13743 * ira-color.c (get_cap_member): New function.
13744 (allocnos_conflict_by_live_ranges_p): Use it.
13745 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
13746 (setup_slot_coalesced_allocno_live_ranges): Ditto.
13748 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
13751 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
13752 (*saddl_se_1): Ditto.
13753 (*ssubsi_1): Ditto.
13754 (*ssubl_se_1): Ditto.
13756 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13758 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
13759 rather than wi::to_widest for DR_INITs.
13760 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
13761 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
13762 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
13764 (vect_analyze_group_access_1): Note that here.
13766 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13768 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
13769 polynomial type sizes.
13771 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13773 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
13774 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
13775 (gimple_add_tmp_var): Likewise.
13777 2018-01-12 Martin Liska <mliska@suse.cz>
13779 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
13780 (gimple_alloc_sizes): Likewise.
13781 (dump_gimple_statistics): Use PRIu64 in printf format.
13782 * gimple.h: Change uint64_t to int.
13784 2018-01-12 Martin Liska <mliska@suse.cz>
13786 * tree-core.h: Use uint64_t instead of int.
13787 * tree.c (tree_node_counts): Likewise.
13788 (tree_node_sizes): Likewise.
13789 (dump_tree_statistics): Use PRIu64 in printf format.
13791 2018-01-12 Martin Liska <mliska@suse.cz>
13793 * Makefile.in: As qsort_chk is implemented in vec.c, add
13794 vec.o to linkage of gencfn-macros.
13795 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
13796 passing the info to record_node_allocation_statistics.
13797 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
13799 * ggc-common.c (struct ggc_usage): Add operator== and use
13800 it in operator< and compare function.
13801 * mem-stats.h (struct mem_usage): Likewise.
13802 * vec.c (struct vec_usage): Remove operator< and compare
13803 function. Can be simply inherited.
13805 2018-01-12 Martin Jambor <mjambor@suse.cz>
13808 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
13809 * tree-ssa-math-opts.c: Include domwalk.h.
13810 (convert_mult_to_fma_1): New function.
13811 (fma_transformation_info): New type.
13812 (fma_deferring_state): Likewise.
13813 (cancel_fma_deferring): New function.
13814 (result_of_phi): Likewise.
13815 (last_fma_candidate_feeds_initial_phi): Likewise.
13816 (convert_mult_to_fma): Added deferring logic, split actual
13817 transformation to convert_mult_to_fma_1.
13818 (math_opts_dom_walker): New type.
13819 (math_opts_dom_walker::after_dom_children): New method, body moved
13820 here from pass_optimize_widening_mul::execute, added deferring logic
13822 (pass_optimize_widening_mul::execute): Moved most of code to
13823 math_opts_dom_walker::after_dom_children.
13824 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
13825 * config/i386/i386.c (ix86_option_override_internal): Added
13826 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
13828 2018-01-12 Richard Biener <rguenther@suse.de>
13831 * dwarf2out.c (gen_variable_die): Do not reset old_die for
13832 inline instance vars.
13834 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
13837 * config/rx/rx.c (rx_is_restricted_memory_address):
13838 Handle SUBREG case.
13840 2018-01-12 Richard Biener <rguenther@suse.de>
13842 PR tree-optimization/80846
13843 * target.def (split_reduction): New target hook.
13844 * targhooks.c (default_split_reduction): New function.
13845 * targhooks.h (default_split_reduction): Declare.
13846 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
13847 target requests first reduce vectors by combining low and high
13849 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
13850 (get_vectype_for_scalar_type_and_size): Export.
13851 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
13852 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
13853 * doc/tm.texi: Regenerate.
13854 * config/i386/i386.c (ix86_split_reduction): Implement
13855 TARGET_VECTORIZE_SPLIT_REDUCTION.
13857 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
13860 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
13861 in PIC mode except for TARGET_VXWORKS_RTP.
13862 * config/sparc/sparc.c: Include cfgrtl.h.
13863 (TARGET_INIT_PIC_REG): Define.
13864 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
13865 (sparc_pic_register_p): New predicate.
13866 (sparc_legitimate_address_p): Use it.
13867 (sparc_legitimize_pic_address): Likewise.
13868 (sparc_delegitimize_address): Likewise.
13869 (sparc_mode_dependent_address_p): Likewise.
13870 (gen_load_pcrel_sym): Remove 4th parameter.
13871 (load_got_register): Adjust call to above. Remove obsolete stuff.
13872 (sparc_expand_prologue): Do not call load_got_register here.
13873 (sparc_flat_expand_prologue): Likewise.
13874 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
13875 (sparc_use_pseudo_pic_reg): New function.
13876 (sparc_init_pic_reg): Likewise.
13877 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
13878 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
13880 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
13882 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
13883 Add item for branch_cost.
13885 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
13887 PR rtl-optimization/83565
13888 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
13889 not extend the result to a larger mode for rotate operations.
13890 (num_sign_bit_copies1): Likewise.
13892 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
13895 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
13897 Use values-Xc.o for -pedantic.
13898 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
13900 2018-01-12 Martin Liska <mliska@suse.cz>
13903 * ipa-devirt.c (final_warning_record::grow_type_warnings):
13905 (possible_polymorphic_call_targets): Use it.
13906 (ipa_devirt): Likewise.
13908 2018-01-12 Martin Liska <mliska@suse.cz>
13910 * profile-count.h (enum profile_quality): Use 0 as invalid
13911 enum value of profile_quality.
13913 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
13915 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
13916 -mext-string options.
13918 2018-01-12 Richard Biener <rguenther@suse.de>
13920 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
13921 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
13922 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
13924 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
13926 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
13928 * configure.ac (--with-long-double-format): Add support for the
13929 configuration option to change the default long double format on
13931 * config.gcc (powerpc*-linux*-*): Likewise.
13932 * configure: Regenerate.
13933 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
13934 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
13935 used without modification.
13937 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
13939 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
13940 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
13941 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
13942 MISC_BUILTIN_SPEC_BARRIER.
13943 (rs6000_init_builtins): Likewise.
13944 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
13946 (speculation_barrier): New define_insn.
13947 * doc/extend.texi: Document __builtin_speculation_barrier.
13949 2018-01-11 Jakub Jelinek <jakub@redhat.com>
13952 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
13953 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
13954 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
13956 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
13957 integral modes instead of "ss" and "sd".
13958 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
13959 vectors with 32-bit and 64-bit elements.
13960 (vecdupssescalarmodesuffix): New mode attribute.
13961 (vec_dup<mode>): Use it.
13963 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
13966 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
13967 frame if argument is passed on stack.
13969 2018-01-11 Jakub Jelinek <jakub@redhat.com>
13972 * ree.c (combine_reaching_defs): Optimize also
13973 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
13974 reg2=any_extend(exp); reg1=reg2;, formatting fix.
13976 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
13978 PR middle-end/83189
13979 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
13981 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
13983 PR middle-end/83718
13984 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
13985 after they are computed.
13987 2018-01-11 Bin Cheng <bin.cheng@arm.com>
13989 PR tree-optimization/83695
13990 * gimple-loop-linterchange.cc
13991 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
13992 reset cached scev information after interchange.
13993 (pass_linterchange::execute): Remove call to scev_reset_htab.
13995 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
13997 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
13998 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
13999 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
14000 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
14001 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
14002 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
14003 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
14004 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
14005 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
14006 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
14007 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
14008 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
14009 (V_lane_reg): Likewise.
14010 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
14012 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
14013 (vfmal_lane_low<mode>_intrinsic,
14014 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
14015 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
14016 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
14017 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
14018 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
14019 vfmsl_lane_high<mode>_intrinsic): New define_insns.
14021 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14023 * config/arm/arm-cpus.in (fp16fml): New feature.
14024 (ALL_SIMD): Add fp16fml.
14025 (armv8.2-a): Add fp16fml as an option.
14026 (armv8.3-a): Likewise.
14027 (armv8.4-a): Add fp16fml as part of fp16.
14028 * config/arm/arm.h (TARGET_FP16FML): Define.
14029 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
14031 * config/arm/arm-modes.def (V2HF): Define.
14032 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
14033 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
14034 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
14035 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
14036 vfmsl_low, vfmsl_high): New set of builtins.
14037 * config/arm/iterators.md (PLUSMINUS): New code iterator.
14038 (vfml_op): New code attribute.
14039 (VFMLHALVES): New int iterator.
14040 (VFML, VFMLSEL): New mode attributes.
14041 (V_reg): Define mapping for V2HF.
14042 (V_hi, V_lo): New mode attributes.
14043 (VF_constraint): Likewise.
14044 (vfml_half, vfml_half_selector): New int attributes.
14045 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
14047 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
14048 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
14050 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
14051 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
14052 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
14053 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
14055 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
14056 Document new effective target and option set.
14058 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14060 * config/arm/arm-cpus.in (armv8_4): New feature.
14061 (ARMv8_4a): New fgroup.
14062 (armv8.4-a): New arch.
14063 * config/arm/arm-tables.opt: Regenerate.
14064 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
14065 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
14066 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
14067 Add matching rules for -march=armv8.4-a and extensions.
14068 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
14070 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
14073 * config/rx/rx.md (BW): New mode attribute.
14074 (sync_lock_test_and_setsi): Add mode suffix to insn output.
14076 2018-01-11 Richard Biener <rguenther@suse.de>
14078 PR tree-optimization/83435
14079 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
14080 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
14081 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
14083 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14084 Alan Hayward <alan.hayward@arm.com>
14085 David Sherwood <david.sherwood@arm.com>
14087 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
14089 (aarch64_classify_address): Initialize it. Track polynomial offsets.
14090 (aarch64_print_address_internal): Use it to check for a zero offset.
14092 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14093 Alan Hayward <alan.hayward@arm.com>
14094 David Sherwood <david.sherwood@arm.com>
14096 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
14097 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
14098 Return a poly_int64 rather than a HOST_WIDE_INT.
14099 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
14100 rather than a HOST_WIDE_INT.
14101 * config/aarch64/aarch64.h (aarch64_frame): Protect with
14102 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
14103 hard_fp_offset, frame_size, initial_adjust, callee_offset and
14104 final_offset from HOST_WIDE_INT to poly_int64.
14105 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
14106 to_constant when getting the number of units in an Advanced SIMD
14108 (aarch64_builtin_vectorized_function): Check for a constant number
14110 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
14112 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
14113 attribute instead of GET_MODE_NUNITS.
14114 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
14115 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
14116 GET_MODE_SIZE for fixed-size registers.
14117 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
14118 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
14119 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
14120 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
14121 (aarch64_print_operand, aarch64_print_address_internal)
14122 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
14123 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
14124 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
14125 Handle polynomial GET_MODE_SIZE.
14126 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
14127 wider than SImode without modification.
14128 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
14129 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
14130 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
14131 passing and returning SVE modes.
14132 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
14133 rather than GEN_INT.
14134 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
14135 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
14136 (aarch64_allocate_and_probe_stack_space): Likewise.
14137 (aarch64_layout_frame): Cope with polynomial offsets.
14138 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
14139 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
14140 polynomial offsets.
14141 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
14142 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
14143 poly_int64 rather than a HOST_WIDE_INT.
14144 (aarch64_get_separate_components, aarch64_process_components)
14145 (aarch64_expand_prologue, aarch64_expand_epilogue)
14146 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
14147 (aarch64_anchor_offset): New function, split out from...
14148 (aarch64_legitimize_address): ...here.
14149 (aarch64_builtin_vectorization_cost): Handle polynomial
14150 TYPE_VECTOR_SUBPARTS.
14151 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
14153 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
14154 number of elements from the PARALLEL rather than the mode.
14155 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
14156 rather than GET_MODE_BITSIZE.
14157 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
14158 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
14159 (aarch64_expand_vec_perm_const_1): Handle polynomial
14160 d->perm.length () and d->perm elements.
14161 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
14162 Apply to_constant to d->perm elements.
14163 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
14164 polynomial CONST_VECTOR_NUNITS.
14165 (aarch64_move_pointer): Take amount as a poly_int64 rather
14167 (aarch64_progress_pointer): Avoid temporary variable.
14168 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
14169 the mode attribute instead of GET_MODE.
14171 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14172 Alan Hayward <alan.hayward@arm.com>
14173 David Sherwood <david.sherwood@arm.com>
14175 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
14176 x exists before using it.
14177 (aarch64_add_constant_internal): Rename to...
14178 (aarch64_add_offset_1): ...this. Replace regnum with separate
14179 src and dest rtxes. Handle the case in which they're different,
14180 including when the offset is zero. Replace scratchreg with an rtx.
14181 Use 2 additions if there is no spare register into which we can
14182 move a 16-bit constant.
14183 (aarch64_add_constant): Delete.
14184 (aarch64_add_offset): Replace reg with separate src and dest
14185 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
14186 Use aarch64_add_offset_1.
14187 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
14188 an rtx rather than an int. Take the delta as a poly_int64
14189 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
14190 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
14191 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
14192 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
14193 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
14194 and aarch64_add_sp.
14195 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
14196 aarch64_add_constant.
14198 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14200 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
14201 Use scalar_float_mode.
14203 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14205 * config/aarch64/aarch64-simd.md
14206 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
14207 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
14208 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
14209 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
14210 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
14211 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
14212 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
14213 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
14214 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
14215 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
14217 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14220 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
14221 targ_options->x_arm_arch_string is non NULL.
14223 2018-01-11 Tamar Christina <tamar.christina@arm.com>
14225 * config/aarch64/aarch64.h
14226 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
14228 2018-01-11 Sudakshina Das <sudi.das@arm.com>
14231 * expmed.c (emit_store_flag_force): Swap if const op0
14232 and change VOIDmode to mode of op0.
14234 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14236 PR rtl-optimization/83761
14237 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
14238 than bytes to mode_for_size.
14240 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14242 PR middle-end/83189
14243 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
14244 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
14247 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14249 PR middle-end/83575
14250 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
14251 when in layout mode.
14252 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
14253 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
14256 2018-01-10 Michael Collison <michael.collison@arm.com>
14258 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
14259 * config/aarch64/aarch64-option-extension.def: Add
14260 AARCH64_OPT_EXTENSION of 'fp16fml'.
14261 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14262 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
14263 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
14264 * config/aarch64/constraints.md (Ui7): New constraint.
14265 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
14266 (VFMLA_SEL_W): Ditto.
14269 (VFMLA16_LOW): New int iterator.
14270 (VFMLA16_HIGH): Ditto.
14271 (UNSPEC_FMLAL): New unspec.
14272 (UNSPEC_FMLSL): Ditto.
14273 (UNSPEC_FMLAL2): Ditto.
14274 (UNSPEC_FMLSL2): Ditto.
14275 (f16mac): New code attribute.
14276 * config/aarch64/aarch64-simd-builtins.def
14277 (aarch64_fmlal_lowv2sf): Ditto.
14278 (aarch64_fmlsl_lowv2sf): Ditto.
14279 (aarch64_fmlalq_lowv4sf): Ditto.
14280 (aarch64_fmlslq_lowv4sf): Ditto.
14281 (aarch64_fmlal_highv2sf): Ditto.
14282 (aarch64_fmlsl_highv2sf): Ditto.
14283 (aarch64_fmlalq_highv4sf): Ditto.
14284 (aarch64_fmlslq_highv4sf): Ditto.
14285 (aarch64_fmlal_lane_lowv2sf): Ditto.
14286 (aarch64_fmlsl_lane_lowv2sf): Ditto.
14287 (aarch64_fmlal_laneq_lowv2sf): Ditto.
14288 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
14289 (aarch64_fmlalq_lane_lowv4sf): Ditto.
14290 (aarch64_fmlsl_lane_lowv4sf): Ditto.
14291 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
14292 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
14293 (aarch64_fmlal_lane_highv2sf): Ditto.
14294 (aarch64_fmlsl_lane_highv2sf): Ditto.
14295 (aarch64_fmlal_laneq_highv2sf): Ditto.
14296 (aarch64_fmlsl_laneq_highv2sf): Ditto.
14297 (aarch64_fmlalq_lane_highv4sf): Ditto.
14298 (aarch64_fmlsl_lane_highv4sf): Ditto.
14299 (aarch64_fmlalq_laneq_highv4sf): Ditto.
14300 (aarch64_fmlsl_laneq_highv4sf): Ditto.
14301 * config/aarch64/aarch64-simd.md:
14302 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
14303 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
14304 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
14305 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
14306 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
14307 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
14308 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
14309 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
14310 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
14311 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
14312 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
14313 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
14314 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
14315 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
14316 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
14317 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
14318 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
14319 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
14320 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
14321 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
14322 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
14323 (vfmlsl_low_u32): Ditto.
14324 (vfmlalq_low_u32): Ditto.
14325 (vfmlslq_low_u32): Ditto.
14326 (vfmlal_high_u32): Ditto.
14327 (vfmlsl_high_u32): Ditto.
14328 (vfmlalq_high_u32): Ditto.
14329 (vfmlslq_high_u32): Ditto.
14330 (vfmlal_lane_low_u32): Ditto.
14331 (vfmlsl_lane_low_u32): Ditto.
14332 (vfmlal_laneq_low_u32): Ditto.
14333 (vfmlsl_laneq_low_u32): Ditto.
14334 (vfmlalq_lane_low_u32): Ditto.
14335 (vfmlslq_lane_low_u32): Ditto.
14336 (vfmlalq_laneq_low_u32): Ditto.
14337 (vfmlslq_laneq_low_u32): Ditto.
14338 (vfmlal_lane_high_u32): Ditto.
14339 (vfmlsl_lane_high_u32): Ditto.
14340 (vfmlal_laneq_high_u32): Ditto.
14341 (vfmlsl_laneq_high_u32): Ditto.
14342 (vfmlalq_lane_high_u32): Ditto.
14343 (vfmlslq_lane_high_u32): Ditto.
14344 (vfmlalq_laneq_high_u32): Ditto.
14345 (vfmlslq_laneq_high_u32): Ditto.
14346 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
14347 (AARCH64_FL_FOR_ARCH8_4): New.
14348 (AARCH64_ISA_F16FML): New ISA flag.
14349 (TARGET_F16FML): New feature flag for fp16fml.
14350 (doc/invoke.texi): Document new fp16fml option.
14352 2018-01-10 Michael Collison <michael.collison@arm.com>
14354 * config/aarch64/aarch64-builtins.c:
14355 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
14356 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14357 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
14358 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
14359 (AARCH64_ISA_SHA3): New ISA flag.
14360 (TARGET_SHA3): New feature flag for sha3.
14361 * config/aarch64/iterators.md (sha512_op): New int attribute.
14362 (CRYPTO_SHA512): New int iterator.
14363 (UNSPEC_SHA512H): New unspec.
14364 (UNSPEC_SHA512H2): Ditto.
14365 (UNSPEC_SHA512SU0): Ditto.
14366 (UNSPEC_SHA512SU1): Ditto.
14367 * config/aarch64/aarch64-simd-builtins.def
14368 (aarch64_crypto_sha512hqv2di): New builtin.
14369 (aarch64_crypto_sha512h2qv2di): Ditto.
14370 (aarch64_crypto_sha512su0qv2di): Ditto.
14371 (aarch64_crypto_sha512su1qv2di): Ditto.
14372 (aarch64_eor3qv8hi): Ditto.
14373 (aarch64_rax1qv2di): Ditto.
14374 (aarch64_xarqv2di): Ditto.
14375 (aarch64_bcaxqv8hi): Ditto.
14376 * config/aarch64/aarch64-simd.md:
14377 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
14378 (aarch64_crypto_sha512su0qv2di): Ditto.
14379 (aarch64_crypto_sha512su1qv2di): Ditto.
14380 (aarch64_eor3qv8hi): Ditto.
14381 (aarch64_rax1qv2di): Ditto.
14382 (aarch64_xarqv2di): Ditto.
14383 (aarch64_bcaxqv8hi): Ditto.
14384 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
14385 (vsha512h2q_u64): Ditto.
14386 (vsha512su0q_u64): Ditto.
14387 (vsha512su1q_u64): Ditto.
14388 (veor3q_u16): Ditto.
14389 (vrax1q_u64): Ditto.
14390 (vxarq_u64): Ditto.
14391 (vbcaxq_u16): Ditto.
14392 * config/arm/types.md (crypto_sha512): New type attribute.
14393 (crypto_sha3): Ditto.
14394 (doc/invoke.texi): Document new sha3 option.
14396 2018-01-10 Michael Collison <michael.collison@arm.com>
14398 * config/aarch64/aarch64-builtins.c:
14399 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
14400 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14401 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
14402 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
14403 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
14404 (AARCH64_ISA_SM4): New ISA flag.
14405 (TARGET_SM4): New feature flag for sm4.
14406 * config/aarch64/aarch64-simd-builtins.def
14407 (aarch64_sm3ss1qv4si): Ditto.
14408 (aarch64_sm3tt1aq4si): Ditto.
14409 (aarch64_sm3tt1bq4si): Ditto.
14410 (aarch64_sm3tt2aq4si): Ditto.
14411 (aarch64_sm3tt2bq4si): Ditto.
14412 (aarch64_sm3partw1qv4si): Ditto.
14413 (aarch64_sm3partw2qv4si): Ditto.
14414 (aarch64_sm4eqv4si): Ditto.
14415 (aarch64_sm4ekeyqv4si): Ditto.
14416 * config/aarch64/aarch64-simd.md:
14417 (aarch64_sm3ss1qv4si): Ditto.
14418 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
14419 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
14420 (aarch64_sm4eqv4si): Ditto.
14421 (aarch64_sm4ekeyqv4si): Ditto.
14422 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
14423 (sm3part_op): Ditto.
14424 (CRYPTO_SM3TT): Ditto.
14425 (CRYPTO_SM3PART): Ditto.
14426 (UNSPEC_SM3SS1): New unspec.
14427 (UNSPEC_SM3TT1A): Ditto.
14428 (UNSPEC_SM3TT1B): Ditto.
14429 (UNSPEC_SM3TT2A): Ditto.
14430 (UNSPEC_SM3TT2B): Ditto.
14431 (UNSPEC_SM3PARTW1): Ditto.
14432 (UNSPEC_SM3PARTW2): Ditto.
14433 (UNSPEC_SM4E): Ditto.
14434 (UNSPEC_SM4EKEY): Ditto.
14435 * config/aarch64/constraints.md (Ui2): New constraint.
14436 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
14437 * config/arm/types.md (crypto_sm3): New type attribute.
14438 (crypto_sm4): Ditto.
14439 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
14440 (vsm3tt1aq_u32): Ditto.
14441 (vsm3tt1bq_u32): Ditto.
14442 (vsm3tt2aq_u32): Ditto.
14443 (vsm3tt2bq_u32): Ditto.
14444 (vsm3partw1q_u32): Ditto.
14445 (vsm3partw2q_u32): Ditto.
14446 (vsm4eq_u32): Ditto.
14447 (vsm4ekeyq_u32): Ditto.
14448 (doc/invoke.texi): Document new sm4 option.
14450 2018-01-10 Michael Collison <michael.collison@arm.com>
14452 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
14453 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
14454 (AARCH64_FL_FOR_ARCH8_4): New.
14455 (AARCH64_FL_V8_4): New flag.
14456 (doc/invoke.texi): Document new armv8.4-a option.
14458 2018-01-10 Michael Collison <michael.collison@arm.com>
14460 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14461 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
14462 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
14463 * config/aarch64/aarch64-option-extension.def: Add
14464 AARCH64_OPT_EXTENSION of 'sha2'.
14465 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
14466 (crypto): Disable sha2 and aes if crypto disabled.
14467 (crypto): Enable aes and sha2 if enabled.
14468 (simd): Disable sha2 and aes if simd disabled.
14469 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
14471 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
14472 (TARGET_SHA2): New feature flag for sha2.
14473 (TARGET_AES): New feature flag for aes.
14474 * config/aarch64/aarch64-simd.md:
14475 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
14476 conditional on TARGET_AES.
14477 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
14478 (aarch64_crypto_sha1hsi): Make pattern conditional
14480 (aarch64_crypto_sha1hv4si): Ditto.
14481 (aarch64_be_crypto_sha1hv4si): Ditto.
14482 (aarch64_crypto_sha1su1v4si): Ditto.
14483 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
14484 (aarch64_crypto_sha1su0v4si): Ditto.
14485 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
14486 (aarch64_crypto_sha256su0v4si): Ditto.
14487 (aarch64_crypto_sha256su1v4si): Ditto.
14488 (doc/invoke.texi): Document new aes and sha2 options.
14490 2018-01-10 Martin Sebor <msebor@redhat.com>
14492 PR tree-optimization/83781
14493 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
14496 2018-01-11 Martin Sebor <msebor@gmail.com>
14497 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14499 PR tree-optimization/83501
14500 PR tree-optimization/81703
14502 * tree-ssa-strlen.c (get_string_cst): Rename...
14503 (get_string_len): ...to this. Handle global constants.
14504 (handle_char_store): Adjust.
14506 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
14507 Jim Wilson <jimw@sifive.com>
14509 * config/riscv/riscv-protos.h (riscv_output_return): New.
14510 * config/riscv/riscv.c (struct machine_function): New naked_p field.
14511 (riscv_attribute_table, riscv_output_return),
14512 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
14513 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
14514 (riscv_compute_frame_info): Only compute frame->mask if not a naked
14516 (riscv_expand_prologue): Add early return for naked function.
14517 (riscv_expand_epilogue): Likewise.
14518 (riscv_function_ok_for_sibcall): Return false for naked function.
14519 (riscv_set_current_function): New.
14520 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
14521 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
14522 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
14523 * doc/extend.texi (RISC-V Function Attributes): New.
14525 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
14527 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
14528 check for 128-bit long double before checking TCmode.
14529 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
14530 128-bit long doubles before checking TFmode or TCmode.
14531 (FLOAT128_IBM_P): Likewise.
14533 2018-01-10 Martin Sebor <msebor@redhat.com>
14535 PR tree-optimization/83671
14536 * builtins.c (c_strlen): Unconditionally return zero for the empty
14538 Use -Warray-bounds for warnings.
14539 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
14540 for non-constant array indices with COMPONENT_REF, arrays of
14541 arrays, and pointers to arrays.
14542 (gimple_fold_builtin_strlen): Determine and set length range for
14543 non-constant character arrays.
14545 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
14547 PR middle-end/81897
14548 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
14551 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
14553 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
14555 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
14558 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
14559 VECTOR_MEM_ALTIVEC_OR_VSX_P.
14560 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
14561 indexed_or_indirect_operand predicate.
14562 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
14563 (*vsx_le_perm_load_v8hi): Likewise.
14564 (*vsx_le_perm_load_v16qi): Likewise.
14565 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
14566 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
14567 (*vsx_le_perm_store_v8hi): Likewise.
14568 (*vsx_le_perm_store_v16qi): Likewise.
14569 (eight unnamed splitters): Likewise.
14571 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
14573 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
14574 * config/rs6000/emmintrin.h: Likewise.
14575 * config/rs6000/mmintrin.h: Likewise.
14576 * config/rs6000/xmmintrin.h: Likewise.
14578 2018-01-10 David Malcolm <dmalcolm@redhat.com>
14581 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
14583 * tree.c (tree_nop_conversion): Return true for location wrapper
14585 (maybe_wrap_with_location): New function.
14586 (selftest::check_strip_nops): New function.
14587 (selftest::test_location_wrappers): New function.
14588 (selftest::tree_c_tests): Call it.
14589 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
14590 (maybe_wrap_with_location): New decl.
14591 (EXPR_LOCATION_WRAPPER_P): New macro.
14592 (location_wrapper_p): New inline function.
14593 (tree_strip_any_location_wrapper): New inline function.
14595 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
14598 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
14599 stack_realign_offset for the largest alignment of stack slot
14601 (ix86_find_max_used_stack_alignment): New function.
14602 (ix86_finalize_stack_frame_flags): Use it. Set
14603 max_used_stack_alignment if we don't realign stack.
14604 * config/i386/i386.h (machine_function): Add
14605 max_used_stack_alignment.
14607 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
14609 * config/arm/arm.opt (-mbranch-cost): New option.
14610 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
14613 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
14616 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
14617 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
14619 2018-01-10 Richard Biener <rguenther@suse.de>
14622 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
14623 early out so it also covers the case where we have a non-NULL
14626 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
14628 PR tree-optimization/83753
14629 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
14630 for non-strided grouped accesses if the number of elements is 1.
14632 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14635 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
14636 * i386.h (TARGET_USE_GATHER): Define.
14637 * x86-tune.def (X86_TUNE_USE_GATHER): New.
14639 2018-01-10 Martin Liska <mliska@suse.cz>
14642 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
14643 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
14645 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
14646 CLEANUP_NO_PARTITIONING is not set.
14648 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
14650 * doc/rtl.texi: Remove documentation of (const ...) wrappers
14651 for vectors, as a partial revert of r254296.
14652 * rtl.h (const_vec_p): Delete.
14653 (const_vec_duplicate_p): Don't test for vector CONSTs.
14654 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
14655 * expmed.c (make_tree): Likewise.
14658 * common.md (E, F): Use CONSTANT_P instead of checking for
14660 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
14661 checking for CONST_VECTOR.
14663 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14665 PR middle-end/83575
14666 * predict.c (force_edge_cold): Handle in more sane way edges
14667 with no prediction.
14669 2018-01-09 Carl Love <cel@us.ibm.com>
14671 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
14673 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
14674 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
14675 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
14676 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
14677 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
14678 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
14679 * config/rs6000/rs6000-protos.h: Add extern defition for
14680 rs6000_generate_float2_double_code.
14681 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
14683 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
14684 (float2_v2df): Add define_expand.
14686 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
14689 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
14690 op_mode in the force_to_mode call.
14692 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
14694 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
14695 instead of checking each element individually.
14696 (aarch64_evpc_uzp): Likewise.
14697 (aarch64_evpc_zip): Likewise.
14698 (aarch64_evpc_ext): Likewise.
14699 (aarch64_evpc_rev): Likewise.
14700 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
14701 instead of checking each element individually. Return true without
14703 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
14704 whether all selected elements come from the same input, instead of
14705 checking each element individually. Remove calls to gen_rtx_REG,
14706 start_sequence and end_sequence and instead assert that no rtl is
14709 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
14711 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
14712 order of HIGH and CONST checks.
14714 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
14716 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
14717 if the destination isn't an SSA_NAME.
14719 2018-01-09 Richard Biener <rguenther@suse.de>
14721 PR tree-optimization/83668
14722 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
14724 (canonicalize_loop_form): ... here, renamed from ...
14725 (canonicalize_loop_closed_ssa_form): ... this and amended to
14726 swap successor edges for loop exit blocks to make us use
14727 the RPO order we need for initial schedule generation.
14729 2018-01-09 Joseph Myers <joseph@codesourcery.com>
14731 PR tree-optimization/64811
14732 * match.pd: When optimizing comparisons with Inf, avoid
14733 introducing or losing exceptions from comparisons with NaN.
14735 2018-01-09 Martin Liska <mliska@suse.cz>
14738 * asan.c (shadow_mem_size): Add gcc_assert.
14740 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
14742 Don't save registers in main().
14745 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
14746 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
14747 * config/avr/avr.c (avr_set_current_function): Don't error if
14748 naked, OS_task or OS_main are specified at the same time.
14749 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
14751 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
14753 * common/config/avr/avr-common.c (avr_option_optimization_table):
14754 Switch on -mmain-is-OS_task for optimizing compilations.
14756 2018-01-09 Richard Biener <rguenther@suse.de>
14758 PR tree-optimization/83572
14759 * graphite.c: Include cfganal.h.
14760 (graphite_transform_loops): Connect infinite loops to exit
14761 and remove fake edges at the end.
14763 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14765 * ipa-inline.c (edge_badness): Revert accidental checkin.
14767 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14770 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
14771 symbols; not inline clones.
14773 2018-01-09 Jakub Jelinek <jakub@redhat.com>
14776 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
14777 hard registers. Formatting fixes.
14779 PR preprocessor/83722
14780 * gcc.c (try_generate_repro): Pass
14781 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
14782 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
14785 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
14786 Kito Cheng <kito.cheng@gmail.com>
14788 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
14789 (riscv_leaf_function_p): Delete.
14790 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
14792 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14794 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
14796 (do_ifelse): New function.
14797 (do_isel): New function.
14798 (do_sub3): New function.
14799 (do_add3): New function.
14800 (do_load_mask_compare): New function.
14801 (do_overlap_load_compare): New function.
14802 (expand_compare_loop): New function.
14803 (expand_block_compare): Call expand_compare_loop() when appropriate.
14804 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
14805 option description.
14806 (-mblock-compare-inline-loop-limit): New option.
14808 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14811 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
14812 Reverse order of second and third operands in first alternative.
14813 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
14814 of first and second elements in UNSPEC_VPERMR vector.
14815 (altivec_expand_vec_perm_le): Likewise.
14817 2018-01-08 Jeff Law <law@redhat.com>
14819 PR rtl-optimizatin/81308
14820 * tree-switch-conversion.c (cfg_altered): New file scoped static.
14821 (process_switch): If group_case_labels makes a change, then set
14823 (pass_convert_switch::execute): If a switch is converted, then
14824 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
14826 PR rtl-optimization/81308
14827 * recog.c (split_all_insns): Conditionally cleanup the CFG after
14830 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
14832 PR target/83663 - Revert r255946
14833 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
14834 generation for cases where splatting a value is not useful.
14835 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
14836 across a vec_duplicate and a paradoxical subreg forming a vector
14837 mode to a vec_concat.
14839 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14841 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
14842 -march=armv8.3-a variants.
14843 * config/arm/t-multilib: Likewise.
14844 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
14846 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14848 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
14850 (cceq_ior_compare_complement): Give it a name so I can use it, and
14851 change boolean_or_operator predicate to boolean_operator so it can
14852 be used to generate a crand.
14853 (eqne): New code iterator.
14854 (bd/bd_neg): New code_attrs.
14855 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
14856 a single define_insn.
14857 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
14858 decrement (bdnzt/bdnzf/bdzt/bdzf).
14859 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
14860 with the new names of the branch decrement patterns, and added the
14861 names of the branch decrement conditional patterns.
14863 2018-01-08 Richard Biener <rguenther@suse.de>
14865 PR tree-optimization/83563
14866 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
14869 2018-01-08 Richard Biener <rguenther@suse.de>
14871 PR middle-end/83713
14872 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
14874 2018-01-08 Richard Biener <rguenther@suse.de>
14876 PR tree-optimization/83685
14877 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
14878 references to abnormals.
14880 2018-01-08 Richard Biener <rguenther@suse.de>
14883 * dwarf2out.c (output_indirect_strings): Handle empty
14884 skeleton_debug_str_hash.
14885 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
14887 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
14889 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
14890 (emit_store_direct): Likewise.
14891 (arc_trampoline_adjust_address): Likewise.
14892 (arc_asm_trampoline_template): New function.
14893 (arc_initialize_trampoline): Use asm_trampoline_template.
14894 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
14895 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
14896 * config/arc/arc.md (flush_icache): Delete pattern.
14898 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
14900 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
14901 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
14904 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
14907 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
14908 by not USED_FOR_TARGET.
14909 (make_pass_resolve_sw_modes): Likewise.
14911 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
14913 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
14916 2018-01-08 Richard Biener <rguenther@suse.de>
14918 PR middle-end/83580
14919 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
14921 2018-01-08 Richard Biener <rguenther@suse.de>
14923 PR middle-end/83517
14924 * match.pd ((t * 2) / 2) -> t): Add missing :c.
14926 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
14928 PR middle-end/81897
14929 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
14930 basic blocks with a small number of successors.
14931 (convert_control_dep_chain_into_preds): Improve handling of
14933 (dump_predicates): Split apart into...
14934 (dump_pred_chain): ...here...
14935 (dump_pred_info): ...and here.
14936 (can_one_predicate_be_invalidated_p): Add debugging printfs.
14937 (can_chain_union_be_invalidated_p): Improve check for invalidation
14939 (uninit_uses_cannot_happen): Avoid unnecessary if
14940 convert_control_dep_chain_into_preds yielded nothing.
14942 2018-01-06 Martin Sebor <msebor@redhat.com>
14944 PR tree-optimization/83640
14945 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
14946 subtracting negative offset from size.
14947 (builtin_access::overlap): Adjust offset bounds of the access to fall
14948 within the size of the object if possible.
14950 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
14952 PR rtl-optimization/83699
14953 * expmed.c (extract_bit_field_1): Restrict the vector usage of
14954 extract_bit_field_as_subreg to cases in which the extracted
14955 value is also a vector.
14957 * lra-constraints.c (process_alt_operands): Test for the equivalence
14958 substitutions when detecting a possible reload cycle.
14960 2018-01-06 Jakub Jelinek <jakub@redhat.com>
14963 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
14964 by default if flag_selective_schedling{,2}. Formatting fixes.
14966 PR rtl-optimization/83682
14967 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
14968 if it has non-VECTOR_MODE element mode.
14969 (vec_duplicate_p): Likewise.
14971 PR middle-end/83694
14972 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
14973 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
14975 2018-01-05 Jakub Jelinek <jakub@redhat.com>
14978 * config/i386/i386-builtin.def
14979 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
14980 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
14981 Require also OPTION_MASK_ISA_AVX512F in addition to
14982 OPTION_MASK_ISA_GFNI.
14983 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
14984 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
14985 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
14986 to OPTION_MASK_ISA_GFNI.
14987 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
14988 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
14989 OPTION_MASK_ISA_AVX512BW.
14990 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
14991 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
14992 addition to OPTION_MASK_ISA_GFNI.
14993 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
14994 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
14995 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
14996 to OPTION_MASK_ISA_GFNI.
14997 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
14998 a requirement for all ISAs rather than any of them with a few
15000 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
15002 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
15003 bitmasks to be enabled with 3 exceptions, instead of requiring any
15004 enabled ISA with lots of exceptions.
15005 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
15006 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
15007 Change avx512bw in isa attribute to avx512f.
15008 * config/i386/sgxintrin.h: Add license boilerplate.
15009 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
15010 to __AVX512F__ and __AVX512VL to __AVX512VL__.
15011 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
15012 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
15014 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
15015 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
15016 temporarily sse2 rather than sse if not enabled already.
15019 * config/i386/sse.md (VI248_VLBW): Rename to ...
15020 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
15021 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
15022 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
15023 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
15024 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
15025 mode iterator instead of VI248_VLBW.
15027 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
15029 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
15030 (record_modified): Skip clobbers; add debug output.
15031 (param_change_prob): Use sreal frequencies.
15033 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15035 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
15036 punt for user-aligned variables.
15038 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15040 * tree-chrec.c (chrec_contains_symbols): Return true for
15043 2018-01-05 Sudakshina Das <sudi.das@arm.com>
15046 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
15047 of (x|y) == x for BICS pattern.
15049 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15051 PR tree-optimization/83605
15052 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
15053 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
15056 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
15058 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
15059 * config/epiphany/rtems.h: New file.
15061 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15062 Uros Bizjak <ubizjak@gmail.com>
15065 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
15066 QIreg_operand instead of register_operand predicate.
15067 * config/i386/i386.c (ix86_rop_should_change_byte_p,
15068 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
15069 comments instead of -fmitigate[-_]rop.
15071 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15074 * cgraphunit.c (symbol_table::compile): Switch to text_section
15075 before calling assembly_start debug hook.
15076 * run-rtl-passes.c (run_rtl_passes): Likewise.
15079 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15081 * tree-vrp.c (extract_range_from_binary_expr_1): Check
15082 range_int_cst_p rather than !symbolic_range_p before calling
15083 extract_range_from_multiplicative_op_1.
15085 2018-01-04 Jeff Law <law@redhat.com>
15087 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
15088 redundant test in assertion.
15090 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15092 * doc/rtl.texi: Document machine_mode wrapper classes.
15094 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15096 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
15097 using tree_to_uhwi.
15099 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15101 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
15102 the VEC_PERM_EXPR fold to fail.
15104 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15107 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
15108 to switched_sections.
15110 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15113 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
15114 test for d.testing.
15116 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
15119 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
15120 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
15122 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15125 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
15126 is BLKmode and bitpos not zero or mode change is needed.
15128 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15131 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
15134 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
15137 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
15138 instead of MULT rtx. Update all corresponding splitters.
15139 (*saddl_se): Ditto.
15140 (*ssub<modesuffix>): Ditto.
15141 (*ssubl_se): Ditto.
15142 (*cmp_sadd_di): Update split patterns.
15143 (*cmp_sadd_si): Ditto.
15144 (*cmp_sadd_sidi): Ditto.
15145 (*cmp_ssub_di): Ditto.
15146 (*cmp_ssub_si): Ditto.
15147 (*cmp_ssub_sidi): Ditto.
15148 * config/alpha/predicates.md (const23_operand): New predicate.
15149 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
15150 Look for ASHIFT, not MULT inner operand.
15151 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
15153 2018-01-04 Martin Liska <mliska@suse.cz>
15155 PR gcov-profile/83669
15156 * gcov.c (output_intermediate_file): Add version to intermediate
15158 * doc/gcov.texi: Document new field 'version' in intermediate
15159 file format. Fix location of '-k' option of gcov command.
15161 2018-01-04 Martin Liska <mliska@suse.cz>
15164 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
15166 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15168 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
15170 2018-01-03 Martin Sebor <msebor@redhat.com>
15172 PR tree-optimization/83655
15173 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
15174 checking calls with invalid arguments.
15176 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15178 * tree-vect-stmts.c (vect_get_store_rhs): New function.
15179 (vectorizable_mask_load_store): Delete.
15180 (vectorizable_call): Return false for masked loads and stores.
15181 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
15182 instead of gimple_assign_rhs1.
15183 (vectorizable_load): Handle IFN_MASK_LOAD.
15184 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
15186 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15188 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
15190 (vectorizable_mask_load_store): ...here.
15191 (vectorizable_load): ...and here.
15193 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15195 * tree-vect-stmts.c (vect_build_all_ones_mask)
15196 (vect_build_zero_merge_argument): New functions, split out from...
15197 (vectorizable_load): ...here.
15199 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15201 * tree-vect-stmts.c (vect_check_store_rhs): New function,
15203 (vectorizable_mask_load_store): ...here.
15204 (vectorizable_store): ...and here.
15206 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15208 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
15210 (vectorizable_mask_load_store): ...here.
15212 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15214 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
15215 (vect_model_store_cost): Take a vec_load_store_type instead of a
15217 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
15218 (vect_model_store_cost): Take a vec_load_store_type instead of a
15220 (vectorizable_mask_load_store): Update accordingly.
15221 (vectorizable_store): Likewise.
15222 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
15224 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15226 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
15227 IFN_MASK_LOAD calls here rather than...
15228 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
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 * expmed.c (extract_bit_field_1): For vector extracts,
15235 fall back to extract_bit_field_as_subreg if vec_extract
15238 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15239 Alan Hayward <alan.hayward@arm.com>
15240 David Sherwood <david.sherwood@arm.com>
15242 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
15243 they are variable or constant sized.
15244 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
15245 slots for constant-sized data.
15247 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15248 Alan Hayward <alan.hayward@arm.com>
15249 David Sherwood <david.sherwood@arm.com>
15251 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
15252 handling COND_EXPRs with boolean comparisons, try to find a better
15253 basis for the mask type than the boolean itself.
15255 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15257 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
15258 is calculated and how it can be overridden.
15259 * genmodes.c (max_bitsize_mode_any_mode): New variable.
15260 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
15262 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
15265 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15266 Alan Hayward <alan.hayward@arm.com>
15267 David Sherwood <david.sherwood@arm.com>
15269 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
15270 Remove the mode argument.
15271 (aarch64_simd_valid_immediate): Remove the mode and inverse
15273 * config/aarch64/iterators.md (bitsize): New iterator.
15274 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
15275 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
15276 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
15277 aarch64_simd_valid_immediate.
15278 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
15279 (aarch64_reg_or_bic_imm): Likewise.
15280 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
15281 with an insn_type enum and msl with a modifier_type enum.
15282 Replace element_width with a scalar_mode. Change the shift
15283 to unsigned int. Add constructors for scalar_float_mode and
15284 scalar_int_mode elements.
15285 (aarch64_vect_float_const_representable_p): Delete.
15286 (aarch64_can_const_movi_rtx_p)
15287 (aarch64_simd_scalar_immediate_valid_for_move)
15288 (aarch64_simd_make_constant): Update call to
15289 aarch64_simd_valid_immediate.
15290 (aarch64_advsimd_valid_immediate_hs): New function.
15291 (aarch64_advsimd_valid_immediate): Likewise.
15292 (aarch64_simd_valid_immediate): Remove mode and inverse
15293 arguments. Rewrite to use the above. Use const_vec_duplicate_p
15294 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
15295 and aarch64_float_const_representable_p on the result.
15296 (aarch64_output_simd_mov_immediate): Remove mode argument.
15297 Update call to aarch64_simd_valid_immediate and use of
15298 simd_immediate_info.
15299 (aarch64_output_scalar_simd_mov_immediate): Update call
15302 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15303 Alan Hayward <alan.hayward@arm.com>
15304 David Sherwood <david.sherwood@arm.com>
15306 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
15307 (mode_nunits): Likewise CONST_MODE_NUNITS.
15308 * machmode.def (ADJUST_NUNITS): Document.
15309 * genmodes.c (mode_data::need_nunits_adj): New field.
15310 (blank_mode): Update accordingly.
15311 (adj_nunits): New variable.
15312 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
15314 (emit_mode_size_inline): Set need_bytesize_adj for all modes
15315 listed in adj_nunits.
15316 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
15317 listed in adj_nunits. Don't emit case statements for such modes.
15318 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
15319 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
15320 nothing if adj_nunits is nonnull.
15321 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
15322 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
15323 (emit_mode_fbit): Update use of print_maybe_const_decl.
15324 (emit_move_size): Likewise. Treat the array as non-const
15326 (emit_mode_adjustments): Handle adj_nunits.
15328 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15330 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
15331 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
15332 (VECTOR_MODES): Use it.
15333 (make_vector_modes): Take the prefix as an argument.
15335 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15336 Alan Hayward <alan.hayward@arm.com>
15337 David Sherwood <david.sherwood@arm.com>
15339 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
15340 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
15341 for MODE_VECTOR_BOOL.
15342 * machmode.def (VECTOR_BOOL_MODE): Document.
15343 * genmodes.c (VECTOR_BOOL_MODE): New macro.
15344 (make_vector_bool_mode): New function.
15345 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
15347 * lto-streamer-in.c (lto_input_mode_table): Likewise.
15348 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
15350 * stor-layout.c (int_mode_for_mode): Likewise.
15351 * tree.c (build_vector_type_for_mode): Likewise.
15352 * varasm.c (output_constant_pool_2): Likewise.
15353 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
15354 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
15355 for MODE_VECTOR_BOOL.
15356 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
15357 of mode class checks.
15358 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
15359 instead of a list of mode class checks.
15360 (expand_vector_scalar_condition): Likewise.
15361 (type_for_widest_vector_mode): Handle BImode as an inner mode.
15363 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15364 Alan Hayward <alan.hayward@arm.com>
15365 David Sherwood <david.sherwood@arm.com>
15367 * machmode.h (mode_size): Change from unsigned short to
15369 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
15370 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
15371 or if measurement_type is not polynomial.
15372 (fixed_size_mode::includes_p): Check for constant-sized modes.
15373 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
15374 return a poly_uint16 rather than an unsigned short.
15375 (emit_mode_size): Change the type of mode_size from unsigned short
15376 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
15377 (emit_mode_adjustments): Cope with polynomial vector sizes.
15378 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15380 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15382 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
15383 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
15384 * caller-save.c (setup_save_areas): Likewise.
15385 (replace_reg_with_saved_mem): Likewise.
15386 * calls.c (emit_library_call_value_1): Likewise.
15387 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
15388 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
15389 (gen_lowpart_for_combine): Likewise.
15390 * convert.c (convert_to_integer_1): Likewise.
15391 * cse.c (equiv_constant, cse_insn): Likewise.
15392 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
15393 (cselib_subst_to_values): Likewise.
15394 * dce.c (word_dce_process_block): Likewise.
15395 * df-problems.c (df_word_lr_mark_ref): Likewise.
15396 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
15397 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
15398 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
15399 (rtl_for_decl_location): Likewise.
15400 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
15401 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
15402 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
15403 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
15404 (expand_expr_real_1): Likewise.
15405 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
15406 (pad_below): Likewise.
15407 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
15408 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
15409 * ira.c (get_subreg_tracking_sizes): Likewise.
15410 * ira-build.c (ira_create_allocno_objects): Likewise.
15411 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
15412 (ira_sort_regnos_for_alter_reg): Likewise.
15413 * ira-costs.c (record_operand_costs): Likewise.
15414 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
15415 (resolve_simple_move): Likewise.
15416 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
15417 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
15418 (lra_constraints): Likewise.
15419 (CONST_POOL_OK_P): Reject variable-sized modes.
15420 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
15421 (add_pseudo_to_slot, lra_spill): Likewise.
15422 * omp-low.c (omp_clause_aligned_alignment): Likewise.
15423 * optabs-query.c (get_best_extraction_insn): Likewise.
15424 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
15425 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
15426 (expand_mult_highpart, valid_multiword_target_p): Likewise.
15427 * recog.c (offsettable_address_addr_space_p): Likewise.
15428 * regcprop.c (maybe_mode_change): Likewise.
15429 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
15430 * regrename.c (build_def_use): Likewise.
15431 * regstat.c (dump_reg_info): Likewise.
15432 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
15433 (find_reloads, find_reloads_subreg_address): Likewise.
15434 * reload1.c (eliminate_regs_1): Likewise.
15435 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
15436 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
15437 (simplify_binary_operation_1, simplify_subreg): Likewise.
15438 * targhooks.c (default_function_arg_padding): Likewise.
15439 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
15440 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
15441 (verify_gimple_assign_ternary): Likewise.
15442 * tree-inline.c (estimate_move_cost): Likewise.
15443 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15444 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
15445 (get_address_cost_ainc): Likewise.
15446 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
15447 (vect_supportable_dr_alignment): Likewise.
15448 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15449 (vectorizable_reduction): Likewise.
15450 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
15451 (vectorizable_operation, vectorizable_load): Likewise.
15452 * tree.c (build_same_sized_truth_vector_type): Likewise.
15453 * valtrack.c (cleanup_auto_inc_dec): Likewise.
15454 * var-tracking.c (emit_note_insn_var_location): Likewise.
15455 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
15456 (ADDR_VEC_ALIGN): Likewise.
15458 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15459 Alan Hayward <alan.hayward@arm.com>
15460 David Sherwood <david.sherwood@arm.com>
15462 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
15464 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
15465 or if measurement_type is polynomial.
15466 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
15467 * combine.c (make_extraction): Likewise.
15468 * dse.c (find_shift_sequence): Likewise.
15469 * dwarf2out.c (mem_loc_descriptor): Likewise.
15470 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
15471 (extract_bit_field, extract_low_bits): Likewise.
15472 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
15473 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
15474 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
15475 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
15476 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
15477 * reload.c (find_reloads): Likewise.
15478 * reload1.c (alter_reg): Likewise.
15479 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
15480 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
15481 * tree-if-conv.c (predicate_mem_writes): Likewise.
15482 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
15483 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
15484 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
15485 * valtrack.c (dead_debug_insert_temp): Likewise.
15486 * varasm.c (mergeable_constant_section): Likewise.
15487 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
15489 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15490 Alan Hayward <alan.hayward@arm.com>
15491 David Sherwood <david.sherwood@arm.com>
15493 * expr.c (expand_assignment): Cope with polynomial mode sizes
15494 when assigning to a CONCAT.
15496 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15497 Alan Hayward <alan.hayward@arm.com>
15498 David Sherwood <david.sherwood@arm.com>
15500 * machmode.h (mode_precision): Change from unsigned short to
15502 (mode_to_precision): Return a poly_uint16 rather than an unsigned
15504 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
15505 or if measurement_type is not polynomial.
15506 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
15507 in which the mode is already known to be a scalar_int_mode.
15508 * genmodes.c (emit_mode_precision): Change the type of mode_precision
15509 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
15511 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15512 for GET_MODE_PRECISION.
15513 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15514 for GET_MODE_PRECISION.
15515 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
15517 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
15518 (expand_field_assignment, make_extraction): Likewise.
15519 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
15520 (get_last_value): Likewise.
15521 * convert.c (convert_to_integer_1): Likewise.
15522 * cse.c (cse_insn): Likewise.
15523 * expr.c (expand_expr_real_1): Likewise.
15524 * lra-constraints.c (simplify_operand_subreg): Likewise.
15525 * optabs-query.c (can_atomic_load_p): Likewise.
15526 * optabs.c (expand_atomic_load): Likewise.
15527 (expand_atomic_store): Likewise.
15528 * ree.c (combine_reaching_defs): Likewise.
15529 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
15530 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
15531 * tree.h (type_has_mode_precision_p): Likewise.
15532 * ubsan.c (instrument_si_overflow): Likewise.
15534 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15535 Alan Hayward <alan.hayward@arm.com>
15536 David Sherwood <david.sherwood@arm.com>
15538 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
15539 polynomial numbers of units.
15540 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
15541 (valid_vector_subparts_p): New function.
15542 (build_vector_type): Remove temporary shim and take the number
15543 of units as a poly_uint64 rather than an int.
15544 (build_opaque_vector_type): Take the number of units as a
15545 poly_uint64 rather than an int.
15546 * tree.c (build_vector_from_ctor): Handle polynomial
15547 TYPE_VECTOR_SUBPARTS.
15548 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
15549 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
15550 (build_vector_from_val): If the number of units is variable,
15551 use build_vec_duplicate_cst for constant operands and
15552 VEC_DUPLICATE_EXPR otherwise.
15553 (make_vector_type): Remove temporary is_constant ().
15554 (build_vector_type, build_opaque_vector_type): Take the number of
15555 units as a poly_uint64 rather than an int.
15556 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
15558 * cfgexpand.c (expand_debug_expr): Likewise.
15559 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
15560 (store_constructor, expand_expr_real_1): Likewise.
15561 (const_scalar_mask_from_tree): Likewise.
15562 * fold-const-call.c (fold_const_reduction): Likewise.
15563 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
15564 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
15565 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
15566 (fold_relational_const): Likewise.
15567 (native_interpret_vector): Likewise. Change the size from an
15568 int to an unsigned int.
15569 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
15570 TYPE_VECTOR_SUBPARTS.
15571 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
15572 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
15573 duplicating a non-constant operand into a variable-length vector.
15574 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
15575 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
15576 * ipa-icf.c (sem_variable::equals): Likewise.
15577 * match.pd: Likewise.
15578 * omp-simd-clone.c (simd_clone_subparts): Likewise.
15579 * print-tree.c (print_node): Likewise.
15580 * stor-layout.c (layout_type): Likewise.
15581 * targhooks.c (default_builtin_vectorization_cost): Likewise.
15582 * tree-cfg.c (verify_gimple_comparison): Likewise.
15583 (verify_gimple_assign_binary): Likewise.
15584 (verify_gimple_assign_ternary): Likewise.
15585 (verify_gimple_assign_single): Likewise.
15586 * tree-pretty-print.c (dump_generic_node): Likewise.
15587 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15588 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
15589 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
15590 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
15591 (vect_shift_permute_load_chain): Likewise.
15592 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
15593 (expand_vector_condition, optimize_vector_constructor): Likewise.
15594 (lower_vec_perm, get_compute_type): Likewise.
15595 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15596 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
15597 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
15598 (vect_recog_mask_conversion_pattern): Likewise.
15599 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
15600 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
15601 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15602 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
15603 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
15604 (vectorizable_shift, vectorizable_operation, vectorizable_store)
15605 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
15606 (supportable_widening_operation): Likewise.
15607 (supportable_narrowing_operation): Likewise.
15608 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
15610 * varasm.c (output_constant): Likewise.
15612 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15613 Alan Hayward <alan.hayward@arm.com>
15614 David Sherwood <david.sherwood@arm.com>
15616 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
15617 so that both the length == 3 and length != 3 cases set up their
15618 own permute vectors. Add comments explaining why we know the
15619 number of elements is constant.
15620 (vect_permute_load_chain): Likewise.
15622 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15623 Alan Hayward <alan.hayward@arm.com>
15624 David Sherwood <david.sherwood@arm.com>
15626 * machmode.h (mode_nunits): Change from unsigned char to
15628 (ONLY_FIXED_SIZE_MODES): New macro.
15629 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
15630 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
15631 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
15633 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
15634 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
15635 or if measurement_type is not polynomial.
15636 * genmodes.c (ZERO_COEFFS): New macro.
15637 (emit_mode_nunits_inline): Make mode_nunits_inline return a
15639 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
15640 Use ZERO_COEFFS when emitting initializers.
15641 * data-streamer.h (bp_pack_poly_value): New function.
15642 (bp_unpack_poly_value): Likewise.
15643 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15644 for GET_MODE_NUNITS.
15645 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15646 for GET_MODE_NUNITS.
15647 * tree.c (make_vector_type): Remove temporary shim and make
15648 the real function take the number of units as a poly_uint64
15649 rather than an int.
15650 (build_vector_type_for_mode): Handle polynomial nunits.
15651 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
15652 * emit-rtl.c (const_vec_series_p_1): Likewise.
15653 (gen_rtx_CONST_VECTOR): Likewise.
15654 * fold-const.c (test_vec_duplicate_folding): Likewise.
15655 * genrecog.c (validate_pattern): Likewise.
15656 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
15657 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
15658 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
15659 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
15660 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
15661 * rtlanal.c (subreg_get_info): Likewise.
15662 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15663 (vect_grouped_load_supported): Likewise.
15664 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
15665 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
15666 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
15667 (simplify_const_unary_operation, simplify_binary_operation_1)
15668 (simplify_const_binary_operation, simplify_ternary_operation)
15669 (test_vector_ops_duplicate, test_vector_ops): Likewise.
15670 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
15671 instead of CONST_VECTOR_NUNITS.
15672 * varasm.c (output_constant_pool_2): Likewise.
15673 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
15674 explicit-encoded elements in the XVEC for variable-length vectors.
15676 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15678 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
15680 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15681 Alan Hayward <alan.hayward@arm.com>
15682 David Sherwood <david.sherwood@arm.com>
15684 * coretypes.h (fixed_size_mode): Declare.
15685 (fixed_size_mode_pod): New typedef.
15686 * builtins.h (target_builtins::x_apply_args_mode)
15687 (target_builtins::x_apply_result_mode): Change type to
15688 fixed_size_mode_pod.
15689 * builtins.c (apply_args_size, apply_result_size, result_vector)
15690 (expand_builtin_apply_args_1, expand_builtin_apply)
15691 (expand_builtin_return): Update accordingly.
15693 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15695 * cse.c (hash_rtx_cb): Hash only the encoded elements.
15696 * cselib.c (cselib_hash_rtx): Likewise.
15697 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
15698 CONST_VECTOR encoding.
15700 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15701 Jeff Law <law@redhat.com>
15704 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
15705 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
15706 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
15707 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
15710 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
15711 explicitly probe *sp in a noreturn function if there were any callee
15712 register saves or frame pointer is needed.
15714 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15717 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
15718 BLKmode for ternary, binary or unary expressions.
15721 * var-tracking.c (delete_vta_debug_insn): New inline function.
15722 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
15723 insns from get_insns () to NULL instead of each bb separately.
15724 Use delete_vta_debug_insn. No longer static.
15725 (vt_debug_insns_local, variable_tracking_main_1): Adjust
15726 delete_vta_debug_insns callers.
15727 * rtl.h (delete_vta_debug_insns): Declare.
15728 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
15729 instead of variable_tracking_main.
15731 2018-01-03 Martin Sebor <msebor@redhat.com>
15733 PR tree-optimization/83603
15734 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
15735 arguments past the endof the argument list in functions declared
15736 without a prototype.
15737 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
15738 Avoid checking when arguments are null.
15740 2018-01-03 Martin Sebor <msebor@redhat.com>
15743 * doc/extend.texi (attribute const): Fix a typo.
15744 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
15745 issuing -Wsuggest-attribute for void functions.
15747 2018-01-03 Martin Sebor <msebor@redhat.com>
15749 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
15750 offset_int::from instead of wide_int::to_shwi.
15751 (maybe_diag_overlap): Remove assertion.
15752 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
15753 * gimple-ssa-sprintf.c (format_directive): Same.
15754 (parse_directive): Same.
15755 (sprintf_dom_walker::compute_format_length): Same.
15756 (try_substitute_return_value): Same.
15758 2018-01-03 Jeff Law <law@redhat.com>
15760 PR middle-end/83654
15761 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
15762 non-constant residual for zero at runtime and avoid probing in
15763 that case. Reorganize code for trailing problem to mirror handling
15766 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15768 PR tree-optimization/83501
15769 * tree-ssa-strlen.c (get_string_cst): New.
15770 (handle_char_store): Call get_string_cst.
15772 2018-01-03 Martin Liska <mliska@suse.cz>
15774 PR tree-optimization/83593
15775 * tree-ssa-strlen.c: Include tree-cfg.h.
15776 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
15777 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
15778 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
15780 (strlen_dom_walker::before_dom_children): Call
15781 gimple_purge_dead_eh_edges. Dump tranformation with details
15783 (strlen_dom_walker::before_dom_children): Update call by adding
15784 new argument cleanup_eh.
15785 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
15787 2018-01-03 Martin Liska <mliska@suse.cz>
15790 * cif-code.def (VARIADIC_THUNK): New enum value.
15791 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
15794 2018-01-03 Jan Beulich <jbeulich@suse.com>
15796 * sse.md (mov<mode>_internal): Tighten condition for when to use
15797 vmovdqu<ssescalarsize> for TI and OI modes.
15799 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15801 Update copyright years.
15803 2018-01-03 Martin Liska <mliska@suse.cz>
15806 * ipa-visibility.c (function_and_variable_visibility): Skip
15807 functions with noipa attribure.
15809 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15811 * gcc.c (process_command): Update copyright notice dates.
15812 * gcov-dump.c (print_version): Ditto.
15813 * gcov.c (print_version): Ditto.
15814 * gcov-tool.c (print_version): Ditto.
15815 * gengtype.c (create_file): Ditto.
15816 * doc/cpp.texi: Bump @copying's copyright year.
15817 * doc/cppinternals.texi: Ditto.
15818 * doc/gcc.texi: Ditto.
15819 * doc/gccint.texi: Ditto.
15820 * doc/gcov.texi: Ditto.
15821 * doc/install.texi: Ditto.
15822 * doc/invoke.texi: Ditto.
15824 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15826 * vector-builder.h (vector_builder::m_full_nelts): Change from
15827 unsigned int to poly_uint64.
15828 (vector_builder::full_nelts): Update prototype accordingly.
15829 (vector_builder::new_vector): Likewise.
15830 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
15831 (vector_builder::operator ==): Likewise.
15832 (vector_builder::finalize): Likewise.
15833 * int-vector-builder.h (int_vector_builder::int_vector_builder):
15834 Take the number of elements as a poly_uint64 rather than an
15836 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
15837 from unsigned int to poly_uint64.
15838 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
15839 (vec_perm_indices::new_vector): Likewise.
15840 (vec_perm_indices::length): Likewise.
15841 (vec_perm_indices::nelts_per_input): Likewise.
15842 (vec_perm_indices::input_nelts): Likewise.
15843 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
15844 number of elements per input as a poly_uint64 rather than an
15845 unsigned int. Use the original encoding for variable-length
15846 vectors, rather than clamping each individual element.
15847 For the second and subsequent elements in each pattern,
15848 clamp the step and base before clamping their sum.
15849 (vec_perm_indices::series_p): Handle polynomial element counts.
15850 (vec_perm_indices::all_in_range_p): Likewise.
15851 (vec_perm_indices_to_tree): Likewise.
15852 (vec_perm_indices_to_rtx): Likewise.
15853 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
15854 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
15855 (tree_vector_builder::new_binary_operation): Handle polynomial
15856 element counts. Return false if we need to know the number
15857 of elements at compile time.
15858 * fold-const.c (fold_vec_perm): Punt if the number of elements
15859 isn't known at compile time.
15861 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15863 * vec-perm-indices.h (vec_perm_builder): Change element type
15864 from HOST_WIDE_INT to poly_int64.
15865 (vec_perm_indices::element_type): Update accordingly.
15866 (vec_perm_indices::clamp): Handle polynomial element_types.
15867 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
15868 (vec_perm_indices::all_in_range_p): Likewise.
15869 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
15871 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
15872 polynomial vec_perm_indices element types.
15873 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
15874 * fold-const.c (fold_vec_perm): Likewise.
15875 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
15876 * tree-vect-generic.c (lower_vec_perm): Likewise.
15877 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
15878 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
15879 element type to HOST_WIDE_INT.
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 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
15886 rather than an int. Use plus_constant.
15887 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
15888 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
15890 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15891 Alan Hayward <alan.hayward@arm.com>
15892 David Sherwood <david.sherwood@arm.com>
15894 * calls.c (emit_call_1, expand_call): Change struct_value_size from
15895 a HOST_WIDE_INT to a poly_int64.
15897 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15898 Alan Hayward <alan.hayward@arm.com>
15899 David Sherwood <david.sherwood@arm.com>
15901 * calls.c (load_register_parameters): Cope with polynomial
15902 mode sizes. Require a constant size for BLKmode parameters
15903 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
15904 forces a parameter to be padded at the lsb end in order to
15905 fill a complete number of words, require the parameter size
15906 to be ordered wrt UNITS_PER_WORD.
15908 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15909 Alan Hayward <alan.hayward@arm.com>
15910 David Sherwood <david.sherwood@arm.com>
15912 * reload1.c (spill_stack_slot_width): Change element type
15913 from unsigned int to poly_uint64_pod.
15914 (alter_reg): Treat mode sizes as polynomial.
15916 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15917 Alan Hayward <alan.hayward@arm.com>
15918 David Sherwood <david.sherwood@arm.com>
15920 * reload.c (complex_word_subreg_p): New function.
15921 (reload_inner_reg_of_subreg, push_reload): Use it.
15923 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15924 Alan Hayward <alan.hayward@arm.com>
15925 David Sherwood <david.sherwood@arm.com>
15927 * lra-constraints.c (process_alt_operands): Reject matched
15928 operands whose sizes aren't ordered.
15929 (match_reload): Refer to this check here.
15931 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15932 Alan Hayward <alan.hayward@arm.com>
15933 David Sherwood <david.sherwood@arm.com>
15935 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
15936 that the mode size is in the set {1, 2, 4, 8, 16}.
15938 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15939 Alan Hayward <alan.hayward@arm.com>
15940 David Sherwood <david.sherwood@arm.com>
15942 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
15943 Use plus_constant instead of gen_rtx_PLUS.
15945 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15946 Alan Hayward <alan.hayward@arm.com>
15947 David Sherwood <david.sherwood@arm.com>
15949 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
15950 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
15951 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
15952 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
15953 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
15954 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
15955 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
15956 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
15957 * config/i386/i386.c (ix86_push_rounding): ...this new function.
15958 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
15960 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
15961 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
15962 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
15963 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
15964 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
15965 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
15966 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
15967 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
15968 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
15969 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
15971 * expr.c (emit_move_resolve_push): Treat the input and result
15972 of PUSH_ROUNDING as a poly_int64.
15973 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
15974 (emit_push_insn): Likewise.
15975 * lra-eliminations.c (mark_not_eliminable): Likewise.
15976 * recog.c (push_operand): Likewise.
15977 * reload1.c (elimination_effects): Likewise.
15978 * rtlanal.c (nonzero_bits1): Likewise.
15979 * calls.c (store_one_arg): Likewise. Require the padding to be
15980 known at compile time.
15982 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15983 Alan Hayward <alan.hayward@arm.com>
15984 David Sherwood <david.sherwood@arm.com>
15986 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
15987 Use plus_constant instead of gen_rtx_PLUS.
15989 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15990 Alan Hayward <alan.hayward@arm.com>
15991 David Sherwood <david.sherwood@arm.com>
15993 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
15994 rather than an int.
15996 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15997 Alan Hayward <alan.hayward@arm.com>
15998 David Sherwood <david.sherwood@arm.com>
16000 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
16001 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
16002 via stack temporaries. Treat the mode size as polynomial too.
16004 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16005 Alan Hayward <alan.hayward@arm.com>
16006 David Sherwood <david.sherwood@arm.com>
16008 * expr.c (expand_expr_real_2): When handling conversions involving
16009 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
16010 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
16011 as a poly_uint64 too.
16013 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16014 Alan Hayward <alan.hayward@arm.com>
16015 David Sherwood <david.sherwood@arm.com>
16017 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
16019 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16020 Alan Hayward <alan.hayward@arm.com>
16021 David Sherwood <david.sherwood@arm.com>
16023 * combine.c (can_change_dest_mode): Handle polynomial
16024 REGMODE_NATURAL_SIZE.
16025 * expmed.c (store_bit_field_1): Likewise.
16026 * expr.c (store_constructor): Likewise.
16027 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
16028 and polynomial REGMODE_NATURAL_SIZE.
16029 (gen_lowpart_common): Likewise.
16030 * reginfo.c (record_subregs_of_mode): Likewise.
16031 * rtlanal.c (read_modify_subreg_p): Likewise.
16033 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16034 Alan Hayward <alan.hayward@arm.com>
16035 David Sherwood <david.sherwood@arm.com>
16037 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
16038 numbers of elements.
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 * match.pd: Cope with polynomial numbers of vector elements.
16046 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16047 Alan Hayward <alan.hayward@arm.com>
16048 David Sherwood <david.sherwood@arm.com>
16050 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
16051 in a POINTER_PLUS_EXPR.
16053 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16054 Alan Hayward <alan.hayward@arm.com>
16055 David Sherwood <david.sherwood@arm.com>
16057 * omp-simd-clone.c (simd_clone_subparts): New function.
16058 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
16059 (ipa_simd_modify_function_body): Likewise.
16061 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16062 Alan Hayward <alan.hayward@arm.com>
16063 David Sherwood <david.sherwood@arm.com>
16065 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
16066 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
16067 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
16068 (expand_vector_condition, vector_element): Likewise.
16069 (subparts_gt): New function.
16070 (get_compute_type): Use subparts_gt.
16071 (count_type_subparts): Delete.
16072 (expand_vector_operations_1): Use subparts_gt instead of
16073 count_type_subparts.
16075 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16076 Alan Hayward <alan.hayward@arm.com>
16077 David Sherwood <david.sherwood@arm.com>
16079 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
16080 (vect_compile_time_alias): ...this new function. Do the calculation
16081 on poly_ints rather than trees.
16082 (vect_prune_runtime_alias_test_list): Update call accordingly.
16084 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16085 Alan Hayward <alan.hayward@arm.com>
16086 David Sherwood <david.sherwood@arm.com>
16088 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
16090 (vect_schedule_slp_instance): Likewise.
16092 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16093 Alan Hayward <alan.hayward@arm.com>
16094 David Sherwood <david.sherwood@arm.com>
16096 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
16097 constant and extern definitions for variable-length vectors.
16098 (vect_get_constant_vectors): Note that the number of units
16099 is known to be constant.
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-stmts.c (vectorizable_conversion): Treat the number
16106 of units as polynomial. Choose between WIDE and NARROW based
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-stmts.c (simd_clone_subparts): New function.
16114 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
16116 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16117 Alan Hayward <alan.hayward@arm.com>
16118 David Sherwood <david.sherwood@arm.com>
16120 * tree-vect-stmts.c (vectorizable_call): Treat the number of
16121 vectors as polynomial. Use build_index_vector for
16122 IFN_GOMP_SIMD_LANE.
16124 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16125 Alan Hayward <alan.hayward@arm.com>
16126 David Sherwood <david.sherwood@arm.com>
16128 * tree-vect-stmts.c (get_load_store_type): Treat the number of
16129 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
16130 for variable-length vectors.
16131 (vectorizable_mask_load_store): Treat the number of units as
16132 polynomial, asserting that it is constant if the condition has
16133 already been enforced.
16134 (vectorizable_store, vectorizable_load): Likewise.
16136 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16137 Alan Hayward <alan.hayward@arm.com>
16138 David Sherwood <david.sherwood@arm.com>
16140 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
16141 of units as polynomial. Punt if we can't tell at compile time
16142 which vector contains the final result.
16144 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16145 Alan Hayward <alan.hayward@arm.com>
16146 David Sherwood <david.sherwood@arm.com>
16148 * tree-vect-loop.c (vectorizable_induction): Treat the number
16149 of units as polynomial. Punt on SLP inductions. Use an integer
16150 VEC_SERIES_EXPR for variable-length integer reductions. Use a
16151 cast of such a series for variable-length floating-point
16154 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16155 Alan Hayward <alan.hayward@arm.com>
16156 David Sherwood <david.sherwood@arm.com>
16158 * tree.h (build_index_vector): Declare.
16159 * tree.c (build_index_vector): New function.
16160 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
16161 of units as polynomial, forcibly converting it to a constant if
16162 vectorizable_reduction has already enforced the condition.
16163 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
16164 to create a {1,2,3,...} vector.
16165 (vectorizable_reduction): Treat the number of units as polynomial.
16166 Choose vectype_in based on the largest scalar element size rather
16167 than the smallest number of units. Enforce the restrictions
16170 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16171 Alan Hayward <alan.hayward@arm.com>
16172 David Sherwood <david.sherwood@arm.com>
16174 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
16175 number of units as polynomial.
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.h (vector_sizes, auto_vector_sizes): New typedefs.
16182 * target.def (autovectorize_vector_sizes): Return the vector sizes
16183 by pointer, using vector_sizes rather than a bitmask.
16184 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
16185 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
16186 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
16188 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
16189 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
16190 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
16191 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
16192 * omp-general.c (omp_max_vf): Likewise.
16193 * omp-low.c (omp_clause_aligned_alignment): Likewise.
16194 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
16195 * tree-vect-loop.c (vect_analyze_loop): Likewise.
16196 * tree-vect-slp.c (vect_slp_bb): Likewise.
16197 * doc/tm.texi: Regenerate.
16198 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
16200 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
16201 the vector size as a poly_uint64 rather than an unsigned int.
16202 (current_vector_size): Change from an unsigned int to a poly_uint64.
16203 (get_vectype_for_scalar_type): Update accordingly.
16204 * tree.h (build_truth_vector_type): Take the size and number of
16205 units as a poly_uint64 rather than an unsigned int.
16206 (build_vector_type): Add a temporary overload that takes
16207 the number of units as a poly_uint64 rather than an unsigned int.
16208 * tree.c (make_vector_type): Likewise.
16209 (build_truth_vector_type): Take the number of units as a poly_uint64
16210 rather than an unsigned int.
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 * target.def (get_mask_mode): Take the number of units and length
16217 as poly_uint64s rather than unsigned ints.
16218 * targhooks.h (default_get_mask_mode): Update accordingly.
16219 * targhooks.c (default_get_mask_mode): Likewise.
16220 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
16221 * doc/tm.texi: Regenerate.
16223 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16224 Alan Hayward <alan.hayward@arm.com>
16225 David Sherwood <david.sherwood@arm.com>
16227 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
16228 * omp-general.c (omp_max_vf): Likewise.
16229 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
16230 (expand_omp_simd): Handle polynomial safelen.
16231 * omp-low.c (omplow_simd_context): Add a default constructor.
16232 (omplow_simd_context::max_vf): Change from int to poly_uint64.
16233 (lower_rec_simd_input_clauses): Update accordingly.
16234 (lower_rec_input_clauses): Likewise.
16236 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16237 Alan Hayward <alan.hayward@arm.com>
16238 David Sherwood <david.sherwood@arm.com>
16240 * tree-vectorizer.h (vect_nunits_for_cost): New function.
16241 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
16242 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
16243 (vect_analyze_slp_cost): Likewise.
16244 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
16245 (vect_model_load_cost): Likewise.
16247 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16248 Alan Hayward <alan.hayward@arm.com>
16249 David Sherwood <david.sherwood@arm.com>
16251 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
16252 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
16253 from an unsigned int * to a poly_uint64_pod *.
16254 (calculate_unrolling_factor): New function.
16255 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
16257 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16258 Alan Hayward <alan.hayward@arm.com>
16259 David Sherwood <david.sherwood@arm.com>
16261 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
16262 from an unsigned int to a poly_uint64.
16263 (_loop_vec_info::slp_unrolling_factor): Likewise.
16264 (_loop_vec_info::vectorization_factor): Change from an int
16266 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
16267 (vect_get_num_vectors): New function.
16268 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
16269 (vect_get_num_copies): Use vect_get_num_vectors.
16270 (vect_analyze_data_ref_dependences): Change max_vf from an int *
16271 to an unsigned int *.
16272 (vect_analyze_data_refs): Change min_vf from an int * to a
16274 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
16275 than an unsigned HOST_WIDE_INT.
16276 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
16277 (vect_analyze_data_ref_dependence): Change max_vf from an int *
16278 to an unsigned int *.
16279 (vect_analyze_data_ref_dependences): Likewise.
16280 (vect_compute_data_ref_alignment): Handle polynomial vf.
16281 (vect_enhance_data_refs_alignment): Likewise.
16282 (vect_prune_runtime_alias_test_list): Likewise.
16283 (vect_shift_permute_load_chain): Likewise.
16284 (vect_supportable_dr_alignment): Likewise.
16285 (dependence_distance_ge_vf): Take the vectorization factor as a
16286 poly_uint64 rather than an unsigned HOST_WIDE_INT.
16287 (vect_analyze_data_refs): Change min_vf from an int * to a
16289 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
16290 vfm1 as a poly_uint64 rather than an int. Make the same change
16291 for the returned bound_scalar.
16292 (vect_gen_vector_loop_niters): Handle polynomial vf.
16293 (vect_do_peeling): Likewise. Update call to
16294 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
16295 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
16297 * tree-vect-loop.c (vect_determine_vectorization_factor)
16298 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
16299 (vect_get_known_peeling_cost): Likewise.
16300 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
16301 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
16302 (vect_transform_loop): Likewise. Use the lowest possible VF when
16303 updating the upper bounds of the loop.
16304 (vect_min_worthwhile_factor): Make static. Return an unsigned int
16305 rather than an int.
16306 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
16307 polynomial unroll factors.
16308 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
16309 (vect_make_slp_decision): Likewise.
16310 (vect_supported_load_permutation_p): Likewise, and polynomial
16312 (vect_analyze_slp_cost): Handle polynomial vf.
16313 (vect_slp_analyze_node_operations): Likewise.
16314 (vect_slp_analyze_bb_1): Likewise.
16315 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
16316 than an unsigned HOST_WIDE_INT.
16317 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
16318 (vectorizable_load): Handle polynomial vf.
16319 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
16321 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
16323 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16324 Alan Hayward <alan.hayward@arm.com>
16325 David Sherwood <david.sherwood@arm.com>
16327 * match.pd: Handle bit operations involving three constants
16328 and try to fold one pair.
16330 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16332 * tree-vect-loop-manip.c: Include gimple-fold.h.
16333 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
16334 niters_maybe_zero parameters. Handle other cases besides a step of 1.
16335 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
16336 Add a path that uses a step of VF instead of 1, but disable it
16338 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
16339 and niters_no_overflow parameters. Update calls to
16340 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
16341 Create a new SSA name if the latter choses to use a ste other
16342 than zero, and return it via niters_vector_mult_vf_var.
16343 * tree-vect-loop.c (vect_transform_loop): Update calls to
16344 vect_do_peeling, vect_gen_vector_loop_niters and
16345 slpeel_make_loop_iterate_ntimes.
16346 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
16347 (vect_gen_vector_loop_niters): Update declarations after above changes.
16349 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
16351 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
16352 128-bit round to integer instructions.
16353 (ceil<mode>2): Likewise.
16354 (btrunc<mode>2): Likewise.
16355 (round<mode>2): Likewise.
16357 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16359 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
16360 unaligned VSX load/store on P8/P9.
16361 (expand_block_clear): Allow the use of unaligned VSX
16362 load/store on P8/P9.
16364 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16366 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
16368 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
16369 swap associated with both a load and a store.
16371 2018-01-02 Andrew Waterman <andrew@sifive.com>
16373 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
16374 * config/riscv/riscv.md (clear_cache): Use it.
16376 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
16378 * web.c: Remove out-of-date comment.
16380 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16382 * expr.c (fixup_args_size_notes): Check that any existing
16383 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
16384 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
16385 (emit_single_push_insn): ...here.
16387 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16389 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
16390 (const_vector_encoded_nelts): New function.
16391 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
16392 (const_vector_int_elt, const_vector_elt): Declare.
16393 * emit-rtl.c (const_vector_int_elt_1): New function.
16394 (const_vector_elt): Likewise.
16395 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
16396 of CONST_VECTOR_ELT.
16398 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16400 * expr.c: Include rtx-vector-builder.h.
16401 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
16402 directly on the tree encoding.
16403 (const_vector_from_tree): Likewise.
16404 * optabs.c: Include rtx-vector-builder.h.
16405 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
16406 sequence of "u" values.
16407 * vec-perm-indices.c: Include rtx-vector-builder.h.
16408 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
16409 directly on the vec_perm_indices encoding.
16411 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16413 * doc/rtl.texi (const_vector): Describe new encoding scheme.
16414 * Makefile.in (OBJS): Add rtx-vector-builder.o.
16415 * rtx-vector-builder.h: New file.
16416 * rtx-vector-builder.c: Likewise.
16417 * rtl.h (rtx_def::u2): Add a const_vector field.
16418 (CONST_VECTOR_NPATTERNS): New macro.
16419 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
16420 (CONST_VECTOR_DUPLICATE_P): Likewise.
16421 (CONST_VECTOR_STEPPED_P): Likewise.
16422 (CONST_VECTOR_ENCODED_ELT): Likewise.
16423 (const_vec_duplicate_p): Check for a duplicated vector encoding.
16424 (unwrap_const_vec_duplicate): Likewise.
16425 (const_vec_series_p): Check for a non-duplicated vector encoding.
16426 Say that the function only returns true for integer vectors.
16427 * emit-rtl.c: Include rtx-vector-builder.h.
16428 (gen_const_vec_duplicate_1): Delete.
16429 (gen_const_vector): Call gen_const_vec_duplicate instead of
16430 gen_const_vec_duplicate_1.
16431 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
16432 (gen_const_vec_duplicate): Use rtx_vector_builder.
16433 (gen_const_vec_series): Likewise.
16434 (gen_rtx_CONST_VECTOR): Likewise.
16435 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
16436 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
16437 Build a new vector rather than modifying a CONST_VECTOR in-place.
16438 (handle_special_swappables): Update call accordingly.
16439 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
16440 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
16441 Build a new vector rather than modifying a CONST_VECTOR in-place.
16442 (handle_special_swappables): Update call accordingly.
16444 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16446 * simplify-rtx.c (simplify_const_binary_operation): Use
16447 CONST_VECTOR_ELT instead of XVECEXP.
16449 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16451 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
16452 the selector elements to be different from the data elements
16453 if the selector is a VECTOR_CST.
16454 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
16455 ssizetype for the selector.
16457 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16459 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
16460 before testing each element individually.
16461 * tree-vect-generic.c (lower_vec_perm): Likewise.
16463 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16465 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
16466 * selftest-run-tests.c (selftest::run_tests): Call it.
16467 * vector-builder.h (vector_builder::operator ==): New function.
16468 (vector_builder::operator !=): Likewise.
16469 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
16470 (vec_perm_indices::all_from_input_p): New function.
16471 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
16472 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
16473 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
16474 instead of reading the VECTOR_CST directly. Detect whether both
16475 vector inputs are the same before constructing the vec_perm_indices,
16476 and update the number of inputs argument accordingly. Use the
16477 utility functions added above. Only construct sel2 if we need to.
16479 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16481 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
16482 the broadcast of the low byte.
16483 (expand_mult_highpart): Use an explicit encoding for the permutes.
16484 * optabs-query.c (can_mult_highpart_p): Likewise.
16485 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
16486 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16487 (vectorizable_bswap): Likewise.
16488 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
16489 explicit encoding for the power-of-2 permutes.
16490 (vect_permute_store_chain): Likewise.
16491 (vect_grouped_load_supported): Likewise.
16492 (vect_permute_load_chain): Likewise.
16494 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16496 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
16497 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
16498 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
16499 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
16500 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
16501 (vect_gen_perm_mask_any): Likewise.
16503 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16505 * int-vector-builder.h: New file.
16506 * vec-perm-indices.h: Include int-vector-builder.h.
16507 (vec_perm_indices): Redefine as an int_vector_builder.
16508 (auto_vec_perm_indices): Delete.
16509 (vec_perm_builder): Redefine as a stand-alone class.
16510 (vec_perm_indices::vec_perm_indices): New function.
16511 (vec_perm_indices::clamp): Likewise.
16512 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
16513 (vec_perm_indices::new_vector): New function.
16514 (vec_perm_indices::new_expanded_vector): Update for new
16515 vec_perm_indices class.
16516 (vec_perm_indices::rotate_inputs): New function.
16517 (vec_perm_indices::all_in_range_p): Operate directly on the
16518 encoded form, without computing elided elements.
16519 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
16520 encoding. Update for new vec_perm_indices class.
16521 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
16522 the given vec_perm_builder.
16523 (expand_vec_perm_var): Update vec_perm_builder constructor.
16524 (expand_mult_highpart): Use vec_perm_builder instead of
16525 auto_vec_perm_indices.
16526 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
16527 vec_perm_indices instead of auto_vec_perm_indices. Use a single
16528 or double series encoding as appropriate.
16529 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
16530 vec_perm_indices instead of auto_vec_perm_indices.
16531 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16532 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16533 (vect_permute_store_chain): Likewise.
16534 (vect_grouped_load_supported): Likewise.
16535 (vect_permute_load_chain): Likewise.
16536 (vect_shift_permute_load_chain): Likewise.
16537 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
16538 (vect_transform_slp_perm_load): Likewise.
16539 (vect_schedule_slp_instance): Likewise.
16540 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16541 (vectorizable_mask_load_store): Likewise.
16542 (vectorizable_bswap): Likewise.
16543 (vectorizable_store): Likewise.
16544 (vectorizable_load): Likewise.
16545 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
16546 vec_perm_indices instead of auto_vec_perm_indices. Use
16547 tree_to_vec_perm_builder to read the vector from a tree.
16548 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
16549 vec_perm_builder instead of a vec_perm_indices.
16550 (have_whole_vector_shift): Use vec_perm_builder and
16551 vec_perm_indices instead of auto_vec_perm_indices. Leave the
16552 truncation to calc_vec_perm_mask_for_shift.
16553 (vect_create_epilog_for_reduction): Likewise.
16554 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
16555 from auto_vec_perm_indices to vec_perm_indices.
16556 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
16557 instead of changing individual elements.
16558 (aarch64_vectorize_vec_perm_const): Use new_vector to install
16559 the vector in d.perm.
16560 * config/arm/arm.c (expand_vec_perm_d::perm): Change
16561 from auto_vec_perm_indices to vec_perm_indices.
16562 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
16563 instead of changing individual elements.
16564 (arm_vectorize_vec_perm_const): Use new_vector to install
16565 the vector in d.perm.
16566 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
16567 Update vec_perm_builder constructor.
16568 (rs6000_expand_interleave): Likewise.
16569 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
16570 (rs6000_expand_interleave): Likewise.
16572 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16574 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
16575 to qimode could truncate the indices.
16576 * optabs.c (expand_vec_perm_var): Likewise.
16578 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16580 * Makefile.in (OBJS): Add vec-perm-indices.o.
16581 * vec-perm-indices.h: New file.
16582 * vec-perm-indices.c: Likewise.
16583 * target.h (vec_perm_indices): Replace with a forward class
16585 (auto_vec_perm_indices): Move to vec-perm-indices.h.
16586 * optabs.h: Include vec-perm-indices.h.
16587 (expand_vec_perm): Delete.
16588 (selector_fits_mode_p, expand_vec_perm_var): Declare.
16589 (expand_vec_perm_const): Declare.
16590 * target.def (vec_perm_const_ok): Replace with...
16591 (vec_perm_const): ...this new hook.
16592 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
16593 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
16594 * doc/tm.texi: Regenerate.
16595 * optabs.def (vec_perm_const): Delete.
16596 * doc/md.texi (vec_perm_const): Likewise.
16597 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
16598 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
16599 expand_vec_perm for constant permutation vectors. Assert that
16600 the mode of variable permutation vectors is the integer equivalent
16601 of the mode that is being permuted.
16602 * optabs-query.h (selector_fits_mode_p): Declare.
16603 * optabs-query.c: Include vec-perm-indices.h.
16604 (selector_fits_mode_p): New function.
16605 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
16606 is defined, instead of checking whether the vec_perm_const_optab
16607 exists. Use targetm.vectorize.vec_perm_const instead of
16608 targetm.vectorize.vec_perm_const_ok. Check whether the indices
16609 fit in the vector mode before using a variable permute.
16610 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
16611 vec_perm_indices instead of an rtx.
16612 (expand_vec_perm): Replace with...
16613 (expand_vec_perm_const): ...this new function. Take the selector
16614 as a vec_perm_indices rather than an rtx. Also take the mode of
16615 the selector. Update call to shift_amt_for_vec_perm_mask.
16616 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
16617 Use vec_perm_indices::new_expanded_vector to expand the original
16618 selector into bytes. Check whether the indices fit in the vector
16619 mode before using a variable permute.
16620 (expand_vec_perm_var): Make global.
16621 (expand_mult_highpart): Use expand_vec_perm_const.
16622 * fold-const.c: Includes vec-perm-indices.h.
16623 * tree-ssa-forwprop.c: Likewise.
16624 * tree-vect-data-refs.c: Likewise.
16625 * tree-vect-generic.c: Likewise.
16626 * tree-vect-loop.c: Likewise.
16627 * tree-vect-slp.c: Likewise.
16628 * tree-vect-stmts.c: Likewise.
16629 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
16631 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
16632 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
16633 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
16634 (aarch64_vectorize_vec_perm_const): ...this new function.
16635 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16636 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16637 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
16638 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
16639 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16640 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16641 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
16643 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
16644 check for NEON modes.
16645 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
16646 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
16647 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
16648 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
16650 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
16651 the old VEC_PERM_CONST conditions.
16652 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
16653 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
16654 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
16655 (ia64_vectorize_vec_perm_const_ok): Merge into...
16656 (ia64_vectorize_vec_perm_const): ...this new function.
16657 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
16658 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
16659 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
16660 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
16661 * config/mips/mips.c (mips_expand_vec_perm_const)
16662 (mips_vectorize_vec_perm_const_ok): Merge into...
16663 (mips_vectorize_vec_perm_const): ...this new function.
16664 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
16665 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
16666 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
16667 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
16668 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
16669 (rs6000_expand_vec_perm_const): Delete.
16670 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
16672 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16673 (altivec_expand_vec_perm_const_le): Take each operand individually.
16674 Operate on constant selectors rather than rtxes.
16675 (altivec_expand_vec_perm_const): Likewise. Update call to
16676 altivec_expand_vec_perm_const_le.
16677 (rs6000_expand_vec_perm_const): Delete.
16678 (rs6000_vectorize_vec_perm_const_ok): Delete.
16679 (rs6000_vectorize_vec_perm_const): New function.
16680 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
16681 an element count and rtx array.
16682 (rs6000_expand_extract_even): Update call accordingly.
16683 (rs6000_expand_interleave): Likewise.
16684 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
16685 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
16686 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
16687 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
16688 (rs6000_expand_vec_perm_const): Delete.
16689 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16690 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16691 (altivec_expand_vec_perm_const_le): Take each operand individually.
16692 Operate on constant selectors rather than rtxes.
16693 (altivec_expand_vec_perm_const): Likewise. Update call to
16694 altivec_expand_vec_perm_const_le.
16695 (rs6000_expand_vec_perm_const): Delete.
16696 (rs6000_vectorize_vec_perm_const_ok): Delete.
16697 (rs6000_vectorize_vec_perm_const): New function. Remove stray
16698 reference to the SPE evmerge intructions.
16699 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
16700 an element count and rtx array.
16701 (rs6000_expand_extract_even): Update call accordingly.
16702 (rs6000_expand_interleave): Likewise.
16703 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
16704 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
16706 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16708 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16710 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
16711 vector mode and that that mode matches the mode of the data
16713 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
16714 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
16715 directly using expand_vec_perm_1 when forcing selectors into
16717 (expand_vec_perm_var): New function, split out from expand_vec_perm.
16719 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16721 * optabs-query.h (can_vec_perm_p): Delete.
16722 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
16723 * optabs-query.c (can_vec_perm_p): Split into...
16724 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
16725 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
16726 particular selector is valid.
16727 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16728 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16729 (vect_grouped_load_supported): Likewise.
16730 (vect_shift_permute_load_chain): Likewise.
16731 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
16732 (vect_transform_slp_perm_load): Likewise.
16733 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16734 (vectorizable_bswap): Likewise.
16735 (vect_gen_perm_mask_checked): Likewise.
16736 * fold-const.c (fold_ternary_loc): Likewise. Don't take
16737 implementations of variable permutation vectors into account
16738 when deciding which selector to use.
16739 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
16740 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
16741 with a false third argument.
16742 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
16743 to test whether the constant selector is valid and can_vec_perm_var_p
16744 to test whether a variable selector is valid.
16746 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16748 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
16749 * optabs-query.c (can_vec_perm_p): Likewise.
16750 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
16751 instead of vec_perm_indices.
16752 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
16753 (vect_gen_perm_mask_checked): Likewise,
16754 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
16755 (vect_gen_perm_mask_checked): Likewise,
16757 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16759 * optabs-query.h (qimode_for_vec_perm): Declare.
16760 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
16761 (qimode_for_vec_perm): ...this new function.
16762 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
16764 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16766 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
16767 does not have a conditional at the top.
16769 2018-01-02 Richard Biener <rguenther@suse.de>
16771 * ipa-inline.c (big_speedup_p): Fix expression.
16773 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
16776 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
16779 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
16783 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
16784 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
16785 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
16786 cond_taken_branch_cost 3->4.
16788 2018-01-01 Jakub Jelinek <jakub@redhat.com>
16790 PR tree-optimization/83581
16791 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
16792 TODO_cleanup_cfg if any changes have been made.
16794 PR middle-end/83608
16795 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
16796 convert_modes if target mode has the right side, but different mode
16799 PR middle-end/83609
16800 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
16801 last argument when extracting from CONCAT. If either from_real or
16802 from_imag is NULL, use expansion through memory. If result is not
16803 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
16804 the parts directly to inner mode, if even that fails, use expansion
16807 PR middle-end/83623
16808 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
16809 check for bswap in mode rather than HImode and use that in expand_unop
16812 Copyright (C) 2018 Free Software Foundation, Inc.
16814 Copying and distribution of this file, with or without modification,
16815 are permitted in any medium without royalty provided the copyright
16816 notice and this notice are preserved.