1 2018-06-18 David Malcolm <dmalcolm@redhat.com>
3 * tree-vect-data-refs.c (vect_analyze_data_ref_dependences):
4 Replace dump_printf_loc call with DUMP_VECT_SCOPE.
5 (vect_slp_analyze_instance_dependence): Likewise.
6 (vect_enhance_data_refs_alignment): Likewise.
7 (vect_analyze_data_refs_alignment): Likewise.
8 (vect_slp_analyze_and_verify_instance_alignment
9 (vect_analyze_data_ref_accesses): Likewise.
10 (vect_prune_runtime_alias_test_list): Likewise.
11 (vect_analyze_data_refs): Likewise.
12 * tree-vect-loop-manip.c (vect_update_inits_of_drs): Likewise.
13 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
14 (vect_analyze_scalar_cycles_1): Likewise.
15 (vect_get_loop_niters): Likewise.
16 (vect_analyze_loop_form_1): Likewise.
17 (vect_update_vf_for_slp): Likewise.
18 (vect_analyze_loop_operations): Likewise.
19 (vect_analyze_loop): Likewise.
20 (vectorizable_induction): Likewise.
21 (vect_transform_loop): Likewise.
22 * tree-vect-patterns.c (vect_pattern_recog): Likewise.
23 * tree-vect-slp.c (vect_analyze_slp): Likewise.
24 (vect_make_slp_decision): Likewise.
25 (vect_detect_hybrid_slp): Likewise.
26 (vect_slp_analyze_operations): Likewise.
27 (vect_slp_bb): Likewise.
28 * tree-vect-stmts.c (vect_mark_stmts_to_be_vectorized): Likewise.
29 (vectorizable_bswap): Likewise.
30 (vectorizable_call): Likewise.
31 (vectorizable_simd_clone_call): Likewise.
32 (vectorizable_conversion): Likewise.
33 (vectorizable_assignment): Likewise.
34 (vectorizable_shift): Likewise.
35 (vectorizable_operation): Likewise.
36 * tree-vectorizer.h (DUMP_VECT_SCOPE): New macro.
38 2018-06-18 Martin Sebor <msebor@redhat.com>
40 PR tree-optimization/81384
41 * builtin-types.def (BT_FN_SIZE_CONST_STRING_SIZE): New.
42 * builtins.c (expand_builtin_strnlen): New function.
43 (expand_builtin): Call it.
44 (fold_builtin_n): Avoid setting TREE_NO_WARNING.
45 * builtins.def (BUILT_IN_STRNLEN): New.
46 * calls.c (maybe_warn_nonstring_arg): Handle BUILT_IN_STRNLEN.
47 Warn for bounds in excess of maximum object size.
48 * tree-ssa-strlen.c (maybe_set_strlen_range): Return tree representing
49 single-value ranges. Handle strnlen.
50 (handle_builtin_strlen): Handle strnlen.
51 (strlen_check_and_optimize_stmt): Same.
52 * doc/extend.texi (Other Builtins): Document strnlen.
54 2018-06-18 Maya Rashish <coypu@sdf.org>
56 * config/alpha/openbsd.h (TARGET_DEFAULT): Define.
57 (LINK_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): Likewise.
58 (INTMAX_TYPE, UINTMAX_TYPE, WINT_TYPE): Likewise.
60 * config/alpha/elf.h (STARTFILE_SPEC, ENDFILE_SPEC): Move from
62 * config/alpha/linux.h (STARTFILE_SPEC, ENDFILE_SPEC): Here.
64 2018-06-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
66 * tree.c (escaped_string::escape): Replace cast to char * by
67 const_cast<char *> (unescaped).
69 2018-06-18 Nick Clifton <nickc@redhat.com>
72 * tree.c (escaped_string): New class. Converts an unescaped
73 string into its escaped equivalent.
74 (warn_deprecated_use): Use the new class to convert the
75 deprecation message, if present.
76 (test_escaped_strings): New self test.
77 (test_c_tests): Add test_escaped_strings.
78 * doc/extend.texi (deprecated): Add a note that the
79 deprecation message is affected by the -fmessage-length
80 option, and that control characters will be escaped.
81 (#pragma GCC error): Document this pragma.
82 (#pragma GCC warning): Likewise.
83 * doc/invoke.texi (-fmessage-length): Document this option's
84 effect on the #warning and #error preprocessor directives and
85 the deprecated attribute.
87 2018-06-18 Eric Botcazou <ebotcazou@adacore.com>
89 * tree.c (decl_value_expr_lookup): Revert latest change.
90 (decl_value_expr_insert): Likewise.
92 2018-06-17 Eric Botcazou <ebotcazou@adacore.com>
94 * gimplify.c (nonlocal_vlas): Delete.
95 (nonlocal_vla_vars): Likewise.
96 (gimplify_var_or_parm_decl): Do not add debug VAR_DECLs for non-local
98 (gimplify_body): Do not create and destroy nonlocal_vlas.
99 * tree-nested.c: Include diagnostic.h.
100 (use_pointer_in_frame): Tweak.
101 (lookup_field_for_decl): Add assertion and declare the transformation.
102 (convert_nonlocal_reference_op) <PARM_DECL>: Rework and issue an
103 internal error when the reference is in a wrong context. Do not
104 create a debug decl by default.
105 (note_nonlocal_block_vlas): Delete.
106 (convert_nonlocal_reference_stmt) <GIMPLE_BIND>: Do not call it.
107 (convert_local_reference_op) <PARM_DECL>: Skip the frame decl. Do not
108 create a debug decl by default.
109 (convert_gimple_call) <GIMPLE_CALL>: Issue an internal error when the
110 call is in a wrong context.
111 (fixup_vla_decls): New function.
112 (finalize_nesting_tree_1): Adjust comment. Call fixup_vla_decls if no
113 debug variables were created.
114 * tree.c (decl_value_expr_lookup): Add checking assertion.
115 (decl_value_expr_insert): Likewise.
117 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
120 * ipa-fnsummary.c (will_be_nonconstant_expr_predicate): Handle CALL_EXPR.
121 * tree-scalar-evolution.c (interpret_expr): Likewise.
122 (expression_expensive_p): Likewise.
123 * tree-ssa-loop-ivopts.c (contains_abnormal_ssa_name_p): Likewise.
124 * tree-ssa-loop-niter.c (number_of_iterations_popcount): New.
125 (number_of_iterations_exit_assumptions): Use number_of_iterations_popcount.
126 (ssa_defined_by_minus_one_stmt_p): New.
128 2018-06-16 Kugan Vivekanandarajah <kuganv@linaro.org>
131 * cfgexpand.c (expand_debug_expr): Hande ABSU_EXPR.
132 * config/i386/i386.c (ix86_add_stmt_cost): Likewise.
133 * dojump.c (do_jump): Likewise.
134 * expr.c (expand_expr_real_2): Check operand type's sign.
135 * fold-const.c (const_unop): Handle ABSU_EXPR.
136 (fold_abs_const): Likewise.
137 * gimple-pretty-print.c (dump_unary_rhs): Likewise.
138 * gimple-ssa-backprop.c (backprop::process_assign_use): Likesie.
139 (strip_sign_op_1): Likesise.
140 * match.pd: Add new pattern to generate ABSU_EXPR.
141 * optabs-tree.c (optab_for_tree_code): Handle ABSU_EXPR.
142 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
143 * tree-eh.c (operation_could_trap_helper_p): Likewise.
144 * tree-inline.c (estimate_operator_cost): Likewise.
145 * tree-pretty-print.c (dump_generic_node): Likewise.
146 * tree-vect-patterns.c (vect_recog_sad_pattern): Likewise.
147 * tree.def (ABSU_EXPR): New.
149 2018-06-16 Jakub Jelinek <jakub@redhat.com>
152 * common.opt (Wunsafe-loop-optimizations): Add Ignore, remove Var,
153 documented as preserved for backward compatibility only.
154 * doc/invoke.texi: Remove -Wunsafe-loop-optimizations documentation.
156 PR rtl-optimization/86108
157 * bb-reorder.c (create_forwarder_block): Renamed to ...
158 (create_eh_forwarder_block): ... this. Split OLD_BB after labels and
159 jump from new landing pad to the second part.
160 (sjlj_fix_up_crossing_landing_pad, dw2_fix_up_crossing_landing_pad):
163 2018-06-15 Jakub Jelinek <jakub@redhat.com>
166 * expr.c (expand_assignment): Remove now redundant COMPLEX_MODE_P
167 check from first store_expr, use to_mode instead of GET_MODE (to_rtx).
168 Only call store_expr for halves if the mode is the same.
171 * match.pd ((X / Y) == 0 -> X < Y): Don't transform complex divisions.
172 Fix up comment formatting.
174 2018-06-15 Bernd Edlinger <bernd.edlinger@hotmail.de>
176 * typed-splay-tree.h (typed_splay_tree::remove): New function.
177 (typed_splay_tree::closure,
178 typed_splay_tree::inner_foreach_fn, typed_splay_tree::m_inner): Deleted.
179 (typed_splay_tree::typed_splay_tree,
180 typed_splay_tree::operator =): Declared private.
181 (typed_splay_tree::splay_tree_key, typed_splay_tree::splay_tree_value,
182 typed_splay_tree::splay_tree_node_s, typed_splay_tree::KDEL,
183 typed_splay_tree::VDEL, typed_splay_tree::splay_tree_delete_helper,
184 typed_splay_tree::rotate_left, typed_splay_tree::rotate_right,
185 typed_splay_tree::splay_tree_splay,
186 typed_splay_tree::splay_tree_foreach_helper,
187 typed_splay_tree::splay_tree_insert,
188 typed_splay_tree::splay_tree_remove,
189 typed_splay_tree::splay_tree_lookup,
190 typed_splay_tree::splay_tree_predecessor,
191 typed_splay_tree::splay_tree_successor,
192 typed_splay_tree::splay_tree_min,
193 typed_splay_tree::splay_tree_max): Took over from splay-tree.c/.h.
194 (typed_splay_tree::root, typed_splay_tree::comp,
195 typed_splay_tree::delete_key,
196 typed_splay_tree::delete_value): New data members.
197 * typed-splay-tree.c (selftest::test_str_to_int): Add a test for
198 typed_splay_tree::remove.
200 2018-06-15 Matthew Fortune <matthew.fortune@mips.com>
202 * config/mips/mips.h (ASM_SPEC): Pass through -mcrc, -mno-crc,
203 -mginv and -mno-ginv to the assembler.
204 * config/mips/mips.opt (-mcrc): New option.
206 * doc/invoke.text (-mcrc): Document.
209 2018-06-15 Nick Clifton <nickc@redhat.com>
212 * tree.c (escaped_string): New class. Converts an unescaped
213 string into its escaped equivalent.
214 (warn_deprecated_use): Use the new class to convert the
215 deprecation message, if present.
216 (test_escaped_strings): New self test.
217 (test_c_tests): Add test_escaped_strings.
218 * doc/extend.texi (deprecated): Add a note that the
219 deprecation message is affected by the -fmessage-length
220 option, and that control characters will be escaped.
221 (#pragma GCC error): Document this pragma.
222 (#pragma GCC warning): Likewise.
223 * doc/invoke.texi (-fmessage-length): Document this option's
224 effect on the #warning and #error preprocessor directives and
225 the deprecated attribute.
227 2018-06-15 Richard Biener <rguenther@suse.de>
229 * tree-vect-slp.c (vect_slp_bb): Dump MSG_OPTIMIZED_LOCATIONS
230 here, also noting vector size used.
231 * tree-vectorizer.c (vectorize_loops): Adjust. Note vector
232 size used in MSG_OPTIMIZED_LOCATIONS dump.
233 (pass_slp_vectorize::execute): Adjust.
235 2018-06-15 Claudiu Zissulescu <claziss@synopsys.com>
238 * config/arc/arc.c (arc_return_address_register): Fix
241 2018-06-15 Richard Biener <rguenther@suse.de>
244 * tree-cfg.c (gimplify_build3): Do not strip sign conversions,
245 leave useless conversion stripping to force_gimple_operand_gsi.
246 (gimplify_build2): Likewise.
247 (gimplify_build1): Likewise.
249 2018-06-15 Richard Biener <rguenther@suse.de>
252 * tree-cfg.c (move_stmt_op): unshare invariant addresses
253 before adjusting their block.
255 2018-06-15 Sebastian Huber <sebastian.huber@embedded-brains.de>
257 * config.gcc (riscv*-*-elf* | riscv*-*-rtems*): Use custom
258 multilibs for *-*-rtems*.
259 * config/riscv/t-rtems: New file.
261 2018-06-14 Jakub Jelinek <jakub@redhat.com>
264 * match.pd ((A +- CST1) +- CST2): Punt if last resort
265 unsigned_type_for returns NULL.
268 * lower-subreg.c (find_decomposable_subregs): Don't decompose float
269 subregs of multi-word pseudos unless the float mode has word size.
271 2018-06-14 Richard Biener <rguenther@suse.de>
274 * tree-vect-generic.c (build_word_mode_vector_type): Remove
275 duplicate and harmful type_hash_canon.
276 * tree.c (type_hash_canon): Assert we didn't find ourselves.
278 2018-06-14 Richard Biener <rguenther@suse.de>
281 * tree-ssa-struct-alias.c (create_variable_info_for): Handle
284 2018-06-14 Sebastian Huber <sebastian.huber@embedded-brains.de>
286 * config/rtems.h (STDINT_LONG32): Define.
288 2018-06-13 Matthew Fortune <matthew.fortune@mips.com>
289 Prachi Godbole <prachi.godbole@imgtec.com>
291 * config/mips/mips-cpus.def: Define P6600.
292 * config/mips/mips-tables.opt: Regenerate.
293 * config/mips/mips.c (mips_ucbranch_type): New enum.
294 (mips_rtx_cost_data): Add support for P6600.
295 (mips_issue_rate): Likewise.
296 (mips_multipass_dfa_lookahead): Likewise.
297 (mips_avoid_hazard): Likewise.
298 (mips_reorg_process_insns): Likewise.
299 (mips_classify_branch_p6600): New function.
300 * config/mips/mips.h (TUNE_P6600): New define.
301 (MIPS_ISA_LEVEL_SPEC): Infer mips64r6 from p6600.
302 (ENABLE_LD_ST_PAIRS): Enable load/store bonding for p6600.
303 * config/mips/mips.md: Include p6600.md.
304 (processor): Add p6600.
305 * config/mips/p6600.md: New file.
306 * doc/invoke.texi: Add p6600 to supported architectures.
308 2018-06-13 Martin Sebor <msebor@redhat.com>
310 PR tree-optimization/86114
311 * gimple-fold.c (gimple_fold_builtin_strlen): Only handle LHS
313 * tree-ssa-strlen.c (maybe_set_strlen_range): Same.
315 2018-06-13 Richard Biener <rguenther@suse.de>
317 * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern):
318 Properly set vector type of the intermediate stmt.
319 * tree-vect-stmts.c (vectorizable_operation): The destination
320 var always has vectype_out type.
322 2018-06-13 Jeff Law <law@redhat.com>
324 * config/rl78/rl78.c (move_elim_pass): Use TDF_NONE rather than
325 integer 0 for argument to print_rtl_with_bb.
326 (rl78_reorg): Likewise.
328 2018-06-13 David Malcolm <dmalcolm@redhat.com>
330 * config/arc/arc.c (hwloop_optimize): Strengthen local "end_label"
331 from rtx to rtx_insn *.
332 * config/bfin/bfin.c (hwloop_optimize): Likewise for local
334 (add_sched_insns_for_speculation): Likewise for local "target",
335 converting usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
336 * config/c6x/c6x.c (reorg_split_calls): Strengthen param "call_labels"
337 from rtx_insn ** to rtx_code_label **.
338 (reorg_emit_nops): Likewise.
339 (c6x_reorg): Likewise for local "call_labels".
340 * config/sh/sh-protos.h (get_dest_uid): Strengthen 1st param from
342 * config/sh/sh.c (dump_table): Strengthen local "lab" from rtx to
343 rtx_code_label *, adding safe_as_a <rtx_code_label *> casts to
344 the loops over LABEL_REFS.
345 (fixup_addr_diff_vecs): Add as_a <rtx_insn *> to usage of
347 (barrier_align): Convert usage of JUMP_LABEL to JUMP_LABEL_AS_INSN.
348 (get_dest_uid): Strengthen param "label" from rtx to rtx_insn *.
349 (split_branches): Strengthen local "olabel" from rtx to
350 rtx_insn *, adding a safe_as_a cast.
351 * emit-rtl.c (next_real_insn): Strengthen param from "rtx"
353 (add_insn_after): Likewise for first two params.
354 (add_insn_before): Likewise.
355 (remove_insn): Likewise for param.
356 (emit_pattern_before_noloc): Likewise for second and third params.
357 (emit_jump_insn_before_noloc): Convert NULL_RTX to NULL.
358 (emit_call_insn_before_noloc): Likewise.
359 (emit_debug_insn_before_noloc): Strengthen "before" param from "rtx"
361 (emit_barrier_before): Likewise.
362 (emit_label_before): Strengthen "label" param from "rtx" to
363 "rtx_code_label *". Strengthen "before" param from "rtx" to
365 (emit_insn_after_1): Strengthen "after" param from "rtx" to
367 (emit_pattern_after_noloc): Likewise.
368 (emit_insn_after_noloc): Likewise.
369 (emit_jump_insn_after_noloc): Likewise.
370 (emit_call_insn_after_noloc): Likewise.
371 (emit_debug_insn_after_noloc): Likewise.
372 (emit_barrier_after): Likewise.
373 (emit_label_after): Likewise for both params.
374 (emit_pattern_after_setloc): Likewise for "after" param. Convert
375 "loc" param from "int" to "location_t".
376 (emit_insn_after_setloc): Likewise.
377 (emit_jump_insn_after_setloc): Likewise.
378 (emit_call_insn_after_setloc): Likewise.
379 (emit_debug_insn_after_setloc): Likewise.
380 (emit_pattern_before_setloc): Likewise for "before" param. Convert
381 "loc" param from "int" to "location_t".
382 (emit_pattern_before): Convert NULL_RTX to NULL.
383 (emit_insn_before_setloc): Convert "loc" param from "int" to
385 (emit_jump_insn_before_setloc): Likewise.
386 (emit_call_insn_before_setloc): Likewise.
387 (emit_debug_insn_before_setloc): Strengthen "before" param from rtx to
388 rtx_insn *. Convert "loc" param from "int" to "location_t".
389 * rtl.h (emit_insn_before_setloc, emit_jump_insn_before_setloc,
390 emit_call_insn_before_setloc, emit_debug_insn_before_setloc):
391 Convert 3rd param from "int" to "location_t".
392 (emit_barrier_before, emit_barrier_after, next_real_insn):
393 Strengthen param from rtx to rtx_insn *.
394 (emit_label_before): Strengthen 1st param from "rtx" to
395 "rtx_code_label *". Strengthen 2nd param from "rtx" to
397 (emit_insn_after_noloc, emit_jump_insn_after_noloc,
398 emit_call_insn_after_noloc, emit_debug_insn_after_noloc):
399 Strengthen 2nd param from "rtx" to "rtx_insn *".
400 (emit_insn_after_setloc, emit_jump_insn_after_setloc)
401 emit_call_insn_after_setloc, emit_debug_insn_after_setloc):
402 Likewise. Convert 3rd param from "int" to "location_t".
403 (emit_label_after): Strengthen 1st param from "rtx" to
405 (next_real_insn, remove_insn): Strengthen param from "rtx" to
407 (add_insn_before, add_insn_after): Strengthen 1st and 2nd params
408 from "rtx" to "rtx_insn *".
410 2018-06-13 Jan Hubicka <hubicka@gcc.gnu.org>
412 * cgraph.c (cgraph_node::get_untransformed_body): Dump function
413 bodies streamed in with -Q.
414 * dumpfile.c (dump_files): Add lto-stream-out dump file.
415 * dumpfile.h (tree_dump_index): Add lto_stream_out.
416 * gimple-streamer-out.c: Include gimple-pretty-print.h
417 (output_bb): Dump stmts streamed.
418 * lto-section-out.c: Include print-tree.h
419 (lto_begin_section): Dump sections created.
420 (lto_output_decl_index): Dump decl encoded.
421 * lto-streamer-out.c: Include print-tree.h
422 (create_output_block): Dump output block created.
423 (DFS::DFS_write_tree_body): Dump DFS SCCs streamed.
424 (output_function): Dump function output.
425 (output_constructor): Dump constructor streamed.
426 (write_global_stream): Output indexes encoded.
427 (produce_asm_for_decls): Dump streams encoded.
428 * lto-streamer.c (streamer_dump_file): New global var.
429 * lto-streamer.h (streamer_dump_file): Declare.
430 * passes.c (ipa_write_summaries): Initialize streamer dump.
431 * varpool.c (varpool_node::get_constructor): Dump constructors streamed
434 2018-06-13 Eric Botcazou <ebotcazou@adacore.com>
437 * config/i386/winnt.c (i386_pe_seh_cold_init): Do not emit negative
438 offsets for register save directives. Emit a second batch of save
439 directives, if need be, when the function accesses prior frames.
441 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
443 * config/arc/fpu.md (fmasf4): Force operand to register.
446 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
448 * config/arc/arc-protos.h (arc_pad_return): Remove.
449 * config/arc/arc.c (machine_function): Remove force_short_suffix
451 (arc_print_operand): Adjust printing of '&'.
452 (arc_verify_short): Remove conditional printing of short suffix.
453 (arc_final_prescan_insn): Remove reference to size_reason.
454 (pad_return): New function.
455 (arc_reorg): Call pad_return.
456 (arc_pad_return): Remove.
457 (arc_init_machine_status): Remove reference to force_short_suffix.
458 * config/arc/arc.md (vunspec): Add VUNSPEC_ARC_BLOCKAGE.
459 (attr length): When attribute iscompact is true force to 2
460 regardless; in the case of maybe check if we want to force the
461 instruction to have 4 bytes length.
462 (nopv): Change it to generate 4 byte long nop as well.
463 (blockage): New pattern.
464 (simple_return): Remove call to arc_pad_return.
465 (p_return_i): Likewise.
467 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
469 * config/arc/elf.h (LINK_GCC_C_SEQUENCE_SPEC): Define.
471 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
473 * config/arc/builtins.def (SYNC): SYNC instruction is valid on all
476 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
478 * config/arc/arc.c (atomic_exchangesi): EX instruction is default
479 for ARC700 and ARCv2.
481 2018-06-13 Chenghua Xu <paul.hua.gm@gmail.com>
484 * config/mips/loongson.md (vec_setv4hi): Gen_lowpart for
485 operands[2] instead of operands[1].
488 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
490 * lra-constraints.c (simplify_operand_subreg): In the paradoxical
491 case, check whether the outer register overlaps an unallocatable
492 register, not just whether it fits the required class.
494 2018-06-12 Richard Sandiford <richard.sandiford@linaro.org>
496 * poly-int.h (can_div_trunc_p): Add new overload in which all values
498 * alias.c (get_addr): Extend CONST_INT handling to poly_int_rtx_p.
499 (memrefs_conflict_p): Likewise.
500 (init_alias_analysis): Likewise.
501 * cfgexpand.c (expand_debug_expr): Likewise.
502 * combine.c (combine_simplify_rtx, force_int_to_mode): Likewise.
503 * cse.c (fold_rtx): Likewise.
504 * explow.c (adjust_stack, anti_adjust_stack): Likewise.
505 * expr.c (emit_block_move_hints): Likewise.
506 (clear_storage_hints, push_block, emit_push_insn): Likewise.
507 (store_expr_with_bounds, reduce_to_bit_field_precision): Likewise.
508 (emit_group_load_1): Use rtx_to_poly_int64 for group offsets.
509 (emit_group_store): Likewise.
510 (find_args_size_adjust): Use strip_offset. Use rtx_to_poly_int64
511 to read the PRE/POST_MODIFY increment.
512 * calls.c (store_one_arg): Use strip_offset.
513 * rtlanal.c (rtx_addr_can_trap_p_1): Extend CONST_INT handling to
515 (set_noop_p): Use rtx_to_poly_int64 for the elements selected
517 * simplify-rtx.c (avoid_constant_pool_reference): Use strip_offset.
518 (simplify_binary_operation_1): Extend CONST_INT handling to
520 * var-tracking.c (compute_cfa_pointer): Take a poly_int64 rather
521 than a HOST_WIDE_INT.
522 (hard_frame_pointer_adjustment): Change from HOST_WIDE_INT to
524 (adjust_mems, add_stores): Update accodingly.
525 (vt_canonicalize_addr): Track polynomial offsets.
526 (emit_note_insn_var_location): Likewise.
527 (vt_add_function_parameter): Likewise.
528 (vt_initialize): Likewise.
530 2018-06-12 Jeff Law <law@redhat.com>
532 * config.gcc (alpha*-*-freebsd*): Remove.
533 * config/alpha/freebsd.h: Remove.
535 2018-06-12 David Malcolm <dmalcolm@redhat.com>
538 * spellcheck-tree.c (levenshtein_distance): Rename to...
539 (get_edit_distance): ...this, and update for underlying renaming.
540 * spellcheck-tree.h (levenshtein_distance): Rename to...
541 (get_edit_distance): ...this.
542 * spellcheck.c (levenshtein_distance): Rename to...
543 (get_edit_distance): ...this. Convert from Levenshtein distance
544 to Damerau-Levenshtein distance by supporting transpositions of
545 adjacent characters. Rename "v1" to "v_next" and "v0" to
547 (selftest::levenshtein_distance_unit_test_oneway): Rename to...
548 (selftest::test_edit_distance_unit_test_oneway): ...this, and
549 update for underlying renaming.
550 (selftest::levenshtein_distance_unit_test): Rename to...
551 (selftest::test_get_edit_distance_unit): ...this, and update for
553 (selftest::test_find_closest_string): Add example from PR 69968
554 where transposition helps
555 (selftest::test_metric_conditions): Update for renaming.
556 (selftest::test_metric_conditions): Likewise.
557 (selftest::spellcheck_c_tests): Likewise.
558 * spellcheck.h (levenshtein_distance): Rename both overloads to...
559 (get_edit_distance): ...this.
560 (best_match::consider): Update for renaming.
562 2018-06-12 Martin Sebor <msebor@redhat.com>
564 PR tree-optimization/85259
565 * builtins.c (compute_objsize): Handle constant offsets.
566 * gimple-ssa-warn-restrict.c (maybe_diag_offset_bounds): Return
567 true iff a warning has been issued.
568 * gimple.h (gimple_nonartificial_location): New function.
569 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Call
570 gimple_nonartificial_location and handle -Wno-system-headers.
571 (handle_builtin_stxncpy): Same.
573 2018-06-12 Martin Sebor <msebor@redhat.com>
576 * fold-const.c (operand_equal_p): Handle SAVE_EXPR.
578 2018-06-12 Will Schmidt <will_schmidt@vnet.ibm.com>
580 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
581 BUILTIN_VEC_XST entries for pointer to double and long long.
583 2018-06-12 H.J. Lu <hongjiu.lu@intel.com>
586 * config/i386/gnu-user.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
588 * config/i386/gnu-user64.h (TARGET_THREAD_SPLIT_STACK_OFFSET):
591 2018-06-12 Martin Liska <mliska@suse.cz>
593 * doc/options.texi: Document IntegerRange.
595 2018-06-12 Martin Liska <mliska@suse.cz>
597 * config/i386/i386.opt: Make MPX-related options as Deprecated.
598 * opt-functions.awk: Handle Deprecated flag.
599 * opts-common.c (decode_cmdline_option): Handle cl_deprecated
601 (read_cmdline_option): Report warning for a deprecated option.
602 * opts.h (struct cl_option): Add new field cl_deprecated.
603 (CL_ERR_DEPRECATED): New.
605 2018-06-12 Martin Liska <mliska@suse.cz>
607 * doc/options.texi: Document Deprecated option flag.
609 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
611 * config/arc/arc-arch.h (arc_extras): New enum.
612 (arc_cpu_t):Add field extra.
613 (arc_cpu_types): Consider the extras.
614 * config/arc/arc-cpus.def: Add extras info.
615 * config/arc/arc-opts.h (processor_type): Consider extra field.
616 * config/arc/arc.c (arc_override_options): Handle extra field.
618 2018-06-12 Claudiu Zissulescu <claziss@synopsys.com>
620 * config/arc/arc-arch.h: Update ARC_OPTX macro.
621 * config/arc/arc-options.def (ARC_OPTX): Introduce a new doc
623 * config/arc/arc.c (arc_init): Update pic warning.
624 (irq_range): Update irq range parsing warnings.
625 (arc_override_options): Update various warning messages.
626 (arc_handle_aux_attribute): Likewise.
628 2018-06-12 Robert Suchanek <robert.suchanek@mips.com>
630 * config/mips/i6400.md (i6400_fpu_fadd): Remove frint.
632 2018-06-12 Jozef Lawrynowicz <jozef.l@mittosystems.com>
634 * doc/sourcebuild.texi: Document usage of line number 0 in verify
635 compiler messages directives.
637 2018-06-12 Matthew Fortune <mfortune@gmail.com>
639 * config/mips/mips-cpus.def: New MIPS_CPU for i6500.
640 * config/mips/mips-tables.opt: Regenerate.
641 * config/mips/mips.h (MIPS_ISA_LEVEL_SPEC): Mark i6500 as
643 * doc/invoke.texi: Document -march=i6500.
645 2018-06-12 Prachi Godbole <prachi.godbole@imgtec.com>
647 * config/mips/i6400.md (i6400_gpmuldiv): Remove cpu_unit.
648 (i6400_gpmul): Add cpu_unit.
649 (i6400_gpdiv): Likewise.
650 (i6400_msa_add_d): Update reservations.
651 (i6400_msa_int_add) Likewise.
652 (i6400_msa_short_logic3) Likewise.
653 (i6400_msa_short_logic2) Likewise.
654 (i6400_msa_short_logic) Likewise.
655 (i6400_msa_move) Likewise.
656 (i6400_msa_cmp) Likewise.
657 (i6400_msa_short_float2) Likewise.
658 (i6400_msa_div_d) Likewise.
659 (i6400_msa_long_logic1) Likewise.
660 (i6400_msa_long_logic2) Likewise.
661 (i6400_msa_mult) Likewise.
662 (i6400_msa_long_float2) Likewise.
663 (i6400_msa_long_float4) Likewise.
664 (i6400_msa_long_float5) Likewise.
665 (i6400_msa_long_float8) Likewise.
666 (i6400_fpu_fadd): Include frint type.
667 (i6400_fpu_store): New define_insn_reservation.
668 (i6400_fpu_load): Likewise.
669 (i6400_fpu_move): Likewise.
670 (i6400_fpu_fcmp): Likewise.
671 (i6400_fpu_fmadd): Likewise.
672 (i6400_int_mult): Include imul3nc type and update reservation.
673 (i6400_int_div): Include idiv3 type and update reservation.
674 (i6400_int_load): Update to check type not move_type.
675 (i6400_int_store): Likewise.
676 (i6400_int_prefetch): Set zero latency.
678 2018-06-12 Eric Botcazou <ebotcazou@adacore.com>
680 * gcc.c: Document new %@{...} sequence.
681 (LINK_COMMAND_SPEC): Use it for the -L switches.
682 (cpp_unique_options): Use it for the -I switches.
683 (at_file_argbuf): New global variable.
684 (in_at_file): Likewise.
685 (alloc_args): Create at_file_argbuf.
686 (clear_args): Truncate at_file_argbuf.
687 (store_arg): If in_at_file, push the argument onto at_file_argbuf.
688 (open_at_file): New function.
689 (close_at_file): Likewise.
690 (create_at_file): Delete.
691 (do_spec_1) <'i'>: Use open_at_file/close_at_file.
694 (validate_switches_from_spec): Deal with %@{...} sequence.
695 (validate_switches): Likewise.
696 (driver::finalize): Call clear_args.
698 2018-06-11 Rasmus Villemoes <rasmus.villemoes@prevas.dk>
700 * config/vx-common.h (USE_TM_CLONE_REGISTRY): #define to 0.
702 2018-06-11 Martin Sebor <msebor@redhat.com>
704 * doc/invoke.texi (-Wall): List -Wc++17-compat.
705 (Wno-class-memaccess): Add @opindex.
706 (Wno-templates, Wno-multiple-inheritance): Same.
707 (Wno-virtual-inheritance, Wno-namespaces): Same.
708 (Wno-pedantic, Wno-chkp, Wcoverage-mismatch): Same.
709 (Wno-format-overflow, Wno-format-truncation): Same.
710 (Wno-unused-local-typedefs, Walloc-size-larger-than): Same.
711 (Wno-alloc-size-larger-than, Wframe-larger-than): Same
712 (Wno-stack-usage, Wno-c++-compat, Wno-c++11-compat): Same.
713 (Wno-c++14-compat, Wno-c++17-compat, Wno-openmp-simd): Same.
714 (Wno-unsuffixed-float-constants, Wno-addr-space-convert): Same.
715 (Wno-misspelled-isr): Same.
717 2018-06-11 Martin Sebor <msebor@redhat.com>
719 * PR tree-optimization/86083
720 * tree-ssa-strlen.c (handle_char_store): Use tree_expr_nonzero_p.
722 2018-06-11 Zhouyi Zhou <zhouzhouyi@gmail.com>
724 * tree-eh.c (lower_eh_constructs_2): Add a comma to comment.
726 2018-06-11 Segher Boessenkool <segher@kernel.crashing.org>
729 * config/rs6000/rs6000.md (*movdi_internal32): Put constraint modifiers
730 on the correct operand.
731 (*movdi_internal64): Ditto.
733 2018-06-11 Martin Liska <mliska@suse.cz>
735 PR tree-optimization/86089
736 * tree-ssa-strlen.c (get_string_length): Move back removed hunk.
738 2018-06-11 Julia Koval <julia.koval@intel.com>
740 * config/i386/vaesintrin.h (_mm_aesdec_epi128, _mm_aesdeclast_epi128)
741 _mm_aesenc_epi128, _mm_aesenclast_epi128: Remove.
742 * config/i386/vpclmulqdqintrin.h (_mm_clmulepi64_epi128): Remove.
744 2018-06-11 Olivier Hainque <hainque@adacore.com>
746 * dwarf2out.c (gen_compile_unit_die): Fallout to DW_LANG_Ada83
747 for Ada with strict dwarf2.
749 2018-06-08 Peter Bergner <bergner@vnet.ibm.com>
752 * config/rs6000/rs6000.c (mem_operand_gpr): Enable PRE_INC and PRE_DEC
755 2018-06-08 Jan Hubicka <hubicka@ucw.cz>
757 * dumpfile.c (FIRST_ME_AUTO_NUMBERED_DUMP): Bump to 4.
759 2018-06-08 David Edelsohn <dje.gcc@gmail.com>
761 * config/rs6000/rs6000.c (rs6000_passes_ieee128): Protect with #if
764 2018-06-08 Martin Liska <mliska@suse.cz>
766 * tree-cfg.h (debug_function): Fix argument type to match
769 2018-06-08 Martin Liska <mliska@suse.cz>
771 * config/powerpcspe/powerpcspe.c (rs6000_xcoff_visibility):
772 Remove usage of MPX-related (and removed) fields.
773 * config/rs6000/rs6000.c (rs6000_xcoff_visibility): Likewise.
775 2018-06-08 David Malcolm <dmalcolm@redhat.com>
777 * cfg.c (debug): Use TDF_NONE rather than 0.
778 * cfghooks.c (debug): Likewise.
779 * dumpfile.c (DUMP_FILE_INFO): Likewise; also for OPTGROUP.
780 (struct dump_option_value_info): Convert to...
781 (struct kv_pair): ...this template type.
782 (dump_options): Convert to kv_pair<dump_flags_t>; use TDF_NONE
784 (optinfo_verbosity_options): Likewise.
785 (optgroup_options): Convert to kv_pair<optgroup_flags_t>; use
787 (gcc::dump_manager::dump_register): Use optgroup_flags_t rather
788 than int for "optgroup_flags" param.
789 (dump_generic_expr_loc): Use dump_flags_t rather than int for
791 (dump_dec): Likewise.
792 (dump_finish): Use TDF_NONE rather than 0.
793 (gcc::dump_manager::opt_info_enable_passes): Use optgroup_flags_t
794 rather than int for "optgroup_flags" param. Use TDF_NONE rather
795 than 0. Update for change to option_ptr.
796 (opt_info_switch_p_1): Convert "optgroup_flags" param from int *
797 to optgroup_flags_t *. Use TDF_NONE and OPTGROUP_NONE rather than
798 0. Update for changes to optinfo_verbosity_options and
800 (opt_info_switch_p): Convert optgroup_flags from int to
802 (dump_basic_block): Use dump_flags_t rather than int
803 for "dump_kind" param.
804 * dumpfile.h (TDF_ADDRESS, TDF_SLIM, TDF_RAW, TDF_DETAILS,
805 TDF_STATS, TDF_BLOCKS, TDF_VOPS, TDF_LINENO, TDF_UID)
806 TDF_STMTADDR, TDF_GRAPH, TDF_MEMSYMS, TDF_RHS_ONLY, TDF_ASMNAME,
807 TDF_EH, TDF_NOUID, TDF_ALIAS, TDF_ENUMERATE_LOCALS, TDF_CSELIB,
808 TDF_SCEV, TDF_GIMPLE, TDF_FOLDING, MSG_OPTIMIZED_LOCATIONS,
809 MSG_MISSED_OPTIMIZATION, MSG_NOTE, MSG_ALL, TDF_COMPARE_DEBUG,
810 TDF_NONE): Convert from macros to...
811 (enum dump_flag): ...this new enum.
812 (dump_flags_t): Update to use enum.
813 (operator|, operator&, operator~, operator|=, operator&=):
814 Implement for dump_flags_t.
815 (OPTGROUP_NONE, OPTGROUP_IPA, OPTGROUP_LOOP, OPTGROUP_INLINE,
816 OPTGROUP_OMP, OPTGROUP_VEC, OPTGROUP_OTHER, OPTGROUP_ALL):
817 Convert from macros to...
818 (enum optgroup_flag): ...this new enum.
819 (optgroup_flags_t): New typedef.
820 (operator|, operator|=): Implement for optgroup_flags_t.
821 (struct dump_file_info): Convert field "alt_flags" to
822 dump_flags_t. Convert field "optgroup_flags" to
824 (dump_basic_block): Use dump_flags_t rather than int for param.
825 (dump_generic_expr_loc): Likewise.
826 (dump_dec): Likewise.
827 (dump_register): Convert param "optgroup_flags" to
829 (opt_info_enable_passes): Likewise.
830 * early-remat.c (early_remat::dump_edge_list): Use TDF_NONE rather
832 * gimple-pretty-print.c (debug): Likewise.
833 * gimple-ssa-store-merging.c (bswap_replace): Likewise.
834 (merged_store_group::apply_stores): Likewise.
835 * gimple-ssa-strength-reduction.c (insert_initializers): Likewise.
836 * gimple.c (verify_gimple_pp): Likewise.
837 * graphite-poly.c (print_pbb_body): Likewise.
838 * passes.c (pass_manager::register_one_dump_file): Convert
839 local "optgroup_flags" to optgroup_flags_t.
840 * print-tree.c (print_node): Use TDF_NONE rather than 0.
842 (debug_body): Likewise.
843 * tree-pass.h (struct pass_data): Convert field "optgroup_flags"
845 * tree-pretty-print.c (print_struct_decl): Use TDF_NONE rather
847 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Likewise.
848 (convert_mult_to_fma): Likewise.
849 * tree-ssa-reassoc.c (undistribute_ops_list): Likewise.
850 * tree-ssa-sccvn.c (vn_eliminate): Likewise.
851 * tree-vect-data-refs.c (dump_lower_bound): Convert param
852 "dump_kind" to dump_flags_t.
854 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
856 * config/rs6000/rs6000.c (min, max): Delete.
858 2018-06-08 Segher Boessenkool <segher@kernel.crashing.org>
860 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete mention of
861 -mabi=spe and -mabi=no-spe.
863 2018-06-08 Martin Liska <mliska@suse.cz>
865 * ipa-pure-const.c (propagate_pure_const): Use ::get at places
866 where we expect an existing summary.
868 2018-06-08 Martin Liska <mliska@suse.cz>
870 * ipa-inline-analysis.c (simple_edge_hints): Use ::get method.
871 * ipa-inline.h (estimate_edge_growth): Likewise.
873 2018-06-08 Martin Liska <mliska@suse.cz>
875 * cgraph.c (function_version_hasher::hash): Use
876 cgraph_node::get_uid ().
877 (function_version_hasher::equal):
878 * cgraph.h (cgraph_node::get_uid): New method.
879 * ipa-inline.c (update_caller_keys): Use
880 cgraph_node::get_uid ().
881 (update_callee_keys): Likewise.
882 * ipa-utils.c (searchc): Likewise.
883 (ipa_reduced_postorder): Likewise.
884 * lto-cgraph.c (input_node): Likewise.
885 * passes.c (is_pass_explicitly_enabled_or_disabled): Likewise.
886 * symbol-summary.h (symtab_insertion): Likewise.
887 (symtab_removal): Likewise.
888 (symtab_duplication): Likewise.
889 * tree-pretty-print.c (dump_function_header): Likewise.
890 * tree-sra.c (convert_callers_for_node): Likewise.
892 2018-06-08 Martin Liska <mliska@suse.cz>
894 * cgraph.c (symbol_table::create_edge): Always assign a new
896 (symbol_table::free_edge): Do not recycle numbers.
897 * cgraph.h (cgraph_edge::get): New method.
898 * symbol-summary.h (symtab_removal): Use it.
899 (symtab_duplication): Likewise.
900 (call_summary::hashable_uid): Remove.
902 2018-06-08 Martin Liska <mliska@suse.cz>
904 * ipa-inline-analysis.c (inline_edge_removal_hook): Remove.
905 (initialize_growth_caches): Remove.
906 (free_growth_caches): Likewise.
907 (do_estimate_edge_time): Use edge_growth_cache.
908 (do_estimate_edge_size): Likewise.
909 (do_estimate_edge_hints): Likewise.
910 * ipa-inline.c (reset_edge_caches): Likewise.
911 (recursive_inlining): Likewise.
912 (inline_small_functions): Likewise.
913 * ipa-inline.h (initialize_growth_caches): Remove.
914 (estimate_edge_size): Likewise.
915 (estimate_edge_time): Likewise.
916 (estimate_edge_hints): Likewise.
917 (reset_edge_growth_cache): Likewise.
918 * symbol-summary.h (call_summary::remove): New method.
920 2018-06-08 Martin Liska <mliska@suse.cz>
922 * ipa-cp.c (class edge_clone_summary): New summary.
923 (grow_edge_clone_vectors): Remove.
924 (ipcp_edge_duplication_hook): Remove.
925 (class edge_clone_summary_t): New call_summary class.
926 (ipcp_edge_removal_hook): Remove.
927 (edge_clone_summary_t::duplicate): New function.
928 (get_next_cgraph_edge_clone): Use edge_clone_summaries.
929 (create_specialized_node): Likewise.
930 (ipcp_driver): Initialize edge_clone_summaries and do not
933 2018-06-08 Martin Liska <mliska@suse.cz>
935 * symbol-summary.h (get): New function.
936 (call_summary::m_initialize_when_cloning): New class member.
938 2018-06-08 Martin Liska <mliska@suse.cz>
940 * cgraph.c (cgraph_node::remove): Do not recycle uid.
941 * cgraph.h (symbol_table::release_symbol): Do not pass uid.
942 (symbol_table::allocate_cgraph_symbol): Do not set uid.
943 * passes.c (uid_hash_t): Record removed_nodes by their uids.
944 (remove_cgraph_node_from_order): Use the removed_nodes set.
945 (do_per_function_toporder): Likwise.
946 * symbol-summary.h (symtab_insertion): Use cgraph_node::uid
947 instead of summary_uid.
948 (symtab_removal): Likewise.
949 (symtab_duplication): Likewise.
951 2018-06-08 Martin Liska <mliska@suse.cz>
953 * ipa-cp.c (ipcp_store_bits_results): Use
954 ipcp_transformation_sum.
955 (ipcp_store_vr_results): Likewise.
956 * ipa-prop.c (ipcp_grow_transformations_if_necessary): Renamed
958 (ipcp_transformation_initialize): ... this.
959 (ipa_set_node_agg_value_chain):
960 (ipa_node_params_t::duplicate): Use ipcp_transformation_sum.
961 (write_ipcp_transformation_info): Likewise.
962 (read_ipcp_transformation_info): Likewise.
963 (ipcp_update_bits): Likewise.
964 (ipcp_update_vr): Likewise.
965 (ipcp_transform_function): Likewise.
966 * ipa-prop.h: Rename ipcp_transformation_summary to
968 (class ipcp_transformation_t): New function summary.
969 (ipcp_get_transformation_summary): Use ipcp_transformation_sum.
970 (ipa_get_agg_replacements_for_node): Likewise.
972 2018-06-08 Martin Liska <mliska@suse.cz>
974 * ipa-pure-const.c (struct funct_state_d): Do it class instead
976 (class funct_state_summary_t): New function_summary class.
977 (has_function_state): Remove.
978 (get_function_state): Likewise.
979 (set_function_state): Likewise.
980 (add_new_function): Likewise.
981 (funct_state_summary_t::insert): New function.
982 (duplicate_node_data): Remove.
983 (remove_node_data): Remove.
984 (funct_state_summary_t::duplicate): New function.
985 (register_hooks): Create new funct_state_summaries.
986 (pure_const_generate_summary): Use it.
987 (pure_const_write_summary): Likewise.
988 (pure_const_read_summary): Likewise.
989 (propagate_pure_const): Likewise.
990 (propagate_nothrow): Likewise.
991 (dump_malloc_lattice): Likewise.
992 (propagate_malloc): Likewise.
993 (execute): Do not register hooks, just remove summary
995 (pass_ipa_pure_const::pass_ipa_pure_const): Simplify
998 2018-06-08 Martin Liska <mliska@suse.cz>
1000 * ipa-reference.c (remove_node_data): Remove.
1001 (duplicate_node_data): Likewise.
1002 (class ipa_ref_var_info_summary_t): New class.
1003 (class ipa_ref_opt_summary_t): Likewise.
1004 (get_reference_vars_info): Use ipa_ref_var_info_summaries.
1005 (get_reference_optimization_summary): Use
1006 ipa_ref_opt_sum_summaries.
1007 (set_reference_vars_info): Remove.
1008 (set_reference_optimization_summary): Likewise.
1009 (ipa_init): Create summaries.
1010 (init_function_info): Use function summary.
1011 (ipa_ref_opt_summary_t::duplicate): New function.
1012 (ipa_ref_opt_summary_t::remove): New function.
1013 (get_read_write_all_from_node): Fix GNU coding style.
1014 (propagate): Use function summary.
1015 (write_node_summary_p): Fix GNU coding style.
1016 (stream_out_bitmap): Likewise.
1017 (ipa_reference_read_optimization_summary): Use function summary.
1018 (ipa_reference_c_finalize): Do not release hooks.
1020 2018-06-08 Martin Liska <mliska@suse.cz>
1022 * ipa-fnsummary.c (dump_ipa_call_summary): Use ::get method.
1023 (analyze_function_body): Extract multiple calls of get_create.
1024 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
1025 * ipa-inline.c (recursive_inlining): Use ::get method.
1026 * ipa-inline.h (estimate_edge_growth): Likewise.
1028 2018-06-08 Martin Liska <mliska@suse.cz>
1030 * hsa-common.h (enum hsa_function_kind): Rename HSA_NONE to
1032 (hsa_function_summary::hsa_function_summary): Use the new enum
1034 (hsa_gpu_implementation_p): Use hsa_summaries::get.
1035 * hsa-gen.c (hsa_get_host_function): Likewise.
1036 (get_brig_function_name): Likewise.
1037 * ipa-hsa.c (process_hsa_functions): Likewise.
1038 (ipa_hsa_write_summary): Likewise.
1039 * symbol-summary.h (symtab_duplication): Use ::get function/
1040 (get): New function.
1042 2018-06-08 Martin Liska <mliska@suse.cz>
1044 * config/i386/i386.c (ix86_can_inline_p): Use get_create instead
1046 * hsa-common.c (hsa_summary_t::link_functions): Likewise.
1047 (hsa_register_kernel): Likewise.
1048 * hsa-common.h (hsa_gpu_implementation_p): Likewise.
1049 * hsa-gen.c (hsa_get_host_function): Likewise.
1050 (get_brig_function_name): Likewise.
1051 (generate_hsa): Likewise.
1052 (pass_gen_hsail::execute): Likewise.
1053 * ipa-cp.c (ipcp_cloning_candidate_p): Likewise.
1054 (devirtualization_time_bonus): Likewise.
1055 (ipcp_propagate_stage): Likewise.
1056 * ipa-fnsummary.c (redirect_to_unreachable): Likewise.
1057 (edge_set_predicate): Likewise.
1058 (evaluate_conditions_for_known_args): Likewise.
1059 (evaluate_properties_for_edge): Likewise.
1060 (ipa_fn_summary::reset): Likewise.
1061 (ipa_fn_summary_t::duplicate): Likewise.
1062 (dump_ipa_call_summary): Likewise.
1063 (ipa_dump_fn_summary): Likewise.
1064 (analyze_function_body): Likewise.
1065 (compute_fn_summary): Likewise.
1066 (estimate_edge_devirt_benefit): Likewise.
1067 (estimate_edge_size_and_time): Likewise.
1068 (estimate_calls_size_and_time): Likewise.
1069 (estimate_node_size_and_time): Likewise.
1070 (inline_update_callee_summaries): Likewise.
1071 (remap_edge_change_prob): Likewise.
1072 (remap_edge_summaries): Likewise.
1073 (ipa_merge_fn_summary_after_inlining): Likewise.
1074 (ipa_update_overall_fn_summary): Likewise.
1075 (read_ipa_call_summary): Likewise.
1076 (inline_read_section): Likewise.
1077 (write_ipa_call_summary): Likewise.
1078 (ipa_fn_summary_write): Likewise.
1079 (ipa_free_fn_summary): Likewise.
1080 * ipa-hsa.c (process_hsa_functions): Likewise.
1081 (ipa_hsa_write_summary): Likewise.
1082 (ipa_hsa_read_section): Likewise.
1083 * ipa-icf.c (sem_function::merge): Likewise.
1084 * ipa-inline-analysis.c (simple_edge_hints): Likewise.
1085 (do_estimate_edge_time): Likewise.
1086 (estimate_size_after_inlining): Likewise.
1087 (estimate_growth): Likewise.
1088 (growth_likely_positive): Likewise.
1089 * ipa-inline-transform.c (clone_inlined_nodes): Likewise.
1090 (inline_call): Likewise.
1091 * ipa-inline.c (caller_growth_limits): Likewise.
1092 (can_inline_edge_p): Likewise.
1093 (can_inline_edge_by_limits_p): Likewise.
1094 (compute_uninlined_call_time): Likewise.
1095 (compute_inlined_call_time): Likewise.
1096 (want_inline_small_function_p): Likewise.
1097 (edge_badness): Likewise.
1098 (update_caller_keys): Likewise.
1099 (update_callee_keys): Likewise.
1100 (recursive_inlining): Likewise.
1101 (inline_small_functions): Likewise.
1102 (inline_to_all_callers_1): Likewise.
1103 (dump_overall_stats): Likewise.
1104 (early_inline_small_functions): Likewise.
1105 (early_inliner): Likewise.
1106 * ipa-inline.h (estimate_edge_growth): Likewise.
1107 * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
1108 * ipa-prop.c (ipa_make_edge_direct_to_target): Likewise.
1109 * ipa-prop.h (IPA_NODE_REF): Likewise.
1110 (IPA_EDGE_REF): Likewise.
1111 * ipa-pure-const.c (malloc_candidate_p): Likewise.
1112 (propagate_malloc): Likewise.
1113 * ipa-split.c (execute_split_functions): Likewise.
1114 * symbol-summary.h: Rename get to get_create.
1116 (get_create): Likewise.
1117 * tree-sra.c (ipa_sra_preliminary_function_checks): Likewise.
1119 2018-06-08 Martin Liska <mliska@suse.cz>
1121 * symbol-summary.h (release): Move definition out of class
1123 (symtab_removal): Likewise.
1124 (symtab_duplication): Likewise.
1126 2018-06-08 Martin Liska <mliska@suse.cz>
1128 * symbol-summary.h (function_summary): Move constructor
1129 implementation out of class declaration.
1130 (release): Likewise.
1131 (symtab_insertion): Likewise.
1132 (symtab_removal): Likewise.
1133 (symtab_duplication): Likewise.
1136 2018-06-08 Martin Liska <mliska@suse.cz>
1138 * Makefile.in: Remove support for MPX (macros, related functions,
1139 fields in cgraph_node, ...).
1140 * builtin-types.def (BT_BND): Likewise.
1141 (BT_FN_BND_CONST_PTR): Likewise.
1142 (BT_FN_CONST_PTR_BND): Likewise.
1143 (BT_FN_VOID_PTR_BND): Likewise.
1144 (BT_FN_BND_CONST_PTR_SIZE): Likewise.
1145 (BT_FN_VOID_CONST_PTR_BND_CONST_PTR): Likewise.
1146 * builtins.c (expand_builtin_memcpy_with_bounds): Likewise.
1147 (expand_builtin_mempcpy_with_bounds): Likewise.
1148 (expand_builtin_memset_with_bounds): Likewise.
1149 (expand_builtin_memset_args): Likewise.
1150 (std_expand_builtin_va_start): Likewise.
1151 (expand_builtin): Likewise.
1152 (expand_builtin_with_bounds): Likewise.
1153 * builtins.def (DEF_BUILTIN_CHKP): Likewise.
1154 (DEF_LIB_BUILTIN_CHKP): Likewise.
1155 (DEF_EXT_LIB_BUILTIN_CHKP): Likewise.
1156 (DEF_CHKP_BUILTIN): Likewise.
1157 (BUILT_IN_MEMCPY): Likewise.
1158 (BUILT_IN_MEMMOVE): Likewise.
1159 (BUILT_IN_MEMPCPY): Likewise.
1160 (BUILT_IN_MEMSET): Likewise.
1161 (BUILT_IN_STPCPY): Likewise.
1162 (BUILT_IN_STRCAT): Likewise.
1163 (BUILT_IN_STRCHR): Likewise.
1164 (BUILT_IN_STRCPY): Likewise.
1165 (BUILT_IN_STRLEN): Likewise.
1166 (BUILT_IN_MEMCPY_CHK): Likewise.
1167 (BUILT_IN_MEMMOVE_CHK): Likewise.
1168 (BUILT_IN_MEMPCPY_CHK): Likewise.
1169 (BUILT_IN_MEMSET_CHK): Likewise.
1170 (BUILT_IN_STPCPY_CHK): Likewise.
1171 (BUILT_IN_STRCAT_CHK): Likewise.
1172 (BUILT_IN_STRCPY_CHK): Likewise.
1173 * calls.c (store_bounds): Likewise.
1174 (emit_call_1): Likewise.
1175 (special_function_p): Likewise.
1176 (maybe_warn_nonstring_arg): Likewise.
1177 (initialize_argument_information): Likewise.
1178 (finalize_must_preallocate): Likewise.
1179 (compute_argument_addresses): Likewise.
1180 (expand_call): Likewise.
1181 * cfgexpand.c (expand_call_stmt): Likewise.
1182 (expand_return): Likewise.
1183 (expand_gimple_stmt_1): Likewise.
1184 (pass_expand::execute): Likewise.
1185 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
1186 (cgraph_node::remove): Likewise.
1187 (cgraph_node::dump): Likewise.
1188 (cgraph_node::verify_node): Likewise.
1189 * cgraph.h (chkp_function_instrumented_p): Likewise.
1190 (symtab_node::get_alias_target): Likewise.
1191 (cgraph_node::can_remove_if_no_direct_calls_and_refs_p): Likewise.
1192 (cgraph_local_p): Likewise.
1193 * cgraphbuild.c (cgraph_edge::rebuild_edges): Likewise.
1194 (cgraph_edge::rebuild_references): Likewise.
1195 * cgraphunit.c (varpool_node::finalize_decl): Likewise.
1196 (walk_polymorphic_call_targets): Likewise.
1197 (cgraph_node::expand_thunk): Likewise.
1198 (symbol_table::output_weakrefs): Likewise.
1199 * common/config/i386/i386-common.c (OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Likewise.
1200 (ix86_handle_option): Likewise.
1201 * config/i386/constraints.md: Likewise.
1202 * config/i386/i386-builtin-types.def (BND): Likewise.
1206 * config/i386/i386-builtin.def (BDESC_END): Likewise.
1207 (BDESC_FIRST): Likewise.
1209 * config/i386/i386-c.c (ix86_target_macros_internal): Likewise.
1210 * config/i386/i386-protos.h (ix86_bnd_prefixed_insn_p): Likewise.
1211 * config/i386/i386.c (enum reg_class): Likewise.
1212 (ix86_target_string): Likewise.
1213 (ix86_option_override_internal): Likewise.
1214 (ix86_conditional_register_usage): Likewise.
1215 (ix86_valid_target_attribute_inner_p): Likewise.
1216 (ix86_set_indirect_branch_type): Likewise.
1217 (ix86_set_current_function): Likewise.
1218 (ix86_function_arg_regno_p): Likewise.
1219 (init_cumulative_args): Likewise.
1220 (ix86_function_arg_advance): Likewise.
1221 (ix86_function_arg): Likewise.
1222 (ix86_pass_by_reference): Likewise.
1223 (ix86_function_value_regno_p): Likewise.
1224 (ix86_function_value_1): Likewise.
1225 (ix86_function_value_bounds): Likewise.
1226 (ix86_return_in_memory): Likewise.
1227 (ix86_setup_incoming_vararg_bounds): Likewise.
1228 (ix86_va_start): Likewise.
1229 (indirect_thunk_need_prefix): Likewise.
1230 (print_reg): Likewise.
1231 (ix86_print_operand): Likewise.
1232 (ix86_expand_call): Likewise.
1233 (ix86_output_function_return): Likewise.
1234 (reg_encoded_number): Likewise.
1235 (BDESC_VERIFYS): Likewise.
1236 (ix86_init_mpx_builtins): Likewise.
1237 (ix86_init_builtins): Likewise.
1238 (ix86_emit_cmove): Likewise.
1239 (ix86_emit_move_max): Likewise.
1240 (ix86_expand_builtin): Likewise.
1241 (ix86_builtin_mpx_function): Likewise.
1242 (ix86_get_arg_address_for_bt): Likewise.
1243 (ix86_load_bounds): Likewise.
1244 (ix86_store_bounds): Likewise.
1245 (ix86_load_returned_bounds): Likewise.
1246 (ix86_store_returned_bounds): Likewise.
1247 (ix86_class_likely_spilled_p): Likewise.
1248 (ix86_hard_regno_mode_ok): Likewise.
1249 (x86_order_regs_for_local_alloc): Likewise.
1250 (ix86_mitigate_rop): Likewise.
1251 (ix86_bnd_prefixed_insn_p): Likewise.
1252 (ix86_mpx_bound_mode): Likewise.
1253 (ix86_make_bounds_constant): Likewise.
1254 (ix86_initialize_bounds): Likewise.
1255 (TARGET_LOAD_BOUNDS_FOR_ARG): Likewise.
1256 (TARGET_STORE_BOUNDS_FOR_ARG): Likewise.
1257 (TARGET_LOAD_RETURNED_BOUNDS): Likewise.
1258 (TARGET_STORE_RETURNED_BOUNDS): Likewise.
1259 (TARGET_CHKP_BOUND_MODE): Likewise.
1260 (TARGET_BUILTIN_CHKP_FUNCTION): Likewise.
1261 (TARGET_CHKP_FUNCTION_VALUE_BOUNDS): Likewise.
1262 (TARGET_CHKP_MAKE_BOUNDS_CONSTANT): Likewise.
1263 (TARGET_CHKP_INITIALIZE_BOUNDS): Likewise.
1264 * config/i386/i386.h (TARGET_MPX): Likewise.
1265 (TARGET_MPX_P): Likewise.
1266 (VALID_BND_REG_MODE): Likewise.
1267 (FIRST_BND_REG): Likewise.
1268 (LAST_BND_REG): Likewise.
1269 (enum reg_class): Likewise.
1270 (BND_REG_P): Likewise.
1271 (BND_REGNO_P): Likewise.
1272 (BNDmode): Likewise.
1273 (ADJUST_INSN_LENGTH): Likewise.
1274 * config/i386/i386.md: Likewise.
1275 * config/i386/i386.opt: Likewise.
1276 * config/i386/linux-common.h (LIBMPX_LIBS): Likewise.
1277 (defined): Likewise.
1278 (LINK_MPX): Likewise.
1279 (MPX_SPEC): Likewise.
1280 (LIBMPX_SPEC): Likewise.
1281 (LIBMPXWRAPPERS_SPEC): Likewise.
1282 (CHKP_SPEC): Likewise.
1283 * config/i386/predicates.md: Likewise.
1284 * dbxout.c (dbxout_type): Likewise.
1285 * doc/extend.texi: Likewise.
1286 * doc/invoke.texi: Likewise.
1287 * doc/md.texi: Likewise.
1288 * doc/tm.texi: Likewise.
1289 * doc/tm.texi.in: Likewise.
1290 * dwarf2out.c (is_base_type): Likewise.
1291 (gen_formal_types_die): Likewise.
1292 (gen_subprogram_die): Likewise.
1293 (gen_type_die_with_usage): Likewise.
1294 (gen_decl_die): Likewise.
1295 (dwarf2out_late_global_decl): Likewise.
1296 * expr.c (expand_assignment): Likewise.
1297 (emit_storent_insn): Likewise.
1298 (store_expr_with_bounds): Likewise.
1299 (store_expr): Likewise.
1300 (expand_expr_real_1): Likewise.
1301 * expr.h (store_expr_with_bounds): Likewise.
1302 * function.c (use_register_for_decl): Likewise.
1303 (struct bounds_parm_data): Likewise.
1304 (assign_parms_augmented_arg_list): Likewise.
1305 (assign_parm_find_entry_rtl): Likewise.
1306 (assign_parm_is_stack_parm): Likewise.
1307 (assign_parm_load_bounds): Likewise.
1308 (assign_bounds): Likewise.
1309 (assign_parms): Likewise.
1310 (expand_function_start): Likewise.
1311 * gcc.c (CHKP_SPEC): Likewise.
1312 * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
1313 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Likewise.
1314 (wrestrict_dom_walker::check_call): Likewise.
1315 * gimple.c (gimple_build_call_from_tree): Likewise.
1316 * gimple.h (enum gf_mask): Likewise.
1317 (gimple_call_with_bounds_p): Likewise.
1318 (gimple_call_set_with_bounds): Likewise.
1319 * gimplify.c (gimplify_init_constructor): Likewise.
1320 * ipa-cp.c (initialize_node_lattices): Likewise.
1321 (propagate_constants_across_call): Likewise.
1322 (find_more_scalar_values_for_callers_subset): Likewise.
1323 * ipa-hsa.c (process_hsa_functions): Likewise.
1324 * ipa-icf-gimple.c (func_checker::compare_gimple_call): Likewise.
1325 * ipa-icf.c (sem_function::merge): Likewise.
1326 * ipa-inline.c (early_inliner): Likewise.
1327 * ipa-pure-const.c (warn_function_noreturn): Likewise.
1328 (warn_function_cold): Likewise.
1329 (propagate_pure_const): Likewise.
1330 * ipa-ref.h (enum GTY): Likewise.
1331 * ipa-split.c (find_retbnd): Likewise.
1332 (consider_split): Likewise.
1333 (split_function): Likewise.
1334 * ipa-visibility.c (cgraph_externally_visible_p): Likewise.
1335 * ipa.c (walk_polymorphic_call_targets): Likewise.
1336 (symbol_table::remove_unreachable_nodes): Likewise.
1337 (process_references): Likewise.
1338 (cgraph_build_static_cdtor_1): Likewise.
1339 * lto-cgraph.c (lto_output_node): Likewise.
1340 (output_refs): Likewise.
1341 (compute_ltrans_boundary): Likewise.
1342 (input_overwrite_node): Likewise.
1343 (input_node): Likewise.
1344 (input_cgraph_1): Likewise.
1345 * params.def (PARAM_CHKP_MAX_CTOR_SIZE): Likewise.
1346 * passes.c (pass_manager::execute_early_local_passes): Likewise.
1347 (class pass_chkp_instrumentation_passes): Likewise.
1348 (make_pass_chkp_instrumentation_passes): Likewise.
1349 * passes.def: Likewise.
1350 * rtl.h (struct GTY): Likewise.
1351 (CALL_EXPR_WITH_BOUNDS_P): Likewise.
1352 * stor-layout.c (layout_type): Likewise.
1353 * symtab.c: Likewise.
1354 * target.def: Likewise.
1355 * targhooks.c (default_chkp_bound_type): Likewise.
1356 (default_chkp_bound_mode): Likewise.
1357 (default_builtin_chkp_function): Likewise.
1358 (default_chkp_function_value_bounds): Likewise.
1359 (default_chkp_make_bounds_constant): Likewise.
1360 (default_chkp_initialize_bounds): Likewise.
1361 * targhooks.h (default_chkp_bound_type): Likewise.
1362 (default_chkp_bound_mode): Likewise.
1363 (default_builtin_chkp_function): Likewise.
1364 (default_chkp_function_value_bounds): Likewise.
1365 (default_chkp_make_bounds_constant): Likewise.
1366 (default_chkp_initialize_bounds): Likewise.
1367 * toplev.c (compile_file): Likewise.
1368 (process_options): Likewise.
1369 * tree-core.h (DEF_BUILTIN): Likewise.
1370 (DEF_BUILTIN_CHKP): Likewise.
1371 * tree-inline.c (declare_return_variable): Likewise.
1372 (remap_gimple_stmt): Likewise.
1373 (copy_bb): Likewise.
1374 (initialize_inlined_parameters): Likewise.
1375 (expand_call_inline): Likewise.
1376 * tree-pass.h (make_pass_ipa_chkp_versioning): Likewise.
1377 (make_pass_ipa_chkp_early_produce_thunks): Likewise.
1378 (make_pass_ipa_chkp_produce_thunks): Likewise.
1379 (make_pass_chkp): Likewise.
1380 (make_pass_chkp_opt): Likewise.
1381 (make_pass_chkp_instrumentation_passes): Likewise.
1382 * tree-pretty-print.c (dump_generic_node): Likewise.
1383 * tree-ssa-ccp.c (insert_clobber_before_stack_restore): Likewise.
1384 * tree-ssa-dce.c (propagate_necessity): Likewise.
1385 (eliminate_unnecessary_stmts): Likewise.
1386 * tree-ssa-pre.c (create_expression_by_pieces): Likewise.
1387 * tree-ssa-sccvn.c (copy_reference_ops_from_call): Likewise.
1388 * tree-ssa-sccvn.h: Likewise.
1389 * tree-ssa-strlen.c (get_string_length): Likewise.
1390 (valid_builtin_call): Likewise.
1391 (adjust_last_stmt): Likewise.
1392 (handle_builtin_strchr): Likewise.
1393 (handle_builtin_strcpy): Likewise.
1394 (handle_builtin_stxncpy): Likewise.
1395 (handle_builtin_memcpy): Likewise.
1396 (handle_builtin_strcat): Likewise.
1397 (strlen_check_and_optimize_stmt): Likewise.
1398 * tree-stdarg.c (expand_ifn_va_arg_1): Likewise.
1399 * tree-streamer-in.c: Likewise.
1400 * tree-streamer.c (record_common_node): Likewise.
1401 * tree.c (tree_code_size): Likewise.
1402 (wide_int_to_tree_1): Likewise.
1403 (type_contains_placeholder_1): Likewise.
1404 (build_common_tree_nodes): Likewise.
1405 * tree.def (POINTER_BOUNDS_TYPE): Likewise.
1406 * tree.h (POINTER_BOUNDS_TYPE_P): Likewise.
1407 (POINTER_BOUNDS_P): Likewise.
1408 (BOUNDED_TYPE_P): Likewise.
1409 (BOUNDED_P): Likewise.
1410 (CALL_WITH_BOUNDS_P): Likewise.
1411 (pointer_bounds_type_node): Likewise.
1412 * value-prof.c (gimple_ic): Likewise.
1413 * var-tracking.c (vt_add_function_parameters): Likewise.
1414 * varasm.c (make_decl_rtl): Likewise.
1415 (assemble_start_function): Likewise.
1416 (output_constant): Likewise.
1417 (maybe_assemble_visibility): Likewise.
1418 * varpool.c (ctor_for_folding): Likewise.
1419 * chkp-builtins.def: Remove.
1420 * ipa-chkp.c: Remove.
1421 * ipa-chkp.h: Remove.
1422 * rtl-chkp.c: Remove.
1423 * rtl-chkp.h: Remove.
1424 * tree-chkp-opt.c: Remove.
1425 * tree-chkp.c: Remove.
1426 * tree-chkp.h: Remove.
1428 2018-06-07 Carl Love <cel@us.ibm.com>
1430 * gcc/config/rs6000/vsx.md (vextract_fp_from_shorth,
1431 vextract_fp_from_shortl): Add BE support.
1433 2018-06-07 Paul Koning <ni1d@arrl.net>
1435 * compare-elim.c (try_merge_compare): Don't merge compare if
1436 address contains a side effect.
1437 (try_eliminate_compare): Likewise.
1439 2018-06-07 Olga Makhotina <olga.makhotina@intel.com>
1441 * config.gcc: Support "tremont".
1442 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "tremont".
1443 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
1445 * config/i386/i386.c (m_TREMONT): Define.
1446 (processor_target_table): Add "tremont".
1447 (PTA_TREMONT): Define.
1448 (ix86_lea_outperforms): Add TARGET_TREMONT.
1449 (get_builtin_code_for_version): Handle PROCESSOR_TREMONT.
1450 (fold_builtin_cpu): Add M_INTEL_TREMONT, replace M_INTEL_GOLDMONT
1451 and M_INTEL_GOLDMONT_PLUS.
1452 (fold_builtin_cpu): Add "tremont".
1453 (ix86_add_stmt_cost): Add TARGET_TREMONT.
1454 (ix86_option_override_internal): Add "tremont".
1455 * config/i386/i386.h (processor_costs): Define TARGET_TREMONT.
1456 (processor_type): Add PROCESSOR_TREMONT.
1457 * config/i386/x86-tune.def: Add m_TREMONT.
1458 * doc/invoke.texi: Add tremont as x86 -march=/-mtune= CPU type.
1460 2018-06-07 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1462 * gcc/config/msp430/msp430.c (msp430_mcu_name): Set the "i" in the
1463 symbol defined for msp430i* devices to be lower case.
1465 gcc/testsuite/gcc.target/msp430/
1466 * msp430i-device-symbol.c: New test.
1467 * msp430f-device-symbol.c: New test.
1468 * msp430.h: New test header file.
1470 2018-06-07 Richard Biener <rguenther@suse.de>
1472 * graphite-sese-to-poly.c (extract_affine): Avoid unneded
1473 wrapping. Properly wrap the result of a BIT_NOT_EXPR.
1474 Properly wrap signed arithmetic if overflow wraps.
1476 2018-06-07 Jakub Jelinek <jakub@redhat.com>
1478 PR tree-optimization/69615
1479 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If rhs2 is lhs
1480 of a cast from a same precision integral SSA_NAME in a bb dominated
1481 by first_bb, retry with rhs2 set to the rhs1 of the cast. Don't emit
1482 cast to utype if rhs2 has already a compatible type.
1484 2018-06-07 Richard Biener <rguenther@suse.de>
1486 PR tree-optimization/85935
1487 * graphite-scop-detection.c (find_params_in_bb): Analyze
1488 condition operands with respect to the correct loop. Assert
1489 the analysis doesn't fail.
1491 2018-06-04 Carl Love <cel@us.ibm.com>
1493 * gcc/config/rs6000/vsx.md (first_match_index_<mode>): Calculate index
1494 using natural element order. Use gen_lshrsi3 instead of gen_ashrsi3
1495 as it is slightly cheaper.
1496 (first_match_or_eos_index_<mode>):
1497 Calculate index using natural element order.
1498 (first_match_index_<mode>):
1499 Calculate index using natural element order.
1500 (first_match_or_eos_index_<mode>):
1501 Calculate index using natural order.
1502 (define_insn vclzlsbb): Change to define_insn vclzlsbb_<mode>.
1503 for BE and LE modes.
1504 * gcc/config/rs6000/rs6000-c.c: Rename P9V_BUILTIN_VCLZLSBB,
1505 P9V_BUILTIN_VCLZLSBB_V16QI.
1506 * gcc/config/rs6000/rs6000-builtin.def: Make VCLZLSBB mode
1509 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
1511 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
1512 indentation and line wrap for many prototypes. Add missing
1513 @smallexample directives around block of prototypes for vec_xl and
1516 2018-06-05 Michael Meissner <meissner@linux.ibm.com>
1518 * config/rs6000/rs6000.c (rs6000_passes_ieee128): New boolean to
1519 track if we pass or return IEEE 128-bit floating point.
1520 (ieee128_mangling_gcc_8_1): New boolean to say whether to generate
1521 C++ mangling that is compatible with GCC 8.1.
1522 (TARGET_ASM_GLOBALIZE_DECL_NAME): Override target hook.
1523 (init_cumulative_args): Note if we pass or return IEEE 128-bit
1524 floating point types.
1525 (rs6000_function_arg_advance_1): Likewise.
1526 (rs6000_mangle_type): Optionally generate mangled names that match
1527 what GCC 8.1 generated for IEEE 128-bit floating point types.
1528 (rs6000_globalize_decl_name): If we have an external function that
1529 passes or returns IEEE 128-bit types, generate a weak reference
1530 from the mangled name used in GCC 8.1 to the current mangled
1532 (rs6000_init_builtins): Make __ibm128 use the long double type if
1533 long double is IBM extended double. Make __float128 use the long
1534 double type if long double is IEEE 128-bit.
1537 * config/rs6000/rs6000-builtin.def (BU_IBM128_2): New helper
1538 macro for __ibm128 built-in functions.
1539 (PACK_IF): Add __ibm128 pack/unpack functions.
1540 (UNPACK_IF): Likewise.
1541 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Do not
1542 enable long double built-in functions if long double is IEEE
1543 128-bit floating point.
1544 (rs6000_invalid_builtin): Update long double built-in function
1546 (rs6000_expand_builtin): For PACK_IF and UNPACK_IF built-in
1547 functions, adjust the built-in function to use the long double
1548 built-in function if __ibm128 and long double are the same type.
1549 * doc/extend.texi (PowerPC builtins): Update documention for
1550 __builtin_{,un}pack_longdouble. Add documentation for
1551 __builtin_{,un}pack_ibm128.
1553 2018-06-06 Jim Wilson <jimw@sifive.com>
1555 * config/riscv/riscv.c (enum riscv_privilege_levels): New.
1556 (struct machine_function): New field interrupt_mode.
1557 (riscv_handle_type_attribute): New function. Add forward declaration.
1558 (riscv_attribute_table) <interrupt>: Use riscv_handle_type_attribute.
1559 (riscv_expand_epilogue): Check interrupt_mode field.
1560 (riscv_set_current_function): Check interrupt attribute args and
1561 set interrupt_mode field.
1562 * config/riscv/riscv.md (UNSPECV_SRET, UNSPECV_URET): New.
1563 (riscv_sret, riscv_uret): New.
1564 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: Document
1565 new arguments to interrupt attribute.
1567 2018-06-06 Peter Bergner <bergner@vnet.ibm.com>
1570 * /config/rs6000/rs6000.h (ASM_CPU_SPEC): Add support for -mpower9.
1571 Don't handle -mcpu=power8 if -mpower9-vector is also used.
1573 2018-06-06 Kelvin Nilsen <kelvin@gcc.gnu.org>
1575 * config/rs6000/rs6000-builtin.def (VSX_BUILTIN_VEC_LD,
1576 VSX_BUILTIN_VEC_ST): Add comment to explain non-traditional uses.
1577 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
1578 several redundant entries.
1580 2018-06-06 David Malcolm <dmalcolm@redhat.com>
1582 * config/i386/i386-protos.h (ix86_expand_call): Strengthen return
1583 type from "rtx" to "rtx_insn *".
1584 * config/i386/i386.c (ix86_expand_split_stack_prologue): Likewise
1585 for local "call_insn", removing cast.
1586 (ix86_expand_call): Likewise, introducing a "call_insn" local.
1588 2018-06-06 Eric Botcazou <ebotcazou@adacore.com>
1590 PR tree-optimization/86066
1591 * gimple-ssa-store-merging.c (process_store): Do not bypass BIT_NOT_EXPR
1592 for BIT_INSERT_EXPR stores.
1594 2018-06-06 Richard Biener <rguenther@suse.de>
1596 PR tree-optimization/86062
1597 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
1598 component refs ontop
1599 of to be offsetted base.
1601 2018-06-06 Jozef Lawrynowicz <jozef.l@mittosystems.com>
1603 * gcc/config/msp430/msp430.c (msp430_attr): Allow interrupt handlers
1604 to be static and remove check on interrupt attribute name.
1606 gcc/testsuite/gcc.target/msp430/
1607 * function-attributes-4.c: New test.
1608 * static-interrupts.c: New test.
1610 2018-06-05 Kelvin Nilsen <kelvin@gcc.gnu.org>
1612 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
1613 volatile qualifier from vec_lvsl and vec_lvsr argument prototypes.
1615 2018-06-05 Steve Ellcey <sellcey@cavium.com>
1618 * config/aarch64/aarch64-protos.h (aarch64_err_no_fpadvsimd): Remove
1620 * config/aarch64/aarch64-protos..c (aarch64_err_no_fpadvsimd):
1621 Remove second argument, change how error is called.
1622 (aarch64_layout_arg): Remove second argument from
1623 aarch64_err_no_fpadvsimd call.
1624 (aarch64_init_cumulative_args): Ditto.
1625 (aarch64_gimplify_va_arg_expr): Ditto.
1626 * config/aarch64/aarch64.md (mov<mode>): Ditto.
1628 2018-06-05 Uros Bizjak <ubizjak@gmail.com>
1630 * config/i386/i386.md (simple_return_indirect_internal): New expander.
1631 (*simple_return_indirect_internal<mode>): Rename from
1632 simple_return_indirect_internal. Use W mode iterator.
1633 (rstorssp): New expander.
1634 (*rstorssp<mode>): Rename from rstorssp. Use P mode iterator.
1635 (clrssbsy): New expander.
1636 (*clrssbsy<mode>): Rename from clrssbsy. Use P mode iterator.
1638 2018-06-05 Andre Vieira <andre.simoesdiasvieira@arm.com>
1640 * config/arm/arm_cmse.h (cmse_nsfptr_create): Change typeof to
1642 (cmse_check_pointed_object): Likewise.
1644 2018-06-05 Martin Liska <mliska@suse.cz>
1646 PR gcov-profile/47618
1647 * doc/invoke.texi: Document how -fprofile-dir format
1650 2018-06-05 Richard Biener <rguenther@suse.de>
1652 * tree-cfgcleanup.c (cleanup_control_flow_pre): For edge
1653 removal pretend DOM info isn't available so we do not update
1654 it and only remove edges, not dominated blocks. Actually free
1655 DOM info in case we removed something. Remove unreachable blocks.
1656 (mfb_keep_latches): Work with either DOM info or marked backedges.
1657 (cleanup_tree_cfg_noloop): Do not remove unreachable blocks
1658 first. Mark backedges if DOM info isn't available.
1659 (Re-)compute DOM info after cleanup_control_flow_pre.
1661 2018-06-05 Richard Biener <rguenther@suse.de>
1663 * tree-cfg.c (struct locus_discrim_map): Store line, not location.
1664 (locus_discrim_hasher::hash): Adjust.
1665 (locus_discrim_hasher::equal): Likewise.
1666 (next_discriminator_for_locus): Work on line directly.
1667 (same_line_p): Pass in expanded locus1 as well.
1668 (assign_discriminators): Avoid redundant location expansions.
1670 2018-06-05 Richard Biener <rguenther@suse.de>
1672 PR tree-optimization/86046
1673 * tree-ssa.c (maybe_optimize_var): Clear DECL_GIMPLE_REG_P
1674 if required after clearing TREE_ADDRESSABLE.
1676 2018-06-05 Richard Biener <rguenther@suse.de>
1678 PR tree-optimization/86047
1679 * tree-ssa-loop.c (for_each_index): Glob handling of all
1680 decls and constants and really handle all of them.
1682 2018-06-05 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1685 * config/arm/arm-builtins.c (arm_type_qualifiers): Add
1686 qualifier_void_pointer and qualifier_const_void_pointer.
1687 (arm_ldc_qualifiers, arm_stc_qualifiers): Use the above.
1688 (arm_init_builtins): Handle the above.
1689 * config/arm/arm_acle.h (__arm_cdp, __arm_ldc, __arm_ldcl, __arm_stc,
1690 __arm_stcl, __arm_mcr, __arm_cdp2, __arm_ldc2, __arm_ldcl2, __arm_stc2,
1691 __arm_stcl2,__arm_mcr2, __arm_mcrr, __arm_mcrr2): Remove return for
1694 2018-06-05 Martin Liska <mliska@suse.cz>
1696 * auto-profile.c (read_autofdo_file): Do not use
1697 gcov_ctr_summary struct.
1698 (afdo_callsite_hot_enough_for_early_inline): Likewise.
1699 * coverage.c (struct counts_entry): Likewise.
1700 (read_counts_file): Read just single summary entry.
1701 (get_coverage_counts): Use gcov_summary struct.
1702 * coverage.h (get_coverage_counts): Likewise.
1703 * gcov-dump.c (dump_working_sets): Likewise.
1704 (tag_summary): Dump just single summary.
1705 * gcov-io.c (gcov_write_summary): Write just histogram
1707 (gcov_read_summary): Read just single summary.
1708 (compute_working_sets): Use gcov_summary struct.
1709 * gcov-io.h (GCOV_TAG_SUMMARY_LENGTH): Remove usage
1710 of GCOV_COUNTERS_SUMMABLE.
1711 (GCOV_COUNTERS_SUMMABLE): Remove.
1712 (GCOV_FIRST_VALUE_COUNTER): Replace with
1713 GCOV_COUNTER_V_INTERVAL.
1714 (struct gcov_ctr_summary): Remove.
1715 (struct gcov_summary): Directly use fields of former
1717 (compute_working_sets): Use gcov_summary struct.
1718 * gcov.c (read_count_file): Do not use ctrs fields.
1719 * lto-cgraph.c (merge_profile_summaries): Use gcov_summary
1721 * lto-streamer.h (struct GTY): Make profile_info gcov_summary
1723 * profile.c: Likewise.
1724 * profile.h: Likewise.
1726 2018-06-05 Martin Liska <mliska@suse.cz>
1728 PR gcov-profile/84846
1729 * gcov.c (output_lines): Print working directory only
1730 in intermediate format.
1732 2018-06-05 Andreas Krebbel <krebbel@linux.ibm.com>
1734 * config/s390/s390-builtin-types.def: Add void function type.
1735 * config/s390/s390-builtins.def: Use the function type for the
1738 2018-06-04 Jim Wilson <jimw@sifive.com>
1740 * config/riscv/riscv-protos.h (riscv_expand_epilogue): Change bool arg
1742 * config/riscv/riscv.c (riscv_for_each_saved_reg): New args epilogue
1743 and maybe_eh_return. Change regno to unsigned int. Use new args to
1744 handle EH_RETURN_DATA_REGNO registers properly.
1745 (riscv_expand_prologue): Pass new args to riscv_for_each_saved_reg.
1746 (riscv_expand_epilogue): Update comment. Change argument name and
1747 type. Update code to use new name and type. Pass new args to
1748 riscv_for_each_saved_reg. Only use EH_RETURN_STACKADJ_RTX when
1750 * config/riscv/riscv.md (NORMAL_RETURN): New.
1751 (SIBCALL_RETURN, EXCEPTION_RETURN): New.
1752 (epilogue, sibcall_epilogue): Update riscv_expand_epilogue arg.
1753 (eh_return): Call gen_eh_return_internal and emit barrier.
1754 (eh_return_internal): Call riscv_expand_epilogue.
1756 2018-06-04 Eric Botcazou <ebotcazou@adacore.com>
1758 * gimple-ssa-store-merging.c (struct merged_store_group): Move up
1759 bit_insertion field and declare can_be_merged_into method.
1760 (merged_store_group::can_be_merged_into): New method.
1761 (imm_store_chain_info::coalesce_immediate): Call it to decide whether
1762 consecutive non-overlapping stores can be merged. Turn MEM_REF stores
1763 into BIT_INSERT_EXPR stores if the group contains a non-MEM_REF store.
1765 2018-06-04 Richard Biener <rguenther@suse.de>
1767 PR tree-optimization/85955
1768 * builtins.c (fold_builtin_sincos): Convert pointers to
1769 destination to appropriate type before dereferencing.
1771 2018-06-04 Segher Boessenkool <segher@kernel.crashing.org>
1773 * config/rs6000/rs6000.md (abs<mode>2 for FLOAT128): Handle IFmode.
1775 2018-06-04 Richard Sandiford <richard.sandiford@linaro.org>
1777 * expr.c (expand_expr_real_1): Force the operand into memory if
1778 its TYPE_MODE is BLKmode and if there is no integer mode for
1779 the number of bits being extracted.
1781 2018-06-04 Jakub Jelinek <jakub@redhat.com>
1785 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
1786 Use vptestnm rather than vptestm in (=Yc,v,C) variant.
1788 2018-06-04 Richard Biener <rguenther@suse.de>
1790 * tree-cfgcleanup.c (cleanup_tree_cfg_1): Fold into...
1791 (cleanup_tree_cfg_noloop): ... single caller. Do
1792 start_recording_case_labels later.
1794 2018-06-04 Sebastian Peryt <sebastian.peryt@intel.com>
1796 * config/i386/cldemoteintrin.h: Change define from _X86INTRIN_H_INCLUDED
1797 to _IMMINTRIN_H_INCLUDED.
1798 * config/i386/pconfigintrin.h: Ditto.
1799 * config/i386/waitpkgintrin.h: Ditto.
1800 * config/i386/immintrin.h: Add includes for sgxintrin.h,
1801 pconfigintrin.h, waitpkgintrin.h and cldemoteintrin.h.
1802 * config/i386/x86intrin.h: Remove includes for mintrin.h, xmmintrin.h,
1803 emmintrin.h, pmmintrin.h, tmmintrin.h, smmintrin.h, wmmintrin.h,
1804 bmiintrin.h, bmi2intrin.h, lzcntintrin.h, sgxintrin.h, pconfigintrin.h,
1805 waitpkgintrin.h and cldemoteintrin.h.
1807 2018-06-04 Richard Biener <rguenther@suse.de>
1809 PR tree-optimization/86038
1810 * tracer.c (find_best_successor): Check probability for
1811 being initialized, bail out if not.
1813 2018-06-04 Richard Earnshaw <rearnsha@arm.com>
1816 * config/arm/arm-cpus.in (ALL_QUIRKS): Add xscale feature to the list
1817 of bits to ignore when comparing architectures.
1819 2018-06-04 Jakub Jelinek <jakub@redhat.com>
1821 PR tree-optimization/69615
1822 * fold-const.c (merge_ranges): If range1 is - [x, x] and x is the
1823 maximum or minimum of the type, try to merge it also as if
1824 range1 is + [-, x - 1] or + [x + 1, -].
1827 * tree.c (inchash::add_expr): Handle IDENTIFIER_NODE.
1829 2018-06-03 Eric Botcazou <ebotcazou@adacore.com>
1831 PR tree-optimization/86034
1832 * gimple-ssa-store-merging.c (output_merged_store): Convert the RHS to
1833 the unsigned bitfield type in a bit insertion sequence if it does not
1834 have a larger precision than the bitfield size.
1835 (process_store): Also bypass widening conversions for BIT_INSERT_EXPR.
1837 2018-06-03 Kito Cheng <kito.cheng@gmail.com>
1839 * config/nds32/nds32-peephole2.md: Add new patterns for code size.
1841 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
1843 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3J.
1844 * config/nds32/nds32.c (nds32_option_override): Consider ARCH_V3J.
1845 * config/nds32/nds32.h (TARGET_ISA_V3): Add ARCH_V3J checking.
1846 * config/nds32/nds32.opt (march): Add enum value Value(ARCH_V3J).
1848 2018-06-03 Chung-Ju Wu <jasonwucj@gmail.com>
1850 * common/config/nds32/nds32-common.c (nds32_option_optimization_table):
1851 Disable -fdelete-null-pointer-checks for ELF toolchain.
1853 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
1854 Kito Cheng <kito.cheng@gmail.com>
1856 * config.gcc (nds32*): Use nds32-linux.opt and nds32-elf.opt.
1857 (nds32le-*-*, nds32be-*-*): Integrate checking process.
1858 (nds32*-*-*): Add glibc and uclibc conditions.
1859 * common/config/nds32/nds32-common.c (nds32_except_unwind_info): New.
1860 (TARGET_EXCEPT_UNWIND_INFO): Define.
1861 * config/nds32/elf.h: New file.
1862 * config/nds32/linux.h: New file.
1863 * config/nds32/nds32-elf.opt: New file.
1864 * config/nds32/nds32-linux.opt: New file.
1865 * config/nds32/nds32-fp-as-gp.c
1866 (pass_nds32_fp_as_gp::gate): Consider TARGET_LINUX_ABI.
1867 * config/nds32/nds32.c (nds32_conditional_register_usage): Consider
1869 (nds32_asm_file_end): Ditto.
1870 (nds32_print_operand): Ditto.
1871 (nds32_insert_attributes): Ditto.
1872 (nds32_init_libfuncs): New function.
1873 (TARGET_HAVE_TLS): Define.
1874 (TARGET_INIT_LIBFUNCS): Define.
1875 * config/nds32/nds32.h (TARGET_DEFAULT_RELAX): Apply different relax
1877 (TARGET_ELF): Apply different mcmodel setting.
1878 (LINK_SPEC, LIB_SPEC, STARTFILE_SPEC, ENDFILE_SPEC): The content has
1879 been migrated into elf.h and linux.h files.
1880 * config/nds32/nds32.md (add_pc): Consider TARGET_LINUX_ABI.
1881 * config/nds32/nds32.opt (mvh): Consider TARGET_LINUX_ABI.
1882 (mcmodel): The content has been migrated into nds32-elf.opt and
1883 nds32-linux.opt files.
1884 * config/nds32/t-elf: New file.
1885 * config/nds32/t-linux: New file.
1887 2018-06-02 Chung-Ju Wu <jasonwucj@gmail.com>
1888 Shiva Chen <shiva0217@gmail.com>
1890 * config/nds32/constants.md (unspec_volatile_element): Add
1891 UNSPEC_VOLATILE_OMIT_FP_BEGIN and UNSPEC_VOLATILE_OMIT_FP_END.
1892 * config/nds32/nds32-fp-as-gp.c: New implementation of fp_as_gp
1894 * config/nds32/nds32-protos.h (nds32_naked_function_p): Declare.
1895 (make_pass_nds32_fp_as_gp): Declare.
1896 * config/nds32/nds32.c (nds32_register_passes): Add fp_as_gp as one
1898 (nds32_asm_function_end_prologue): Remove unused asm output.
1899 (nds32_asm_function_begin_epilogue): Remove unused asm output.
1900 (nds32_asm_file_start): Output necessary fp_as_gp information.
1901 (nds32_option_override): Adjust register usage.
1902 (nds32_expand_prologue): Consider fp_as_gp situation.
1903 (nds32_expand_prologue_v3push): Consider fp_as_gp situation.
1904 * config/nds32/nds32.md (prologue): Check fp_as_gp_p and naked_p.
1907 (simple_return): Ditto.
1908 (omit_fp_begin): Output special directive for fp_as_gp.
1909 (omit_fp_end): Output special directive for fp_as_gp.
1910 * config/nds32/nds32.opt (mfp-as-gp, mno-fp-as-gp, mforce-fp-as-gp,
1911 mforbid-fp-as-gp): New options.
1913 2018-06-01 Mark Wielaard <mark@klomp.org>
1915 * dwarf2out.c (dwarf2out_finish): Remove generation of
1916 DW_AT_loclists_base.
1918 2018-06-01 Eric Botcazou <ebotcazou@adacore.com>
1920 * gimple-ssa-store-merging.c: Include gimple-fold.h.
1921 (struct store_immediate_info): Document BIT_INSERT_EXPR stores.
1922 (struct merged_store_group): Add bit_insertion field.
1923 (dump_char_array): Use standard hexadecimal format.
1924 (merged_store_group::merged_store_group): Set bit_insertion to false.
1925 (merged_store_group::apply_stores): Use optimal buffer size. Deal
1926 with BIT_INSERT_EXPR stores. Move up code updating the mask and
1927 also print the mask in the dump file.
1928 (pass_store_merging::gate): Minor tweak.
1929 (imm_store_chain_info::coalesce_immediate): Fix wrong association
1930 of stores with groups in dump. Allow coalescing of BIT_INSERT_EXPR
1931 stores with INTEGER_CST stores.
1932 (count_multiple_uses) <BIT_INSERT_EXPR>: New case.
1933 (imm_store_chain_info::output_merged_store): Add try_bitpos variable
1934 and use it throughout. Generate bit insertion sequences if need be.
1935 (pass_store_merging::process_store): Remove redundant condition.
1936 Record stores from a SSA name to a bit-field with BIT_INSERT_EXPR.
1938 2018-06-01 Segher Boessenkool <segher@kernel.crashing.org>
1940 * config/rs6000/rs6000.c (rs6000_mangle_type): Change the mangling of
1941 the 128-bit floating point types. Fix function comment.
1943 2018-06-01 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
1945 * config/aarch64/aarch64-simd.md
1946 (aarch64_simd_vec_unpack<su>_lo_<mode>): Use UXTL and SXTL assembler
1948 (aarch64_simd_vec_unpack<su>_hi_<mode>): Use UXTL2 and SXTL2 assembler
1951 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
1953 PR tree-optimization/85989
1954 * gimple-ssa-backprop.c (backprop::m_visited_phis): New member
1956 (backprop::intersect_uses): Check it when deciding whether this
1957 is a backedge reference.
1958 (backprop::process_block): Add each phi to m_visited_phis
1959 after visiting it, then clear it at the end.
1961 2018-06-01 Richard Biener <rguenther@suse.de>
1963 * tree-vectorizer.h (vect_dr_stmt): New function.
1964 (vect_get_load_cost): Adjust.
1965 (vect_get_store_cost): Likewise.
1966 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence):
1967 Use vect_dr_stmt instead of DR_SMTT.
1968 (vect_record_base_alignments): Likewise.
1969 (vect_calculate_target_alignment): Likewise.
1970 (vect_compute_data_ref_alignment): Likewise and make static.
1971 (vect_update_misalignment_for_peel): Likewise.
1972 (vect_verify_datarefs_alignment): Likewise.
1973 (vector_alignment_reachable_p): Likewise.
1974 (vect_get_data_access_cost): Likewise. Pass down
1975 vinfo to vect_get_load_cost/vect_get_store_cost instead of DR.
1976 (vect_get_peeling_costs_all_drs): Likewise.
1977 (vect_peeling_hash_get_lowest_cost): Likewise.
1978 (vect_enhance_data_refs_alignment): Likewise.
1979 (vect_find_same_alignment_drs): Likewise.
1980 (vect_analyze_data_refs_alignment): Likewise.
1981 (vect_analyze_group_access_1): Likewise.
1982 (vect_analyze_group_access): Likewise.
1983 (vect_analyze_data_ref_access): Likewise.
1984 (vect_analyze_data_ref_accesses): Likewise.
1985 (vect_vfa_segment_size): Likewise.
1986 (vect_small_gap_p): Likewise.
1987 (vectorizable_with_step_bound_p): Likewise.
1988 (vect_prune_runtime_alias_test_list): Likewise.
1989 (vect_analyze_data_refs): Likewise.
1990 (vect_supportable_dr_alignment): Likewise.
1991 * tree-vect-loop-manip.c (get_misalign_in_elems): Likewise.
1992 (vect_gen_prolog_loop_niters): Likewise.
1993 * tree-vect-loop.c (vect_analyze_loop_2): Likewise.
1994 * tree-vect-patterns.c (vect_recog_bool_pattern): Do not
1996 (vect_recog_mask_conversion_pattern): Likewise.
1997 (vect_try_gather_scatter_pattern): Likewise.
1998 * tree-vect-stmts.c (vect_model_store_cost): Pass stmt_info
1999 to vect_get_store_cost.
2000 (vect_get_store_cost): Get stmt_info instead of DR.
2001 (vect_model_load_cost): Pass stmt_info to vect_get_load_cost.
2002 (vect_get_load_cost): Get stmt_info instead of DR.
2004 2018-06-01 Richard Biener <rguenther@suse.de>
2007 * gimple-fold.c (var_decl_component_p): Also allow offsetted
2008 vars wrapped in MEM_REFs.
2010 2018-06-01 Richard Sandiford <richard.sandiford@linaro.org>
2012 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
2013 Fix subreg tests so that we only return a choice between
2014 GENERAL_REGS and FP_REGS if the original classes included both.
2016 2018-06-01 Richard Biener <rguenther@suse.de>
2019 * tree-ssa-structalias.c (get_function_part_constraint):
2020 Handle NULL fi->decl.
2021 (find_func_aliases_for_call): Properly handle indirect
2022 fi from direct call.
2023 (find_func_clobbers): Likewise.
2024 (ipa_pta_execute): Likewise.
2025 (create_variable_info_for): For functions that are ifunc_resolver
2026 resolve to a varinfo that contains the result of the resolver call.
2027 (associate_varinfo_to_alias): Do not treat ifunc resolvers as
2030 2018-05-31 Michael Collison <michael.collison@arm.com>
2032 * config/aarch64/aarch64.md:
2033 (*fix_to_zero_extenddfdi2): New pattern.
2034 * gcc.target/aarch64/fix_extend1.c: New testcase.
2036 2018-05-31 Qing Zhao <qing.zhao@oracle.com>
2040 * builtins.c (expand_builtin): Add the handling of BUILT_IN_STRCMP_EQ
2041 and BUILT_IN_STRNCMP_EQ.
2042 * builtins.def: Add new builtins BUILT_IN_STRCMP_EQ and
2043 BUILT_IN_STRNCMP_EQ.
2044 * gimple-fold.c (gimple_fold_builtin_string_compare): Add the
2045 handling of BUILTIN_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
2046 (gimple_fold_builtin): Likewise.
2047 * tree-ssa-strlen.c (compute_string_length): New function.
2048 (determine_min_obsize): New function.
2049 (handle_builtin_string_cmp): New function to handle calls to
2050 string compare functions.
2051 (strlen_optimize_stmt): Add handling to builtin string compare
2053 * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
2054 Add the handling of BUILT_IN_STRCMP_EQ and BUILT_IN_STRNCMP_EQ.
2055 * tree.c (build_common_builtin_nodes): Add new defines of
2056 BUILT_IN_STRNCMP_EQ and BUILT_IN_STRCMP_EQ.
2058 2018-05-31 Jakub Jelinek <jakub@redhat.com>
2061 * bb-reorder.c (pass_partition_blocks::gate): Return false for
2062 functions with naked attribute.
2064 2018-05-31 Uros Bizjak <ubizjak@gmail.com>
2066 * config/i386/sse.md (avx_vec_concat<mode>):
2067 Substitute concat_tg_mode mode attribute with xtg_mode.
2068 (<mask_codefor>avx512dq_broadcast<mode><mask_name>_1): Ditto.
2069 (concat_tg_mode): Remove mode attribute.
2071 2018-05-31 Martin Sebor <msebor@redhat.com>
2074 * calls.c (alloc_max_size): Correct a logic error/typo.
2075 Treat excessive arguments as infinite. Warn for invalid arguments.
2076 * doc/invoke.texi (-Walloc-size-larger-than): Update.
2078 2018-05-31 H.J. Lu <hongjiu.lu@intel.com>
2081 * config/i386/x86-tune.def: Re-enable partial_reg_dependency
2082 and movx for Haswell.
2084 2018-05-31 Chung-Lin Tang <cltang@codesourcery.com>
2085 Cesar Philippidis <cesar@codesourcery.com>
2088 * gimplify.c (gimplify_adjust_omp_clauses): Add 'remove = true'
2089 when emitting error on private/firstprivate reductions.
2090 * omp-low.c (lower_omp_target): Avoid reference-type processing
2091 on pointers for firstprivate clause.
2093 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2095 * config/aarch64/aarch64-simd-builtins.def (ld1x3): New.
2098 * config/aarch64/aarch64-simd.md
2099 (aarch64_ld1x3<VALLDIF:mode>): New pattern.
2100 (aarch64_ld1_x3_<mode>): Likewise
2101 (aarch64_st1x2<VALLDIF:mode>): Likewise
2102 (aarch64_st1_x2_<mode>): Likewise
2103 (aarch64_st1x3<VALLDIF:mode>): Likewise
2104 (aarch64_st1_x3_<mode>): Likewise
2105 * config/aarch64/arm_neon.h (vld1_u8_x3): New function.
2106 (vld1_s8_x3): Likewise.
2107 (vld1_u16_x3): Likewise.
2108 (vld1_s16_x3): Likewise.
2109 (vld1_u32_x3): Likewise.
2110 (vld1_s32_x3): Likewise.
2111 (vld1_u64_x3): Likewise.
2112 (vld1_s64_x3): Likewise.
2113 (vld1_f16_x3): Likewise.
2114 (vld1_f32_x3): Likewise.
2115 (vld1_f64_x3): Likewise.
2116 (vld1_p8_x3): Likewise.
2117 (vld1_p16_x3): Likewise.
2118 (vld1_p64_x3): Likewise.
2119 (vld1q_u8_x3): Likewise.
2120 (vld1q_s8_x3): Likewise.
2121 (vld1q_u16_x3): Likewise.
2122 (vld1q_s16_x3): Likewise.
2123 (vld1q_u32_x3): Likewise.
2124 (vld1q_s32_x3): Likewise.
2125 (vld1q_u64_x3): Likewise.
2126 (vld1q_s64_x3): Likewise.
2127 (vld1q_f16_x3): Likewise.
2128 (vld1q_f32_x3): Likewise.
2129 (vld1q_f64_x3): Likewise.
2130 (vld1q_p8_x3): Likewise.
2131 (vld1q_p16_x3): Likewise.
2132 (vld1q_p64_x3): Likewise.
2133 (vst1_s64_x2): Likewise.
2134 (vst1_u64_x2): Likewise.
2135 (vst1_f64_x2): Likewise.
2136 (vst1_s8_x2): Likewise.
2137 (vst1_p8_x2): Likewise.
2138 (vst1_s16_x2): Likewise.
2139 (vst1_p16_x2): Likewise.
2140 (vst1_s32_x2): Likewise.
2141 (vst1_u8_x2): Likewise.
2142 (vst1_u16_x2): Likewise.
2143 (vst1_u32_x2): Likewise.
2144 (vst1_f16_x2): Likewise.
2145 (vst1_f32_x2): Likewise.
2146 (vst1_p64_x2): Likewise.
2147 (vst1q_s8_x2): Likewise.
2148 (vst1q_p8_x2): Likewise.
2149 (vst1q_s16_x2): Likewise.
2150 (vst1q_p16_x2): Likewise.
2151 (vst1q_s32_x2): Likewise.
2152 (vst1q_s64_x2): Likewise.
2153 (vst1q_u8_x2): Likewise.
2154 (vst1q_u16_x2): Likewise.
2155 (vst1q_u32_x2): Likewise.
2156 (vst1q_u64_x2): Likewise.
2157 (vst1q_f16_x2): Likewise.
2158 (vst1q_f32_x2): Likewise.
2159 (vst1q_f64_x2): Likewise.
2160 (vst1q_p64_x2): Likewise.
2161 (vst1_s64_x3): Likewise.
2162 (vst1_u64_x3): Likewise.
2163 (vst1_f64_x3): Likewise.
2164 (vst1_s8_x3): Likewise.
2165 (vst1_p8_x3): Likewise.
2166 (vst1_s16_x3): Likewise.
2167 (vst1_p16_x3): Likewise.
2168 (vst1_s32_x3): Likewise.
2169 (vst1_u8_x3): Likewise.
2170 (vst1_u16_x3): Likewise.
2171 (vst1_u32_x3): Likewise.
2172 (vst1_f16_x3): Likewise.
2173 (vst1_f32_x3): Likewise.
2174 (vst1_p64_x3): Likewise.
2175 (vst1q_s8_x3): Likewise.
2176 (vst1q_p8_x3): Likewise.
2177 (vst1q_s16_x3): Likewise.
2178 (vst1q_p16_x3): Likewise.
2179 (vst1q_s32_x3): Likewise.
2180 (vst1q_s64_x3): Likewise.
2181 (vst1q_u8_x3): Likewise.
2182 (vst1q_u16_x3): Likewise.
2183 (vst1q_u32_x3): Likewise.
2184 (vst1q_u64_x3): Likewise.
2185 (vst1q_f16_x3): Likewise.
2186 (vst1q_f32_x3): Likewise.
2187 (vst1q_f64_x3): Likewise.
2188 (vst1q_p64_x3): Likewise.
2190 2018-05-30 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2192 * config/msp430/msp430.c (msp430_output_labelref): Prepend
2193 user_label_prefix to name.
2195 * tree-core.h: Update comment about the format of NAME string
2196 passed to handler in attribute_spec.
2198 * config/msp430/msp430.md: Remove erroneous subreg expression from
2199 zero_extendqisi2 insn pattern. Remove msp430x ISA restriction on
2200 zero_extend{q,h}isi2.
2202 2018-05-30 Borislav Petkov <bp@suse.de>
2204 * doc/extend.texi: Document some architecture specific
2205 constraints and sort entries.
2207 2018-05-30 Martin Sebor <msebor@redhat.com>
2210 * builtins.c (expand_builtin_stpcpy_1): New function.
2211 (expand_builtin_stpcpy): Call it, and call maybe_warn_nonstring_arg
2212 only if the former succeeds.
2214 2018-05-31 Sameera Deshpande <sameera.deshpande@linaro.org>
2216 * config/aarch64/aarch64-cores.def (saphira) : Add support for ARMv8.4
2219 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2221 * doc/invoke.texi (-flinker-output): Document
2223 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2225 * passes.c (ipa_write_summaries): Only modify statements if body
2227 * cgraphunit.c (ipa_passes): Also produce intermeidate code when
2228 incrementally linking.
2229 (ipa_passes): Likewise.
2230 * lto-cgraph.c (lto_output_node): When incrementally linking do not
2231 pass down resolution info.
2232 * common.opt (flag_incremental_link): Update info.
2233 * gcc.c (plugin specs): Turn flinker-output=* to
2234 -plugin-opt=-linker-output-known
2235 * toplev.c (compile_file): Also cut compilation when doing incremental
2237 * flag-types. (enum lto_partition_model): Add
2238 LTO_LINKER_OUTPUT_NOLTOREL.
2239 (invoke.texi): Add -flinker-output docs.
2240 * ipa.c (symbol_table::remove_unreachable_nodes): Handle LTO incremental
2241 link same way as WPA; do not stream in dead initializers.
2243 * dwarf2out.c (dwarf2out_die_ref_for_decl,
2244 darf2out_register_external_decl): Support incremental link.
2246 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2248 * lto-opts.c (lto_write_options): Skip OPT_dumpdir, OPT_fresolution_.
2250 2018-05-30 Jan Hubicka <hubicka@ucw.cz>
2252 * lto-wrapper.c (debug_objcopy): Add rename parameter; pass
2253 it down to simple_object_copy_lto_debug_sections.
2254 (run_gcc): Determine incremental LTO link time and configure
2255 lto1 into non-wpa mode, disable renaming of debug sections.
2257 2018-05-30 Kelvin Nilsen <kelvin@gcc.gnu.org>
2259 * doc/extend.texi (PowerPC AltiVec Built-in Functions): Remove
2260 descriptions of various incorrectly documented functions.
2262 2018-05-30 Andre Vieira <andre.simoesdiasvieira@arm.com>
2265 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
2266 address check not strict.
2268 2018-05-30 Richard Biener <rguenther@suse.de>
2270 PR tree-optimization/85964
2271 * tracer.c (better_p): Drop initialized count check, we only
2272 call the function with initialized counts now.
2273 (find_best_successor): Do find a best edge if one
2274 has uninitialized count.
2275 (find_best_predecessor): Likewise. Do BB frequency check only
2276 if count is initialized.
2278 2017-05-30 Jackson Woodruff <jackson.woodruff@arm.com>
2280 * config/aarch64/aarch64.c (aarch64_host_wide_int_compare): New.
2281 (aarch64_ldrstr_offset_compare): New.
2282 (aarch64_operands_adjust_ok_for_ldpstp): Update to consider all
2283 load/store orderings.
2284 (aarch64_gen_adjusted_ldpstp): Likewise.
2286 2018-05-30 Wilco Dijkstra <wdijkstr@arm.com>
2288 * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
2289 Check for subset of GENERAL_REGS and FP_REGS.
2290 * config/aarch64/aarch64-simd.md (aarch64_get_lane): Increase cost of
2293 2018-05-30 Richard Sandiford <richard.sandiford@linaro.org>
2295 * alias.c (adjust_offset_for_component_ref): Use poly_int_tree_p
2296 and wi::to_poly_offset. Add the current offset and then check
2297 whether the sum fits, rather than using an unchecked addition of
2298 a checked term. Check for a shwi rather than a uhwi.
2299 * expr.c (get_bit_range): Use tree_to_poly_uint64.
2300 (store_constructor): Use poly_int_tree_p.
2301 (expand_expr_real_1): Likewise.
2302 * function.c (assign_temp): Likewise.
2303 * fold-const.c (const_binop): Use poly_int_tree_p and
2305 (fold_indirect_ref_1): Likewise. Use multiple_p to attempt an exact
2307 * ipa-icf-gimple.c (func_checker::compare_operand): Use
2308 to_poly_offset for MEM offsets.
2309 * ipa-icf.c (sem_variable::equals): Likewise.
2310 * stor-layout.c (compute_record_mode): Use poly_int_tree_p.
2311 * tree-ssa-sccvn.c (ao_ref_init_from_vn_reference): Use
2312 wi::to_poly_offset for BIT_FIELD_REF offsets.
2313 (vn_reference_maybe_forwprop_address): Use poly_int_tree_p and
2315 * var-tracking.c (emit_note_insn_var_location): Use
2316 tree_to_poly_uint64.
2318 2018-05-29 Jim Wilson <jimw@sifive.com>
2320 * config/riscv/riscv.c (riscv_interrupt_type): Fix comment typo.
2322 2018-05-29 Uros Bizjak <ubizjak@gmail.com>
2325 * config/i386/i386.md (l<rounding_insn><MODEF:mode><SWI48:mode>2):
2326 Enable for TARGET_SSE4_1 and generate rounds{s,d} and cvtts{s,d}2si{,q}
2328 (sse4_1_round<mode>2): Use nonimmediate_operand
2329 for operand 1 predicate.
2331 2018-05-29 Martin Sebor <msebor@redhat.com>
2332 Richard Biener <rguenther@suse.de>
2335 * calls.c (get_size_range): Call determine_value_range instead
2336 of get_value_range..
2337 * tree-vrp.h (determine_value_range): Declared new function.
2338 * tree-vrp.c (determine_value_range_1, determine_value_range): New.
2340 2018-05-29 Richard Biener <rguenther@suse.de>
2342 * tree-vect-data-refs.c (vect_preserves_scalar_order_p): Make
2343 sure to use non-pattern stmts for get_earlier_stmt arguments.
2344 * tree-vectorizer.h (get_earlier_stmt): Assert we do not get
2345 called on pattern stmts.
2346 (get_later_stmt): Likewise.
2348 2018-05-29 Martin Liska <mliska@suse.cz>
2350 PR gcov-profile/85759
2351 * doc/gcov.texi: Document GCOV_ERROR_FILE and GCOV_EXIT_AT_ERROR
2354 2018-05-29 Jakub Jelinek <jakub@redhat.com>
2356 * tree-cfg.c (verify_gimple_assign_unary): Add checking for
2358 (verify_gimple_assign_binary): Check TYPE_VECTOR_SUBPARTS for
2362 * tree.def (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
2363 VEC_PACK_FLOAT_EXPR): New tree codes.
2364 * tree-pretty-print.c (op_code_prio): Handle
2365 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR.
2366 (dump_generic_node): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
2367 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
2368 * tree-inline.c (estimate_operator_cost): Likewise.
2369 * gimple-pretty-print.c (dump_binary_rhs): Handle VEC_PACK_FLOAT_EXPR.
2370 * fold-const.c (const_binop): Likewise.
2371 (const_unop): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR and
2372 VEC_UNPACK_FIX_TRUNC_LO_EXPR.
2373 * tree-cfg.c (verify_gimple_assign_unary): Likewise.
2374 (verify_gimple_assign_binary): Handle VEC_PACK_FLOAT_EXPR.
2375 * cfgexpand.c (expand_debug_expr): Handle VEC_UNPACK_FIX_TRUNC_HI_EXPR,
2376 VEC_UNPACK_FIX_TRUNC_LO_EXPR and VEC_PACK_FLOAT_EXPR.
2377 * expr.c (expand_expr_real_2): Likewise.
2378 * optabs.def (vec_packs_float_optab, vec_packu_float_optab,
2379 vec_unpack_sfix_trunc_hi_optab, vec_unpack_sfix_trunc_lo_optab,
2380 vec_unpack_ufix_trunc_hi_optab, vec_unpack_ufix_trunc_lo_optab): New
2382 * optabs.c (expand_widen_pattern_expr): For
2383 VEC_UNPACK_FIX_TRUNC_HI_EXPR and VEC_UNPACK_FIX_TRUNC_LO_EXPR use
2384 sign from result type rather than operand's type.
2385 (expand_binop_directly): For vec_packu_float_optab and
2386 vec_packs_float_optab allow result type to be different from operand's
2388 * optabs-tree.c (optab_for_tree_code): Handle
2389 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
2390 VEC_PACK_FLOAT_EXPR. Formatting fixes.
2391 * tree-vect-generic.c (expand_vector_operations_1): Handle
2392 VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR and
2393 VEC_PACK_FLOAT_EXPR.
2394 * tree-vect-stmts.c (supportable_widening_operation): Handle
2396 (supportable_narrowing_operation): Handle FLOAT_EXPR.
2397 * config/i386/i386.md (fixprefix, floatprefix): New code attributes.
2398 * config/i386/sse.md (*float<floatunssuffix>v2div2sf2): Rename to ...
2399 (float<floatunssuffix>v2div2sf2): ... this. Formatting fix.
2400 (vpckfloat_concat_mode, vpckfloat_temp_mode, vpckfloat_op_mode): New
2402 (vec_pack<floatprefix>_float_<mode>): New expander.
2403 (vunpckfixt_mode, vunpckfixt_model, vunpckfixt_extract_mode): New mode
2405 (vec_unpack_<fixprefix>fix_trunc_lo_<mode>,
2406 vec_unpack_<fixprefix>fix_trunc_hi_<mode>): New expanders.
2407 * doc/md.texi (vec_packs_float_@var{m}, vec_packu_float_@var{m},
2408 vec_unpack_sfix_trunc_hi_@var{m}, vec_unpack_sfix_trunc_lo_@var{m},
2409 vec_unpack_ufix_trunc_hi_@var{m}, vec_unpack_ufix_trunc_lo_@var{m}):
2411 * doc/generic.texi (VEC_UNPACK_FLOAT_HI_EXPR,
2412 VEC_UNPACK_FLOAT_LO_EXPR): Fix pasto in description.
2413 (VEC_UNPACK_FIX_TRUNC_HI_EXPR, VEC_UNPACK_FIX_TRUNC_LO_EXPR,
2414 VEC_PACK_FLOAT_EXPR): Document.
2416 2018-05-29 Richard Biener <rguenther@suse.de>
2418 * tree-vectorizer.h (struct vec_info): Add stmt_vec_infos
2420 (stmt_vec_info_vec): Make pointer.
2421 (init_stmt_vec_info_vec): Remove.
2422 (free_stmt_vec_info_vec): Likewise.
2423 (set_stmt_vec_info_vec): New function.
2424 (free_stmt_vec_infos): Likewise.
2425 (vinfo_for_stmt): Adjust for stmt_vec_info_vec indirection.
2426 (set_vinfo_for_stmt): Likewise.
2427 (get_earlier_stmt): Likewise.
2428 (get_later_stmt): Likewise.
2429 * tree-vectorizer.c (stmt_vec_info_vec): Make pointer.
2430 (vec_info::vec_info): Allocate stmt_vec_infos and set the global.
2431 (vec_info::~vec_info): Free stmt_vec_infos.
2432 (vectorize_loops): Set the global stmt_vec_info_vec to NULL.
2433 Remove old init_stmt_vec_info_vec/free_stmt_vec_info_vec calls.
2434 (pass_slp_vectorize::execute): Likewise.
2435 * tree-vect-stmts.c (init_stmt_vec_info_vec): Remove.
2436 (free_stmt_vec_info_vec): Likewise.
2437 (set_stmt_vec_info_vec): New function.
2438 (free_stmt_vec_infos): Likewise.
2439 * tree-vect-loop.c (_loop_vec_info::~_loop_vec_info): Set
2440 the global stmt_vec_info_vec.
2441 * tree-parloops.c (gather_scalar_reductions): Use
2442 set_stmt_vec_info_vec/free_stmt_vec_infos and maintain a local
2445 2018-05-29 Richard Biener <rguenther@suse.de>
2447 * dominance.c (iterate_fix_dominators): Push/pop TV_DOMINANCE.
2449 2018-05-29 Martin Liska <mliska@suse.cz>
2450 David Malcolm <dmalcolm@redhat.com>
2452 * vec.c (test_reverse): New.
2453 (vec_c_tests): Add new test.
2454 * vec.h (vl_ptr>::reverse): New function.
2456 2018-05-29 Gerald Pfeifer <gerald@pfeifer.com>
2458 * config.gcc: Identify FreeBSD 3.x and 4.x as unsupported.
2460 * config/freebsd-spec.h (FBSD_LIB_SPEC): Only consider FreeBSD 5
2463 2018-05-28 Bernd Edlinger <bernd.edlinger@hotmail.de>
2465 * tree-dump.c (dump_node): Use splay_tree_delete_pointers.
2467 2018-05-28 Richard Biener <rguenther@suse.de>
2469 PR tree-optimization/85933
2470 * tree-vect-data-refs.c (vect_record_base_alignments): Only
2471 look at stmts marked as vectorizable.
2473 2018-05-28 Richard Biener <rguenther@suse.de>
2475 PR tree-optimization/85934
2476 * tree-vect-generic.c (expand_vector_operations_1): Hoist
2477 vector boolean check before scalar optimization.
2479 2018-05-28 Jakub Jelinek <jakub@redhat.com>
2481 * doc/invoke.texi (ARM Options): Use @item instead of @itemx
2484 2018-05-28 Mark Wielaard <mark@klomp.org>
2486 * dwarf2asm.c (dw2_asm_output_delta_uleb128): Add brackets around lab2
2487 if it is an expression containing a minus sign.
2489 2018-05-27 John David Anglin <danglin@gcc.gnu.org>
2491 * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Define to 0.
2493 2018-05-27 Paul Koning <ni1d@arrl.net>
2495 * config/pdp11/pdp11.md (truncsihi2): Remove.
2497 2018-05-27 Monk Chiang <sh.chiang04@gmail.com>
2498 Chung-Ju Wu <jasonwucj@gmail.com>
2500 * config/nds32/nds32-intrinsic.md (unaligned_storedi): Modify patterns
2502 (unaligned_store_dw): Ditto.
2503 * config/nds32/nds32-memory-manipulation.c
2504 (nds32_expand_movmemsi_loop_known_size): Refactoring implementation.
2505 (nds32_gen_dup_4_byte_to_word_value): Rename to ...
2506 (nds32_gen_dup_4_byte_to_word_value_aux): ... this.
2507 (emit_setmem_word_loop): Rename to ...
2508 (emit_setmem_doubleword_loop): ... this.
2509 (nds32_gen_dup_4_byte_to_word_value): New function.
2510 (nds32_gen_dup_8_byte_to_double_word_value): New function.
2511 (nds32_expand_setmem_loop): Refine implementation.
2512 (nds32_expand_setmem_loop_v3m): Ditto.
2513 * config/nds32/nds32-multiple.md (unaligned_store_update_base_dw): New
2516 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
2518 * config/nds32/nds32.md (bswapsi2, bswaphi2): New patterns.
2520 2018-05-27 Chung-Ju Wu <jasonwucj@gmail.com>
2522 * config/nds32/nds32.c (nds32_attribute_table): Add "no_prologue".
2523 (nds32_init_machine_status): Initialize machine->attr_naked_p and
2524 machine->attr_no_prologue_p.
2525 (nds32_compute_stack_frame): Check "naked" and "no_prologue" attributes.
2526 (nds32_naked_function_p): Handle "naked" and "no_prologue" attributes.
2527 (nds32_expand_epilogue): Consider attr_naked_p.
2528 (nds32_expand_epilogue_v3pop): Likewise.
2529 (nds32_can_use_return_insn): Likewise.
2530 * config/nds32/nds32.h (machine_function): Add attr_naked_p and
2531 attr_no_prologue_p fields.
2532 * config/nds32/nds32.opt (mret-in-naked-func): New option.
2534 2018-05-27 Jakub Jelinek <jakub@redhat.com>
2537 * config/i386/i386.md (fixunssuffix, floatunssuffix): New code
2539 * config/i386/sse.md
2540 (<floatsuffix>float<sseintvecmodelower><mode>2<mask_name><round_name>):
2542 (float<floatunssuffix><sseintvecmodelower><mode>2<mask_name><round_name>):
2544 (<floatsuffix>float<sselongvecmodelower><mode>2<mask_name><round_name>):
2546 (float<floatunssuffix><sselongvecmodelower><mode>2<mask_name><round_name>):
2548 (*<floatsuffix>floatv2div2sf2): Rename to ...
2549 (*float<floatunssuffix>v2div2sf2): ... this.
2550 (<floatsuffix>floatv2div2sf2_mask): Rename to ...
2551 (float<floatunssuffix>v2div2sf2_mask): ... this.
2552 (*<floatsuffix>floatv2div2sf2_mask_1): Rename to ...
2553 (*float<floatunssuffix>v2div2sf2_mask_1): ... this.
2554 (<fixsuffix>fix_truncv8dfv8si2<mask_name><round_saeonly_name>): Rename
2556 (fix<fixunssuffix>_truncv8dfv8si2<mask_name><round_saeonly_name>):
2558 (<fixsuffix>fix_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
2560 (fix<fixunssuffix>_trunc<mode><sseintvecmodelower>2<mask_name><round_saeonly_name>):
2562 (<fixsuffix>fix_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
2564 (fix<fixunssuffix>_trunc<mode><sselongvecmodelower>2<mask_name><round_saeonly_name>):
2566 (<fixsuffix>fix_truncv2sfv2di2<mask_name>): Rename to ...
2567 (fix<fixunssuffix>_truncv2sfv2di2<mask_name>): ... this.
2568 (vec_pack_ufix_trunc_<mode>): Use gen_fixuns_truncv8dfv8si2 instead of
2569 gen_ufix_truncv8dfv8si2.
2570 * config/i386/i386-builtin.def (__builtin_ia32_cvttpd2uqq256_mask,
2571 __builtin_ia32_cvttpd2uqq128_mask, __builtin_ia32_cvttps2uqq256_mask,
2572 __builtin_ia32_cvttps2uqq128_mask, __builtin_ia32_cvtuqq2ps256_mask,
2573 __builtin_ia32_cvtuqq2ps128_mask, __builtin_ia32_cvtuqq2pd256_mask,
2574 __builtin_ia32_cvtuqq2pd128_mask, __builtin_ia32_cvttpd2udq512_mask,
2575 __builtin_ia32_cvtuqq2ps512_mask, __builtin_ia32_cvtuqq2pd512_mask,
2576 __builtin_ia32_cvttps2uqq512_mask, __builtin_ia32_cvttpd2uqq512_mask):
2577 Use fixuns instead ufix or floatuns instead ufloat in CODE_FOR_ names.
2579 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
2583 * varasm.c (assemble_alias): Lookup ifunc attribute on error.
2585 2018-05-25 Jim Wilson <jimw@sifive.com>
2587 * config/riscv/riscv-protos.h (riscv_epilogue_uses): New.
2588 * config/riscv/riscv.c (struct machine_function): Add
2589 interrupt_handler_p and attribute_checked_p fields.
2590 (riscv_attribute_table): Add interrupt.
2591 (riscv_interrupt_type_p): New.
2592 (riscv_save_reg_p): Save extra regs for interrupt handler.
2593 (riscv_use_save_libcall): Return false for interrupt handler.
2594 (riscv_first_stack_step): Add forward declaration.
2595 (riscv_compute_frame_info): New local interrupt_save_t1. Set it
2596 for interrupt handler with large frame. Use it for saved reg list.
2597 (riscv_expand_prologue): Move flag_stack_usage_info support to
2598 eliminate duplication.
2599 (riscv_expand_epilogue): Generate mret for interrupt handler.
2600 (riscv_epilogue_uses): New.
2601 (riscv_can_use_return_insn): Return false for interrupt handler.
2602 (riscv_function_ok_for_sibcall): Likewise.
2603 (riscv_set_current_function): Add interrupt handler support.
2604 * config/riscv/riscv.h (EPILOGUE_USES): Call riscv_epilogue_uses.
2605 * config/riscv/riscv.md (UNSPECV_MRET): New.
2607 (riscv_frflags, riscv_fsflags): Use tab after opcode.
2609 * doc/extend.texi (RISC-V Function Attributes) <interrupt>: New.
2611 2018-05-25 Bill Schmidt <wschmidt@linux.ibm.com>
2613 PR tree-optimization/85712
2614 * gimple-ssa-strength-reduction.c (replace_one_candidate): Skip if
2615 this candidate has already been replaced in-situ by a copy.
2617 2018-05-25 Jason Merrill <jason@redhat.com>
2619 PR c++/80485 - inline function non-zero address.
2620 * symtab.c (nonzero_address): Check DECL_COMDAT.
2622 2018-05-25 Uros Bizjak <ubizjak@gmail.com>
2625 * config/alpha/alpha.md (ashlsi3): New insn pattern.
2626 (*ashlsi_se): Rename from *ashldi_se. Define as sign
2627 extension of SImode operation. Use const123_operand predicate.
2628 (*saddsi_1): Remove.
2629 (*saddl_se_1): Ditto.
2631 (*ssubl_se_1): Ditto.
2632 * config/alpha/predicates.md (const123_operand): New predicate.
2633 * config/alpha/constraints.md (P): Use IN_RANGE.
2635 2018-05-25 Richard Biener <rguenther@suse.de>
2637 * tree-ssa-alias.h (refs_may_alias_p): Add tbaa_p bool parameter,
2639 (ref_maybe_used_by_stmt_p): Likewise.
2640 (stmt_may_clobber_ref_p): Likewise.
2641 (stmt_may_clobber_ref_p_1): Likewise.
2642 * tree-ssa-alias.c (refs_may_alias_p): Add tbaa_p bool parameter
2644 (ref_maybe_used_by_stmt_p): Likewise.
2645 (stmt_may_clobber_ref_p): Likewise.
2646 (stmt_may_clobber_ref_p_1): Likewise.
2647 * tree-vect-data-refs.c (vect_slp_analyze_node_dependences): Use
2648 the alias oracle to disambiguate DRs with stmts DR analysis
2650 (vect_analyze_data_refs): Do not give up on not analyzable
2651 DRs for BB vectorization. Remove code truncating the dataref
2654 2018-05-25 Jakub Jelinek <jakub@redhat.com>
2657 * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>_1):
2658 Add (=Yk,v,C) variant using vptestm insn. Use TARGET_AVX512BW
2659 in test instead of TARGET_AVX512F for VI12_AVX512VL iterator.
2661 2018-05-25 Richard Biener <rguenther@suse.de>
2663 * tree-vect-data-refs.c (vect_find_stmt_data_reference): New
2664 function, combining stmt data ref gathering and fatal analysis
2666 (vect_analyze_data_refs): Remove now redudnant code and simplify.
2667 * tree-vect-loop.c (vect_get_datarefs_in_loop): Factor out from
2668 vect_analyze_loop_2 and use vect_find_stmt_data_reference.
2669 * tree-vect-slp.c (vect_slp_bb): Use vect_find_stmt_data_reference.
2670 * tree-vectorizer.h (vect_find_stmt_data_reference): Declare.
2672 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2674 PR tree-optimization/85720
2675 * tree-loop-distribution.c (break_alias_scc_partitions): Don't merge
2676 SCC if all partitions are builtins.
2677 (version_loop_by_alias_check): New parameter. Generate cancelable
2678 runtime alias check if all partitions are builtins.
2679 (distribute_loop): Update call to above function.
2681 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2683 * tree-outof-ssa.c (tree-ssa.h, tree-dfa.h): Include header files.
2684 (create_default_def, for_all_parms): Moved from tree-ssa-coalesce.c.
2685 (parm_default_def_partition_arg): Ditto.
2686 (set_parm_default_def_partition): Ditto.
2687 (get_parm_default_def_partitions): Ditto and make it static.
2688 (get_undefined_value_partitions): Ditto and make it static.
2689 (remove_ssa_form): Refactor call to init_var_map here.
2690 * tree-ssa-coalesce.c (build_ssa_conflict_graph): Support live range
2691 computation for loop region.
2692 (coalesce_partitions, compute_optimized_partition_bases): Ditto.
2693 (register_default_def): Delete.
2694 (for_all_parms, create_default_def): Move to tree-outof-ssa.c.
2695 (parm_default_def_partition_arg): Ditto.
2696 (set_parm_default_def_partition): Ditto.
2697 (get_parm_default_def_partitions): Ditto and make it static.
2698 (get_undefined_value_partitions): Ditto and make it static.
2699 (coalesce_with_default, coalesce_with_default): Update comment.
2700 (create_coalesce_list_for_region): New func factored out from
2701 create_outofssa_var_map.
2702 (populate_coalesce_list_for_outofssa): New func factored out from
2703 create_outofssa_var_map and coalesce_ssa_name.
2704 (create_outofssa_var_map): Delete.
2705 (coalesce_ssa_name): Refactor to support live range computation.
2706 * tree-ssa-coalesce.h (coalesce_ssa_name): Change decl.
2707 (get_parm_default_def_partitions): Delete.
2708 (get_undefined_value_partitions): Ditto.
2709 * tree-ssa-live.c (init_var_map, delete_var_map): Support live range
2710 computation for loop region.
2711 (new_tree_live_info, loe_visit_block): Ditto.
2712 (live_worklist, set_var_live_on_entry): Ditto.
2713 (calculate_live_on_exit, verify_live_on_entry): Ditto.
2714 * tree-ssa-live.h (struct _var_map): New fields.
2715 (init_var_map): Change decl.
2716 (region_contains_p): New.
2718 2018-05-25 Bin Cheng <bin.cheng@arm.com>
2720 * tree-ssa-live.h (live_merge_and_clear): Delete.
2722 2018-05-25 Richard Biener <rguenther@suse.de>
2725 * tree-dump.c (dequeue_and_dump): Remove access to removed
2726 operand 2 of a SWITCH_EXPR.
2728 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2730 * doc/sourcebuild.texi (vect_double_cond_arith): Include
2731 multiplication and division.
2732 * doc/md.texi (cond_mul@var{m}, cond_div@var{m}, cond_mod@var{m})
2733 (cond_udiv@var{m}, cond_umod@var{m}): Document.
2734 * optabs.def (cond_smul_optab, cond_sdiv_optab, cond_smod_optab)
2735 (cond_udiv_optab, cond_umod_optab): New optabs.
2736 * internal-fn.def (IFN_COND_MUL, IFN_COND_DIV, IFN_COND_MOD)
2737 (IFN_COND_RDIV): New internal functions.
2738 * internal-fn.c (get_conditional_internal_fn): Handle TRUNC_DIV_EXPR,
2739 TRUNC_MOD_EXPR and RDIV_EXPR.
2740 * match.pd (UNCOND_BINARY, COND_BINARY): Handle them.
2741 * config/aarch64/iterators.md (UNSPEC_COND_MUL, UNSPEC_COND_DIV):
2743 (SVE_INT_BINARY): Include mult.
2744 (SVE_COND_FP_BINARY): Include UNSPEC_MUL and UNSPEC_DIV.
2745 (optab, sve_int_op): Handle mult.
2746 (optab, sve_fp_op, commutative): Handle UNSPEC_COND_MUL and
2748 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New pattern
2749 for SVE_INT_BINARY_SD.
2751 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2753 * config/aarch64/iterators.md (SVE_INT_BINARY_SD): New code iterator.
2754 (optab, sve_int_op): Handle div and udiv.
2755 * config/aarch64/aarch64-sve.md (<optab><mode>3): New expander
2756 for SVE_INT_BINARY_SD.
2757 (*<optab><mode>3): New insn for the same.
2759 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2761 * tree-vect-patterns.c: Include predict.h.
2762 (vect_recog_divmod_pattern): Restrict check for division support
2763 to when optimizing for size.
2765 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2767 * doc/sourcebuild.texi (vect_double_cond_arith: Document.
2768 * gimple-match.h (gimple_match_op::MAX_NUM_OPS): Bump to 4.
2769 (gimple_match_op::gimple_match_op): Add an overload for 4 operands.
2770 (gimple_match_op::set_op): Likewise.
2771 (gimple_resimplify4): Declare.
2772 * genmatch.c (get_operand_type): Handle CFN_COND_* functions.
2773 (expr::gen_transform): Likewise.
2774 (decision_tree::gen): Generate a simplification routine for 4 operands.
2775 * gimple-match-head.c (gimple_simplify): Add an overload for
2776 4 operands. In the top-level function, handle up to 4 call
2777 arguments and call gimple_resimplify4.
2778 (gimple_resimplify4): New function.
2779 (build_call_internal): Pass a fourth operand.
2780 (maybe_push_to_seq): Likewise.
2781 * match.pd (UNCOND_BINARY, COND_BINARY): New operator lists.
2782 Fold VEC_COND_EXPRs of an operation and a default value into
2783 an IFN_COND_* function if possible.
2784 * config/aarch64/iterators.md (UNSPEC_COND_MAX, UNSPEC_COND_MIN):
2786 (SVE_COND_FP_BINARY): Include them.
2787 (optab, sve_fp_op): Handle them.
2788 (SVE_INT_BINARY_REV): New code iterator.
2789 (SVE_COND_FP_BINARY_REV): New int iterator.
2790 (commutative): New int attribute.
2791 * config/aarch64/aarch64-protos.h (aarch64_sve_prepare_conditional_op):
2793 * config/aarch64/aarch64.c (aarch64_sve_prepare_conditional_op): New
2795 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Use it.
2796 (*cond_<optab><mode>): New patterns for reversed operands.
2798 2018-05-25 Richard Biener <rguenther@suse.de>
2800 * tree-vectorizer.h (STMT_VINFO_GROUP_*, GROUP_*): Remove.
2801 (DR_GROUP_*): New, assert we have non-NULL ->data_ref_info.
2802 (REDUC_GROUP_*): New, assert we have NULL ->data_ref_info.
2803 (STMT_VINFO_GROUPED_ACCESS): Adjust.
2804 * tree-vect-data-refs.c (everywhere): Adjust users.
2805 * tree-vect-loop.c (everywhere): Likewise.
2806 * tree-vect-slp.c (everywhere): Likewise.
2807 * tree-vect-stmts.c (everywhere): Likewise.
2808 * tree-vect-patterns.c (vect_reassociating_reduction_p): Likewise.
2810 2018-05-25 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
2812 * configure.ac (gcc_cv_as_section_has_e): Move to common section.
2814 (gcc_cv_as_section_exclude): ... this.
2815 Try Solaris as #exclude syntax.
2816 * configure: Regenerate.
2817 * config.in: Regenerate.
2818 * config/i386/i386.c (i386_solaris_elf_named_section): Handle
2820 * config/sparc/sparc.c (sparc_solaris_elf_asm_named_section)
2821 [HAVE_GAS_SECTION_EXCLUDE]: Handle SECTION_EXCLUDE.
2823 * varasm.c (default_elf_asm_named_section): Don't check if
2824 HAVE_GAS_SECTION_EXCLUDE is defined.
2826 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2828 * doc/md.texi: Update the documentation of the cond_* optabs
2829 to mention the new final operand. Fix GET_MODE_NUNITS call.
2830 Describe the scalar case too.
2831 * internal-fn.def (IFN_EXTRACT_LAST): Change type to fold_left.
2832 * internal-fn.c (expand_cond_unary_optab_fn): Expect 3 operands
2834 (expand_cond_binary_optab_fn): Expect 4 operands instead of 3.
2835 (get_conditional_internal_fn): Update comment.
2836 * tree-vect-loop.c (vectorizable_reduction): Pass the original
2837 accumulator value as a final argument to conditional functions.
2838 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): Turn into
2839 a define_expand and add an "else" operand. Assert for now that
2840 the else operand is equal to operand 2. Use SVE_INT_BINARY and
2841 SVE_COND_FP_BINARY instead of SVE_COND_INT_OP and SVE_COND_FP_OP.
2842 (*cond_<optab><mode>): New patterns.
2843 * config/aarch64/iterators.md (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX)
2844 (UNSPEC_COND_SMIN, UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
2845 (UNSPEC_COND_EOR): Delete.
2846 (optab): Remove associated mappings.
2847 (SVE_INT_BINARY): New code iterator.
2848 (sve_int_op): Remove int attribute and add "minus" to the code
2850 (SVE_COND_INT_OP): Delete.
2851 (SVE_COND_FP_OP): Rename to...
2852 (SVE_COND_FP_BINARY): ...this.
2854 2018-05-25 Richard Sandiford <richard.sandiford@linaro.org>
2856 * optabs.c (can_reuse_operands_p): New function.
2857 (maybe_legitimize_operands): Try to reuse the results for
2860 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
2862 * config/i386/sse.md (cvtusi2<ssescalarmodesuffix>64<round_name>):
2863 Add {q} suffix to insn mnemonic.
2865 2018-05-23 Jozef Lawrynowicz <jozef.l@mittosystems.com>
2867 * config/msp430/msp430.c (TARGET_WARN_FUNC_RETURN): Define.
2868 (msp430_warn_func_return): New.
2870 2018-05-24 Roger Sayle <roger@nextmovesoftware.com>
2872 * fold-const.c (tree_nonzero_bits): New function.
2873 * fold-const.h (tree_nonzero_bits): Likewise.
2874 * match.pd (POPCOUNT): New patterns to fold BUILTIN_POPCOUNT and
2875 friends. POPCOUNT(x&1) => x&1, POPCOUNT(x)==0 => x==0, etc.
2877 2018-05-24 H.J. Lu <hongjiu.lu@intel.com>
2881 * varasm.c (assemble_alias): Check ifunc_resolver only on
2884 2018-05-24 Uros Bizjak <ubizjak@gmail.com>
2887 * config/i386/sse.md (movdi_to_sse): Do not generate pseudo
2888 when memory input operand is handled.
2890 2018-05-24 Luis Machado <luis.machado@linaro.org>
2892 * config/aarch64/aarch64.c (qdf24xx_addrcost_table): New static
2894 (qdf24xx_tunings) <addr_costs>: Set to qdf24xx_addrcost_table.
2896 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
2898 * match.pd: Delay FMA folds until after vectorization.
2900 2018-05-24 Andre Vieira <andre.simoesdiasvieira@arm.com>
2903 * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): Make
2904 address check not strict.
2906 2018-05-24 Richard Sandiford <richard.sandiford@linaro.org>
2908 * gimple-match.h (gimple_match_op): New class.
2909 (mprts_hook): Replace parameters with a gimple_match_op *.
2910 (maybe_build_generic_op): Likewise.
2911 (gimple_simplified_result_is_gimple_val): Replace parameters with
2912 a const gimple_match_op *.
2913 (gimple_simplify): Replace code_helper * and tree * parameters with
2914 a gimple_match_op * parameter.
2915 (gimple_resimplify1): Replace code_helper *, tree and tree *
2916 parameters with a gimple_match_op * parameter.
2917 (gimple_resimplify2): Likewise.
2918 (gimple_resimplify3): Likewise.
2919 (maybe_push_res_to_seq): Replace code_helper, tree and tree *
2920 parameters with a gimple_match_op * parameter.
2921 * gimple-match-head.c (gimple_simplify): Change prototypes of
2922 auto-generated functions to take a gimple_match_op * instead of
2923 separate code_helper * and tree * parameters. Make the same
2924 change in the top-level overload and update calls to the
2925 gimple_resimplify routines. Update calls to the auto-generated
2926 functions and to maybe_push_res_to_seq in the publicly-facing
2927 operation-specific gimple_simplify overloads.
2928 (gimple_match_op::MAX_NUM_OPS): Define.
2929 (gimple_resimplify1): Replace rcode and ops with a single res_op
2930 parameter. Update call to gimple_simplify.
2931 (gimple_resimplify2): Likewise.
2932 (gimple_resimplify3): Likewise.
2933 (mprts_hook): Replace parameters with a gimple_match_op *.
2934 (maybe_build_generic_op): Likewise.
2935 (build_call_internal): Replace type, nargs and ops with
2936 a gimple_match_op *.
2937 (maybe_push_res_to_seq): Replace res_code, type and ops parameters
2938 with a single gimple_match_op *. Update calls to mprts_hook,
2939 build_call_internal and gimple_simplified_result_is_gimple_val.
2940 Factor out code that is common to the tree_code and combined_fn cases.
2941 * genmatch.c (expr::gen_transform): Replace tem_code and
2942 tem_ops with a gimple_match_op called tem_op. Update calls
2943 to the gimple_resimplify functions and maybe_push_res_to_seq.
2944 (dt_simplify::gen_1): Manipulate res_op instead of res_code and
2945 res_ops. Update call to the gimple_resimplify functions.
2946 (dt_simplify::gen): Pass res_op instead of res_code and res_ops.
2947 (decision_tree::gen): Make the functions take a gimple_match_op *
2948 called res_op instead of separate res_code and res_ops parameters.
2949 Update call accordingly.
2950 * gimple-fold.c (replace_stmt_with_simplification): Replace rcode
2951 and ops with a single res_op parameter. Update calls to
2952 maybe_build_generic_op and maybe_push_res_to_seq.
2953 (fold_stmt_1): Update calls to gimple_simplify and
2954 replace_stmt_with_simplification.
2955 (gimple_fold_stmt_to_constant_1): Update calls to gimple_simplify
2956 and gimple_simplified_result_is_gimple_val.
2957 * tree-cfgcleanup.c (cleanup_control_expr_graph): Update call to
2959 * tree-ssa-sccvn.c (vn_lookup_simplify_result): Replace parameters
2960 with a gimple_match_op *.
2961 (vn_nary_build_or_lookup): Likewise. Update call to
2962 vn_nary_build_or_lookup_1.
2963 (vn_nary_build_or_lookup_1): Replace rcode, type and ops with a
2964 gimple_match_op *. Update calls to the gimple_resimplify routines
2965 and to gimple_simplified_result_is_gimple_val.
2966 (vn_nary_simplify): Update call to vn_nary_build_or_lookup_1.
2967 Use gimple_match_op::MAX_NUM_OPS instead of a hard-coded 3.
2968 (vn_reference_lookup_3): Update call to vn_nary_build_or_lookup.
2969 (visit_nary_op): Likewise.
2970 (visit_reference_op_load): Likewise.
2972 2018-05-23 Luis Machado <luis.machado@linaro.org>
2974 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Use correct type
2975 modifier for printing the step amount.
2977 2018-05-23 Jozef Lawrynowicz <jozef.l@somniumtech.com>
2980 * gcc/tree.c (build_common_tree_nodes): Dont set TYPE_SIZE for __intN
2983 2018-05-23 Segher Boessenkool <segher@kernel.crashing.org>
2985 * doc/sourcebuild.texi (Endianness): New subsubsection.
2987 2018-05-23 Luis Machado <luis.machado@linaro.org>
2989 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
2990 <prefetch_dynamic_strides>: New const bool field.
2991 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
2992 prefetch_dynamic_strides.
2993 (exynosm1_prefetch_tune): Likewise.
2994 (thunderxt88_prefetch_tune): Likewise.
2995 (thunderx_prefetch_tune): Likewise.
2996 (thunderx2t99_prefetch_tune): Likewise.
2997 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides to
2999 (aarch64_override_options_internal): Update to set
3000 PARAM_PREFETCH_DYNAMIC_STRIDES.
3001 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
3002 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
3003 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
3004 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
3005 prefetch-dynamic-strides setting.
3007 2018-05-23 Luis Machado <luis.machado@linaro.org>
3009 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
3010 <minimum_stride>: New const int field.
3011 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
3012 minimum_stride field defaulting to -1.
3013 (exynosm1_prefetch_tune): Likewise.
3014 (thunderxt88_prefetch_tune): Likewise.
3015 (thunderx_prefetch_tune): Likewise.
3016 (thunderx2t99_prefetch_tune): Likewise.
3017 (qdf24xx_prefetch_tune) <minimum_stride>: Set to 2048.
3018 <default_opt_level>: Set to 3.
3019 (aarch64_override_options_internal): Update to set
3020 PARAM_PREFETCH_MINIMUM_STRIDE.
3021 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
3022 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
3023 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
3024 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
3025 stride is constant and is below the minimum stride threshold.
3027 2018-05-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3029 * config/arm/arm-cpus.in (mode26): Delete.
3030 (armv4): Delete mode26 reference.
3031 * config/arm/arm.c (arm_configure_build_target): Delete use of
3034 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
3036 * config/i386/i386.md (*floatuns<SWI48:mode><MODEF:mode>2_avx512):
3038 (floatunssi<mode>2): Also enable for AVX512F and TARGET_SSE_MATH.
3039 Rewrite expander pattern. Emit gen_floatunssi<mode>2_i387_with_xmm
3041 (floatunsdisf2): Rewrite expander pattern. Hanlde TARGET_AVX512F.
3042 (floatunsdidf2): Ditto.
3044 2018-05-23 Uros Bizjak <ubizjak@gmail.com>
3046 * config/i386/i386.md (fixuns_trunc<mode>di2): New insn pattern.
3047 (fixuns_trunc<mode>si2_avx512f): Ditto.
3048 (*fixuns_trunc<mode>si2_avx512f_zext): Ditto.
3049 (fixuns_trunc<mode>si2): Also enable for AVX512F and TARGET_SSE_MATH.
3050 Emit fixuns_trunc<mode>si2_avx512f for AVX512F targets.
3052 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
3054 PR rtl-optimization/79985
3055 * df-scan.c (df_insn_refs_collect): Remove special case for
3056 global registers and asm statements.
3058 2018-05-23 Alexander Monakov <amonakov@ispras.ru>
3060 * extend.texi (Global Register Variables): Rewrite the bullet list.
3061 Note that the register is available for allocation. Note that access
3062 via inline asm must use constraints. Add note about async-signal
3063 handlers. Remove paragraph about automagic register selection.
3065 2018-05-23 Richard Biener <rguenther@suse.de>
3067 * tree-ssa-sccvn.c (vn_reference_lookup_3): Remove restriction
3068 of fixed offset from memset VN.
3070 2018-05-23 Bill Schmidt <wschmidt@linux.ibm.com>
3072 * gimple-ssa-strength-reduction.c (struct slsr_cand_d): Add
3074 (alloc_cand_and_find_basis): Initialize first_interp field.
3075 (slsr_process_mul): Modify first_interp field.
3076 (slsr_process_add): Likewise.
3077 (slsr_process_cast): Modify first_interp field for each new
3079 (slsr_process_copy): Likewise.
3080 (dump_candidate): Dump first_interp field.
3081 (replace_mult_candidate): Process all interpretations, not just
3083 (replace_rhs_if_not_dup): Likewise.
3084 (replace_one_candidate): Likewise.
3086 2018-05-23 Wilco Dijkstra <wdijkstr@arm.com>
3088 * config/aarch64/aarch64.c (aarch64_use_frame_pointer):
3090 (aarch64_needs_frame_chain): New function.
3091 (aarch64_parse_override_string): Set aarch64_use_frame_pointer.
3093 2018-05-23 Sudakshina Das <sudi.das@arm.com>
3096 * common/config/aarch64/aarch64-common.c (aarch64_handle_option):
3097 Check val before adding MASK_STRICT_ALIGN to opts->x_target_flags.
3098 * config/aarch64/aarch64.opt (mstrict-align): Remove RejectNegative.
3099 * config/aarch64/aarch64.c (aarch64_attributes): Mark allow_neg
3100 as true for strict-align.
3101 (aarch64_can_inline_p): Perform checks even when callee has no
3102 attributes to check for strict alignment.
3103 * doc/extend.texi (AArch64 Function Attributes): Document
3105 * doc/invoke.texi: (AArch64 Options): Likewise.
3107 2018-05-23 Richard Sandiford <richard.sandiford@linaro.org>
3109 PR tree-optimization/85853
3110 * tree-vect-slp.c (vect_slp_analyze_node_operations): Split out
3111 the handling of the root of the node to...
3112 (vect_slp_analyze_node_operations_1): ...this new function,
3113 and run the whole thing with the child nodes' def types
3114 set according to their SLP node's def type.
3116 2018-05-23 Richard Biener <rguenther@suse.de>
3119 * tree-data-ref.c (create_runtime_alias_checks): Defer
3120 and ignore overflow warnings.
3122 2018-05-23 Yury Gribov <tetra2005@gmail.com>
3124 PR tree-optimization/85822
3125 * tree-vrp.c (is_masked_range_test): Fix handling of negative
3128 2018-05-23 Richard Biener <rguenther@suse.de>
3130 * tree-ssa-sccvn.c (vn_reference_lookup_3): Handle arbitrary
3131 memset constants via native_interpret_expr.
3133 2018-05-22 H.J. Lu <hongjiu.lu@intel.com>
3136 * cgraph.h (cgraph_node::create): Set ifunc_resolver for ifunc
3138 (cgraph_node::create_alias): Likewise.
3139 (cgraph_node::get_availability): Check ifunc_resolver instead
3140 of looking up ifunc attribute.
3141 * cgraphunit.c (maybe_diag_incompatible_alias): Likewise.
3142 * varasm.c (do_assemble_alias): Likewise.
3143 (assemble_alias): Likewise.
3144 (default_binds_local_p_3): Likewise.
3145 * cgraph.h (cgraph_node): Add ifunc_resolver.
3146 (cgraph_node::only_called_directly_or_aliased_p): Return false
3148 * lto-cgraph.c (input_node): Set ifunc_resolver for ifunc
3150 * symtab.c (symtab_node::verify_base): Verify that ifunc_resolver
3151 is equivalent to lookup_attribute ("ifunc", DECL_ATTRIBUTES (decl)).
3152 (symtab_node::binds_to_current_def_p): Check ifunc_resolver
3153 instead of looking up ifunc attribute.
3155 2018-05-22 Luis Machado <luis.machado@linaro.org>
3157 * config/aarch64/aarch64.md (*ashift<mode>_extv_bfiz): New pattern.
3159 2018-05-22 Martin Sebor <msebor@redhat.com>
3162 * builtins.c (expand_builtin_strcpy): Call maybe_warn_nonstring_arg
3163 only when expasion succeeds.
3164 (expand_builtin_strcmp): Same.
3165 (expand_builtin_strncmp): Same.
3167 2018-05-22 Martin Sebor <msebor@redhat.com>
3169 * calls.c (maybe_warn_nonstring_arg): Fix a typo in a comment.
3171 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3172 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3174 * config/aarch64/aarch64-ldpstp.md: Replace uses of
3175 aarch64_mem_pair_operand with memory_operand and delete operand swapping
3177 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3178 Add check for legitimate_address.
3179 (aarch64_gen_adjusted_ldpstp): Swap operands where appropriate.
3180 (aarch64_swap_ldrstr_operands): New.
3181 * config/aarch64/aarch64-protos.h (aarch64_swap_ldrstr_operands):
3184 2018-05-22 Jackson Woodruff <jackson.woodruff@arm.com>
3185 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3187 * config/aarch64/aarch64.md: New patterns to generate stp
3189 (store_pair_sw, store_pair_dw): New patterns to generate stp for
3190 single words and double words.
3191 (load_pair_sw, load_pair_dw): Likewise.
3192 (store_pair_sf, store_pair_df, store_pair_si, store_pair_di):
3194 (load_pair_sf, load_pair_df, load_pair_si, load_pair_di):
3196 * config/aarch64/aarch64-ldpstp.md: Modify peephole
3197 for different mode ldpstp and add peephole for merged zero stores.
3199 * config/aarch64/aarch64.c (aarch64_operands_ok_for_ldpstp):
3201 (aarch64_gen_store_pair): Rename calls to match new patterns.
3202 (aarch64_gen_load_pair): Rename calls to match new patterns.
3203 * config/aarch64/aarch64-simd.md (load_pair<mode>): Rename to...
3204 (load_pair<DREG:mode><DREG2:mode>): ... This.
3205 (store_pair<mode>): Rename to...
3206 (vec_store_pair<DREG:mode><DREG2:mode>): ... This.
3207 * config/aarch64/iterators.md (DREG, DREG2, DX2, SX, SX2, DSX):
3209 (V_INT_EQUIV): Handle SImode.
3210 * config/aarch64/predicates.md (aarch64_reg_zero_or_fp_zero):
3213 2018-05-22 Martin Sebor <msebor@redhat.com>
3216 * calls.c (maybe_warn_nonstring_arg): Use string length to set
3217 or ajust the presumed bound on an operation to avoid unnecessary
3220 2018-05-22 Martin Sebor <msebor@redhat.com>
3222 PR tree-optimization/85826
3223 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Avoid
3224 assuming that a DECL necesarily has a constant size.
3226 2018-05-22 Richard Sandiford <richard.sandiford@linaro.org>
3229 * internal-fn.c (expand_direct_optab_fn): Cope with a null lhs.
3231 2018-05-22 Richard Biener <rguenther@suse.de>
3233 PR tree-optimization/85834
3234 * tree-ssa-sccvn.c (vn_reference_lookup_3): Properly handle
3235 non-constant and non-zero memset arguments.
3237 2018-05-22 Martin Liska <mliska@suse.cz>
3240 * ipa-icf.c (sem_item::add_type): Do not ICE for incomplete types.
3242 2018-05-22 Richard Biener <rguenther@suse.de>
3244 PR tree-optimization/85863
3245 * tree-vect-stmts.c (vect_is_simple_cond): Only widen invariant
3246 comparisons when vectype is specified.
3247 (vectorizable_condition): Do not specify vectype for
3248 vect_is_simple_cond when SLP vectorizing.
3250 2018-05-21 Michael Meissner <meissner@linux.ibm.com>
3253 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Do not
3254 define __ibm128 as long double.
3255 * config/rs6000/rs6000.c (rs6000_init_builtins): Create __ibm128
3256 as a distinct type when IEEE 128-bit support is enabled.
3257 (init_float128_ieee): Fix up conversions between IFmode and IEEE
3258 128-bit types to use the correct functions.
3259 (rs6000_expand_float128_convert): Use explicit FLOAT_EXTEND to
3260 convert between 128-bit floating point types that have different
3261 modes but the same representation, instead of using gen_lowpart to
3263 * config/rs6000/rs6000.md (IFKF): New iterator for IFmode and
3265 (IFKF_reg): New attributes to give the register constraints for
3267 (extend<mode>tf2_internal): New insns to mark an explicit
3268 conversion between 128-bit floating point types that have a
3269 different mode but share the same representation.
3271 2018-05-21 Richard Sandiford <richard.sandiford@linaro.org>
3273 PR tree-optimization/85814
3274 * tree-ssa-strlen.c (get_stridx_plus_constant): Cope with
3275 a null return from get_strinfo when unsharing the next
3276 strinfo in the chain.
3278 2018-05-21 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
3281 * varasm.c (weak_finish): Clean up weak_decls.
3283 2018-05-21 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3285 * config/aarch64/aarch64.md ("unspec"): Define UNSPEC_SABAL,
3286 UNSPEC_SABDL2, UNSPEC_SADALP, UNSPEC_UABAL, UNSPEC_UABDL2,
3287 UNSPEC_UADALP values.
3288 * config/aarch64/iterators.md (ABAL): New int iterator.
3291 (sur): Add mappings for the above.
3292 * config/aarch64/aarch64-simd.md (aarch64_<sur>abdl2<mode>_3):
3294 (aarch64_<sur>abal<mode>_4): Likewise.
3295 (aarch64_<sur>adalp<mode>_3): Likewise.
3296 (<sur>sadv16qi): New define_expand.
3298 2018-05-21 Alexander Nesterovskiy <alexander.nesterovskiy@intel.com>
3300 * config/i386/i386.md (*movsf_internal): AVX falsedep fix.
3301 (*movdf_internal): Ditto.
3302 (*rcpsf2_sse): Ditto.
3303 (*rsqrtsf2_sse): Ditto.
3304 (*sqrt<mode>2_sse): Ditto.
3306 2018-05-21 Tamar Christina <tamar.christina@arm.com>
3308 * config/aarch64/aarch64-simd.md (aarch64_eor3qv8hi): Change to
3310 (aarch64_bcaxqv8hi): Change to bcaxq<mode>4.
3311 * config/aarch64/aarch64-simd-builtins.def (veor3q_u8, veor3q_u32,
3312 veor3q_u64, veor3q_s8, veor3q_s16, veor3q_s32, veor3q_s64, vbcaxq_u8,
3313 vbcaxq_u32, vbcaxq_u64, vbcaxq_s8, vbcaxq_s16, vbcaxq_s32,
3315 * config/aarch64/arm_neon.h: Likewise.
3316 * config/aarch64/iterators.md (VQ_I): New.
3318 2018-05-21 Alexey Brodkin <abrodkin@synopsys.com>
3320 * config.gcc: Add arc/t-multilib-linux to tmake_file for
3322 * config/arc/t-multilib-linux: Specify MULTILIB_OPTIONS and
3325 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
3327 * config/nds32/constraints.md (S): New constraint.
3328 * config/nds32/nds32.md (call_internal): Use constraint S.
3329 (call_value_internal): Likewise.
3330 (sibcall_internal): Likewise.
3331 (sibcall_value_internal): Likewise.
3333 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
3334 Chung-Ju Wu <jasonwucj@gmail.com>
3336 * config/nds32/nds32.c (nds32_register_move_cost): Take garywolf cpu
3339 2018-05-20 Kito Cheng <kito.cheng@gmail.com>
3340 Chung-Ju Wu <jasonwucj@gmail.com>
3342 * config/nds32/nds32-cost.c (rtx_cost_model_t): New structure.
3343 (insn_size_16bit, insn_size_32bit): New variables for cost evaluation.
3344 (nds32_rtx_costs_impl): Simplify.
3345 (nds32_address_cost_impl): Simplify.
3346 (nds32_init_rtx_costs): New function.
3347 (nds32_rtx_costs_speed_prefer): Likewise.
3348 (nds32_rtx_costs_size_prefer): Likewise.
3349 (nds32_address_cost_speed_prefer): Likewise.
3350 (nds32_address_cost_speed_fwprop): Likewise.
3351 (nds32_address_cost_size_prefer): Likewise.
3352 * config/nds32/nds32-protos.h (nds32_init_rtx_costs): Declare.
3353 * config/nds32/nds32.c (nds32_option_override): Use
3354 nds32_init_rtx_costs function.
3356 2018-05-20 Chung-Ju Wu <jasonwucj@gmail.com>
3358 * config/nds32/nds32.c (nds32_asm_file_start): Output pipeline model.
3359 * config/nds32/nds32.h (TARGET_PIPELINE_N7): Define.
3360 (TARGET_PIPELINE_N8): Likewise.
3361 (TARGET_PIPELINE_N10): Likewise.
3362 (TARGET_PIPELINE_N13): Likewise.
3363 (TARGET_PIPELINE_GRAYWOLF): Likewise.
3365 2018-05-19 Monk Chiang <sh.chiang04@gmail.com>
3367 * config/nds32/nds32-fpu.md: Update copyright year.
3369 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3371 * config/nds32/nds32.h (ASM_SPEC): Adjust spec rule.
3373 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3375 * config/nds32/nds32.c
3376 (nds32_md_asm_adjust): Consider flag_inline_asm_r15 variable.
3377 * config/nds32/nds32.opt (minline-asm-r15): New option.
3379 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3381 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS): Add
3383 * config/nds32/nds32.md (abssi2): New pattern.
3385 2018-05-19 Uros Bizjak <ubizjak@gmail.com>
3387 * config/i386/i386.md (rex64namesuffix): New mode attribute.
3388 * config/i386/sse.md (sse_cvtsi2ss<rex64namesuffix><round_name>):
3389 Merge insn pattern from sse_cvtsi2ss<round_name> and
3390 sse_cvtsi2ssq<round_name> using SWI48 mode iterator.
3391 (sse_cvtss2si<rex64namesuffix><round_name>): Merge insn pattern
3392 from sse_cvtss2si<round_name> and sse_cvtss2siq<round_name>
3393 using SWI48 mode iterator.
3394 (sse_cvtss2si<rex64namesuffix>_2): Merge insn pattern from
3395 sse_cvtss2si_2 and sse_cvtss2siq_2 using SWI48 mode iterator.
3396 (sse_cvttss2si<rex64namesuffix><round_saeonly_name>): Merge insn
3397 pattern from sse_cvttss2si<round_saeonly_name>
3398 and sse_cvttss2siq<round_saeonly_name> using SWI48 mode iterator.
3399 (avx512f_vcvtss2usi<rex64namesuffix><round_name>): Merge insn pattern
3400 from avx512f_vcvtss2usi<round_name> and avx512f_vcvtss2usiq<round_name>
3401 using SWI48 mode iterator.
3402 (avx512f_vcvttss2usi<rex64namesuffix><round_saeonly_name>): Merge
3403 insn pattern from avx512f_vcvttss2usi<round_saeonly_name> and
3404 avx512f_vcvttss2usiq<round_saeonly_name> using SWI48 mode iterator.
3405 (avx512f_vcvtsd2usi<rex64namesuffix><round_name>): Merge insn pattern
3406 from avx512f_vcvtsd2usi<round_name> and avx512f_vcvtsd2usiq<round_name>
3407 using SWI48 mode iterator.
3408 (avx512f_vcvttsd2usi<rex64namesuffix><round_saeonly_name>): Merge
3409 insn pattern from avx512f_vcvttsd2usi<round_saeonly_name> and
3410 avx512f_vcvttsd2usiq<round_saeonly_name> using SWI48 mode iterator.
3411 (sse2_cvtsd2si<rex64namesuffix><round_name>): Merge insn pattern from
3412 sse2_cvtsd2si<round_name> and sse2_cvtsd2siq<round_name> using
3413 SWI48 mode iterator.
3414 (sse2_cvtsd2si<rex64namesuffix>_2): Merge insn pattern from
3415 sse2_cvtsd2si_2 and sse2_cvtsd2siq_2 using SWI48 mode iterator.
3416 (sse_cvttsd2si<rex64namesuffix><round_saeonly_name>): Merge insn
3417 pattern from sse_cvttsd2si<round_saeonly_name>
3418 and sse_cvttsd2siq<round_saeonly_name> using SWI48 mode iterator.
3420 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3422 * config/nds32/nds32-md-auxiliary.c
3423 (nds32_valid_smw_lwm_base_p): Refine.
3424 (nds32_output_smw_single_word): Refine.
3425 (nds32_output_smw_double_word): New.
3426 * config/nds32/nds32-protos.h (nds32_output_smw_double_word): New.
3428 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3430 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push): Refine.
3431 (nds32_output_stack_pop): Refine.
3432 (nds32_expand_unaligned_load): Refine.
3433 (nds32_expand_unaligned_store): Refine.
3435 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3436 Chung-Ju Wu <jasonwucj@gmail.com>
3438 * config/nds32/constants.md: Add TP_REGNUM constant.
3439 (unspec_element): Add UNSPEC_GOTINIT, UNSPEC_GOT, UNSPEC_GOTOFF,
3440 UNSPEC_PLT, UNSPEC_TLSGD, UNSPEC_TLSLD, UNSPEC_TLSIE, UNSPEC_TLSLE and
3442 * config/nds32/nds32-doubleword.md: Consider flag_pic.
3443 * config/nds32/nds32-dspext.md (mov<mode>): Expand TLS and PIC cases.
3444 * config/nds32/nds32-predicates.c (nds32_const_unspec_p): New.
3445 * config/nds32/nds32-md-auxiliary.c: Implementation that support TLS
3446 and PIC code generation.
3447 * config/nds32/nds32-protos.h: Declarations that support TLS and PIC
3449 * config/nds32/nds32-relax-opt.c: Consider TLS and PIC for relax
3451 * config/nds32/nds32.md: Support TLS and PIC.
3452 * config/nds32/nds32.c: Support TLS and PIC.
3453 * config/nds32/nds32.h (nds32_relax_insn_type): New enum type.
3454 * config/nds32/predicates.md (nds32_nonunspec_symbolic_operand): New
3457 2018-05-19 Chung-Ju Wu <jasonwucj@gmail.com>
3459 * config/nds32/nds32-predicates.c (const_vector_to_hwint): Use machine
3460 mode with E_ prefix.
3462 2018-05-19 Kuan-Lin Chen <kuanlinchentw@gmail.com>
3463 Chung-Ju Wu <jasonwucj@gmail.com>
3465 * config/nds32/constants.md (unspec_element): Add UNSPEC_ICT.
3466 * config/nds32/nds32-md-auxiliary.c
3467 (symbolic_reference_mentioned_p): New.
3468 (nds32_legitimize_ict_address): New.
3469 (nds32_expand_ict_move): New.
3470 (nds32_indirect_call_referenced_p): New.
3471 (nds32_symbol_binds_local_p): Delete.
3472 (nds32_long_call_p): Modify.
3473 * config/nds32/nds32-opts.h (nds32_ict_model_type): New enum type.
3474 * config/nds32/nds32-protos.h
3475 (symbolic_reference_mentioned_p): Declare.
3476 (nds32_legitimize_ict_address): Declare.
3477 (nds32_expand_ict_move): Declare.
3478 (nds32_indirect_call_referenced_p): Declare.
3479 * config/nds32/nds32-relax-opt.c (nds32_ict_const_p): New.
3480 (nds32_relax_group): Use nds32_ict_const_p as condition.
3481 * config/nds32/nds32.c (nds32_attribute_table): Add "indirect_call".
3482 (nds32_asm_file_start): Output ict_model directive in asm code.
3483 (nds32_legitimate_address_p): Consider indirect call.
3484 (nds32_print_operand): Consider indirect call.
3485 (nds32_print_operand_address): Consider indirect call.
3486 (nds32_insert_attributes): Handle "indirect_call" attribute.
3487 (TARGET_LEGITIMATE_ADDRESS_P): Define.
3488 (TARGET_LEGITIMATE_CONSTANT_P): Define.
3489 (TARGET_CANNOT_FORCE_CONST_MEM): Define.
3490 (TARGET_DELEGITIMIZE_ADDRESS): Define.
3491 (TARGET_ASM_OUTPUT_ADDR_CONST_EXTRA): Define.
3492 * config/nds32/nds32.h (SYMBOLIC_CONST_P): Define.
3493 (TARGET_ICT_MODEL_SMALL): Define.
3494 (TARGET_ICT_MODEL_LARGE): Define.
3495 * config/nds32/nds32.md (movsi): Consider ict model.
3496 (call, call_value): Consider ict model.
3497 (sibcall, sibcall_value): Consider ict model.
3498 * config/nds32/nds32.opt (mict-model): New option.
3499 * config/nds32/predicates.md (nds32_symbolic_operand): Consider ict
3502 2018-05-18 Kito Cheng <kito.cheng@gmail.com>
3503 Monk Chiang <sh.chiang04@gmail.com>
3504 Jim Wilson <jimw@sifive.com>
3506 * common/config/riscv/riscv-common.c (riscv_parse_arch_string):
3507 Add support to parse rv32e*. Clear MASK_RVE for rv32i and rv64i.
3508 * config.gcc (riscv*-*-*): Add support for rv32e* and ilp32e.
3509 * config/riscv/riscv-c.c (riscv_cpu_cpp_builtins): Define
3510 __riscv_32e when TARGET_RVE. Handle ABI_ILP32E as soft-float ABI.
3511 * config/riscv/riscv-opts.h (riscv_abi_type): Add ABI_ILP32E.
3512 * config/riscv/riscv.c (riscv_compute_frame_info): When TARGET_RVE,
3513 compute save_libcall_adjustment properly.
3514 (riscv_option_override): Call error if TARGET_RVE and not ABI_ILP32E.
3515 (riscv_conditional_register_usage): Handle TARGET_RVE and ABI_ILP32E.
3516 * config/riscv/riscv.h (UNITS_PER_FP_ARG): Handle ABI_ILP32E.
3517 (STACK_BOUNDARY, ABI_STACK_BOUNDARY): Handle TARGET_RVE.
3518 (GP_REG_LAST, MAX_ARGS_IN_REGISTERS): Likewise.
3519 (ABI_SPEC): Handle mabi=ilp32e.
3520 * config/riscv/riscv.opt (abi_type): Add ABI_ILP32E.
3521 (RVE): Add RVE mask.
3522 * doc/invoke.texi (RISC-V options) <-mabi>: Add ilp32e info.
3523 <-march>: Add rv32e as an example.
3525 2018-05-18 Marc Glisse <marc.glisse@inria.fr>
3528 * tree-ssa-structalias.c (create_variable_info_for_1): Extra argument.
3529 (intra_create_variable_infos): Handle C++ constructors.
3531 2018-05-18 Martin Liska <mliska@suse.cz>
3533 * passes.def: Remove a redundant pass.
3535 2018-05-18 Eric Botcazou <ebotcazou@adacore.com>
3538 * config/sparc/sparc.c (sparc_expand_builtin): Always initialize op[0].
3540 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3542 * config/arm/arm-cpus.in (armv3m, mode32): Delete features.
3544 (ARMv2, ARMv3, ARMv3m): Delete fgroups.
3546 (armv2, armv2a, armv3, armv3m): Delete architectures.
3547 (arm2, arm250, arm3, arm6, arm60, arm600, arm610, arm620,
3548 arm7, arm7d, arm7di, arm70, arm700, arm700i, arm710, arm720,
3549 arm710c, arm7100, arm7500, arm7500fe, arm7m, arm7dm, arm7dmi):
3551 * config/arm/arm.md (maddsidi4): Remove check for arm_arch3m.
3552 (*mulsidi3adddi): Likewise.
3553 (mulsidi3): Likewise.
3554 (*mulsidi3_nov6): Likewise.
3555 (umulsidi3): Likewise.
3556 (umulsidi3_nov6): Likewise.
3557 (umaddsidi4): Likewise.
3558 (*umulsidi3adddi): Likewise.
3559 (smulsi3_highpart): Likewise.
3560 (*smulsi3_highpart_nov6): Likewise.
3561 (umulsi3_highpart): Likewise.
3562 (*umulsi3_highpart_nov6): Likewise.
3563 * config/arm/arm.h (arm_arch3m): Delete.
3564 * config/arm/arm.c (arm_arch3m): Delete.
3565 (arm_option_override_internal): Update armv3-related comment.
3566 (arm_configure_build_target): Delete use of isa_bit_mode32.
3567 (arm_option_reconfigure_globals): Delete set of arm_ach3m.
3568 (arm_rtx_costs_internal): Delete check of arm_arch3m.
3569 * config/arm/arm-fixed.md (mulsq3): Delete check for arm_arch3m.
3571 (mulusa3): Likewise.
3572 * config/arm/arm-protos.h (arm_arch3m): Delete.
3573 * config/arm/arm-tables.opt: Regenerate.
3574 * config/arm/arm-tune.md: Likewise.
3575 * config/arm/t-arm-elf (all_early_nofp): Delete mentions of
3576 deleted architectures.
3578 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3580 * config/arm/arm-cpus.in (armv5, armv5e): Delete features.
3581 (armv5t, armv5te): New features.
3582 (ARMv5, ARMv5e): Delete fgroups.
3583 (ARMv5t, ARMv5te): Adjust for above changes.
3585 (armv5, armv5e): Delete arches.
3586 * config/arm/arm.md (*call_reg_armv5): Use arm_arch5t instead of
3588 (*call_reg_arm): Likewise.
3589 (*call_value_reg_armv5): Likewise.
3590 (*call_value_reg_arm): Likewise.
3591 (*call_symbol): Likewise.
3592 (*call_value_symbol): Likewise.
3593 (*sibcall_insn): Likewise.
3594 (*sibcall_value_insn): Likewise.
3596 (prefetch): Likewise.
3597 (define_split and define_peephole2 dependent on arm_arch5):
3599 * config/arm/arm.h (TARGET_LDRD): Use arm_arch5te instead of
3601 (TARGET_ARM_QBIT): Likewise.
3602 (TARGET_DSP_MULTIPLY): Likewise.
3603 (enum base_architecture): Delete BASE_ARCH_5, BASE_ARCH_5E.
3604 (arm_arch5, arm_arch5e): Delete.
3605 (arm_arch5t, arm_arch5te): Declare.
3606 * config/arm/arm.c (arm_arch5, arm_arch5e): Delete.
3607 (arm_arch5t): Declare.
3608 (arm_option_reconfigure_globals): Update for the above.
3609 (arm_options_perform_arch_sanity_checks): Update comment, replace
3610 use of arm_arch5 with arm_arch5t.
3611 (use_return_insn): Likewise.
3612 (arm_emit_call_insn): Likewise.
3613 (output_return_instruction): Likewise.
3614 (arm_final_prescan_insn): Likewise.
3615 (arm_coproc_builtin_available): Likewise.
3616 * config/arm/arm-c.c (arm_cpu_builtins): Replace arm_arch5 and
3617 arm_arch5e with arm_arch5t and arm_arch5te.
3618 * config/arm/arm-protos.h (arm_arch5, arm_arch5e): Delete.
3619 (arm_arch5t, arm_arch5te): Declare.
3620 * config/arm/arm-tables.opt: Regenerate.
3621 * config/arm/t-arm-elf: Remove references to armv5, armv5e.
3622 * config/arm/t-multilib: Likewise.
3623 * config/arm/thumb1.md (*call_reg_thumb1_v5): Check arm_arch5t
3624 instead of arm_arch5.
3625 (*call_reg_thumb1): Likewise.
3626 (*call_value_reg_thumb1_v5): Likewise.
3627 (*call_value_reg_thumb1): Likewise.
3628 * config/arm/vxworks.h (TARGET_OS_CPP_BUILTINS): Remove now
3630 * doc/invoke.texi (ARM Options): Remove references to armv5, armv5e.
3632 2018-05-18 Martin Liska <mliska@suse.cz>
3634 PR gcov-profile/84846
3635 * doc/gcov.texi: Document -t option of gcov tool.
3637 2018-05-18 Martin Liska <mliska@suse.cz>
3639 PR gcov-profile/84846
3640 * gcov.c (print_usage): Add new -t option.
3641 (process_args): Handle the option.
3642 (generate_results): Use stdout as output when requested by
3645 2018-05-18 Martin Liska <mliska@suse.cz>
3647 PR gcov-profile/84846
3648 * coverage.c (coverage_init): Write PWD to .gcno file.
3649 * doc/gcov.texi: Document how working directory is printed.
3650 * gcov-dump.c (dump_gcov_file): Print PWD.
3651 * gcov.c (output_intermediate_file): Likewise.
3652 (read_graph_file): Read PWD string.
3653 (output_lines): Print PWD.
3655 2018-05-18 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
3658 * ipa-pure-const.c (malloc_candidate_p): Remove the check integer_zerop
3659 for retval and return false if all args to phi are zero.
3661 2018-05-18 Richard Biener <rguenther@suse.de>
3663 * gimple-ssa-evrp.c (class evrp_folder): Add simplify_stmt_using_ranges
3665 (evrp_dom_walker::before_dom_children): Call it.
3667 2018-05-18 Richard Biener <rguenther@suse.de>
3669 * tree-dfa.c (get_ref_base_and_extent): Use range-info to refine
3670 results when processing array refs with variable index.
3672 2018-05-18 Toon Moene <toon@moene.org>
3674 * doc/invoke.texi: Move -floop-unroll-and-jam documentation
3675 directly after that of -floop-interchange. Indicate that both
3676 options are enabled by default when specifying -O3.
3678 2018-05-18 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3680 * config/aarch64/aarch64-simd.md (vec_set<mode>): Use VALL_F16 mode
3681 iterator. Delete separate integer-mode vec_set<mode> expander.
3682 (aarch64_simd_vec_setv2di): Delete.
3683 (vec_setv2di): Delete.
3684 (aarch64_simd_vec_set<mode>): Delete all other patterns with that name.
3685 Use VALL_F16 mode iterator. Add LD1 alternative and use vwcore for
3686 the "w, r" alternative.
3688 2018-05-18 Martin Liska <mliska@suse.cz>
3690 * passes.def: Add pass_lower_switch and pass_lower_switch_O0.
3691 * tree-pass.h (make_pass_lower_switch_O0): New function.
3692 * tree-switch-conversion.c (node_has_low_bound): Remove.
3693 (node_has_high_bound): Likewise.
3694 (node_is_bounded): Likewise.
3695 (class pass_lower_switch): Make it a template type and create
3697 (pass_lower_switch::execute): Add template argument.
3698 (make_pass_lower_switch): New function.
3699 (make_pass_lower_switch_O0): New function.
3700 (do_jump_if_equal): Remove.
3701 (emit_case_nodes): Simplify to just handle all 3 cases and leave
3702 all the hard work to tree optimization passes.
3704 2018-05-18 Martin Liska <mliska@suse.cz>
3706 * dbgcnt.c (limit_low): Renamed from limit.
3707 (limit_high): New variable.
3708 (dbg_cnt_is_enabled): Check for upper limit.
3709 (dbg_cnt): Adjust dumping.
3710 (dbg_cnt_set_limit_by_index): Add new argument for high
3712 (dbg_cnt_set_limit_by_name): Likewise.
3713 (dbg_cnt_process_single_pair): Parse new format.
3714 (dbg_cnt_process_opt): Use strtok.
3715 (dbg_cnt_list_all_counters): Remove 'value' and add
3717 * doc/invoke.texi: Document changes.
3719 2018-05-18 Richard Sandiford <richard.sandiford@linaro.org>
3721 * doc/sourcebuild.texi (scalar_all_fma): Document.
3722 * tree.def (FMA_EXPR): Delete.
3723 * internal-fn.def (FMA, FMS, FNMA, FNMS): New internal functions.
3724 * internal-fn.c (ternary_direct): New macro.
3725 (expand_ternary_optab_fn): Likewise.
3726 (direct_ternary_optab_supported_p): Likewise.
3727 * Makefile.in (build/genmatch.o): Depend on case-fn-macros.h.
3728 * builtins.c (fold_builtin_fma): Delete.
3729 (fold_builtin_3): Don't call it.
3730 * cfgexpand.c (expand_debug_expr): Remove FMA_EXPR handling.
3731 * expr.c (expand_expr_real_2): Likewise.
3732 * fold-const.c (operand_equal_p): Likewise.
3733 (fold_ternary_loc): Likewise.
3734 * gimple-pretty-print.c (dump_ternary_rhs): Likewise.
3735 * gimple.c (DEFTREECODE): Likewise.
3736 * gimplify.c (gimplify_expr): Likewise.
3737 * optabs-tree.c (optab_for_tree_code): Likewise.
3738 * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
3739 * tree-eh.c (operation_could_trap_p): Likewise.
3740 (stmt_could_throw_1_p): Likewise.
3741 * tree-inline.c (estimate_operator_cost): Likewise.
3742 * tree-pretty-print.c (dump_generic_node): Likewise.
3743 (op_code_prio): Likewise.
3744 * tree-ssa-loop-im.c (stmt_cost): Likewise.
3745 * tree-ssa-operands.c (get_expr_operands): Likewise.
3746 * tree.c (commutative_ternary_tree_code, add_expr): Likewise.
3747 * fold-const-call.h (fold_fma): Delete.
3748 * fold-const-call.c (fold_const_call_ssss): Handle CFN_FMS,
3749 CFN_FNMA and CFN_FNMS.
3751 * genmatch.c (combined_fn): New enum.
3752 (commutative_ternary_tree_code): Remove FMA_EXPR handling.
3753 (commutative_op): New function.
3754 (commutate): Use it. Handle more than 2 operands.
3755 (dt_operand::gen_gimple_expr): Use commutative_op.
3756 (parser::parse_expr): Allow :c to be used with non-binary
3757 operators if the commutative operand is known.
3758 * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Handle
3759 CFN_FMS, CFN_FNMA and CFN_FNMS.
3760 (backprop::process_assign_use): Remove FMA_EXPR handling.
3761 * hsa-gen.c (gen_hsa_insns_for_operation_assignment): Likewise.
3762 (gen_hsa_fma): New function.
3763 (gen_hsa_insn_for_internal_fn_call): Use it for IFN_FMA, IFN_FMS,
3764 IFN_FNMA and IFN_FNMS.
3765 * match.pd: Add folds for IFN_FMS, IFN_FNMA and IFN_FNMS.
3766 * gimple-fold.h (follow_all_ssa_edges): Declare.
3767 * gimple-fold.c (follow_all_ssa_edges): New function.
3768 * tree-ssa-math-opts.c (convert_mult_to_fma_1): Use the
3769 gimple_build interface and use follow_all_ssa_edges to fold the result.
3770 (convert_mult_to_fma): Use direct_internal_fn_suppoerted_p
3771 instead of checking for optabs directly.
3772 * config/i386/i386.c (ix86_add_stmt_cost): Recognize FMAs as calls
3773 rather than FMA_EXPRs.
3774 * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Create a
3775 call to IFN_FMA instead of an FMA_EXPR.
3777 2018-05-17 Jim Wilson <jimw@sifive.com>
3779 * expr.c (do_tablejump): When converting index to Pmode, if we have a
3780 sign extended promoted subreg, and the range does not have the sign bit
3781 set, then do a sign extend.
3783 * config/riscv/riscv.c (riscv_extend_comparands): In unsigned QImode
3784 test, check for sign extended subreg and/or constant operands, and
3785 do a sign extend in that case.
3787 2018-05-17 Steve Ellcey <sellcey@cavium.com>
3789 * config/aarch64/thunderx2t99.md (thunderx2t99_ls_both): Delete.
3790 (thunderx2t99_multiple): Delete psuedo-units from used cpus.
3792 (thunderx2t99_alu_shift): Remove alu_shift_reg, alus_shift_reg.
3793 Change logics_shift_reg to logics_shift_imm.
3794 (thunderx2t99_fp_loadpair_basic): Delete.
3795 (thunderx2t99_fp_storepair_basic): Delete.
3796 (thunderx2t99_asimd_int): Add neon_sub and neon_sub_q types.
3797 (thunderx2t99_asimd_polynomial): Delete.
3798 (thunderx2t99_asimd_fp_simple): Add neon_fp_mul_s_scalar_q
3799 and neon_fp_mul_d_scalar_q.
3800 (thunderx2t99_asimd_fp_conv): Add *int_to_fp* types.
3801 (thunderx2t99_asimd_misc): Delete neon_dup and neon_dup_q.
3802 (thunderx2t99_asimd_recip_step): Add missing *sqrt* types.
3803 (thunderx2t99_asimd_lut): Add missing tbl types.
3804 (thunderx2t99_asimd_ext): Delete.
3805 (thunderx2t99_asimd_load1_1_mult): Delete.
3806 (thunderx2t99_asimd_load1_2_mult): Delete.
3807 (thunderx2t99_asimd_load1_ldp): New.
3808 (thunderx2t99_asimd_load1): New.
3809 (thunderx2t99_asimd_load2): Add missing *load2* types.
3810 (thunderx2t99_asimd_load3): New.
3811 (thunderx2t99_asimd_load4): New.
3812 (thunderx2t99_asimd_store1_1_mult): Delete.
3813 (thunderx2t99_asimd_store1_2_mult): Delete.
3814 (thunderx2t99_asimd_store2_mult): Delete.
3815 (thunderx2t99_asimd_store2_onelane): Delete.
3816 (thunderx2t99_asimd_store_stp): New.
3817 (thunderx2t99_asimd_store1): New.
3818 (thunderx2t99_asimd_store2): New.
3819 (thunderx2t99_asimd_store3): New.
3820 (thunderx2t99_asimd_store4): New.
3822 2018-05-17 Jerome Lambourg <lambourg@adacore.com>
3824 * config/arm/arm_cmse.h (cmse_nsfptr_create, cmse_is_nsfptr): Remove
3825 #include <stdint.h>. Replace intptr_t with __INTPTR_TYPE__.
3827 2018-05-17 Pat Haugen <pthaugen@us.ibm.com>
3828 Segher Boessenkool <segher@kernel.crashing.org>
3831 * config/rs6000/rs6000.c (rs6000_output_move_128bit): Check dest
3834 2018-05-17 Richard Biener <rguenther@suse.de>
3836 * tree-ssa-dse.c (dse_classify_store): Fix iterator increment
3837 for pruning loop and prune defs feeding only already visited PHIs.
3839 2018-05-17 Richard Biener <rguenther@suse.de>
3841 * tree-ssa-sccvn.c (vn_reference_lookup_3): Improve memset handling.
3843 2018-05-17 Bin Cheng <bin.cheng@arm.com>
3844 Richard Biener <rguenther@suse.de>
3846 PR tree-optimization/85793
3847 * tree-vect-stmts.c (vectorizable_load): Handle 1 element-wise load
3848 for VMAT_ELEMENTWISE.
3850 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
3852 * internal-fn.h (lookup_internal_fn): Declare
3853 * internal-fn.c (lookup_internal_fn): New function.
3854 * gimple.c (gimple_build_call_from_tree): Handle calls to
3856 * gimple-pretty-print.c (dump_gimple_call): Print "." before
3857 internal function names.
3858 * tree-pretty-print.c (dump_generic_node): Likewise.
3859 * tree-ssa-scopedtables.c (expr_hash_elt::print): Likewise.
3861 2018-05-17 Richard Sandiford <richard.sandiford@linaro.org>
3863 * gimple-fold.h (gimple_build): Make the function forms take
3864 combined_fn rather than built_in_function.
3865 (gimple_simplify): Likewise.
3866 * gimple-match-head.c (gimple_simplify): Likewise.
3867 * gimple-fold.c (gimple_build): Likewise.
3868 * tree-vect-loop.c (get_initial_def_for_reduction): Use gimple_build
3869 rather than gimple_build_call_internal.
3870 (get_initial_defs_for_reduction): Likewise.
3871 (vect_create_epilog_for_reduction): Likewise.
3872 (vectorizable_live_operation): Likewise.
3874 2018-05-17 Martin Liska <mliska@suse.cz>
3876 * gimple-ssa-sprintf.c (format_directive): Do not use
3877 space in between 'G_' and '('.
3879 2018-05-17 Jakub Jelinek <jakub@redhat.com>
3882 * config/i386/i386.c (ix86_fold_builtin): Handle masked shifts
3883 even if the mask is not all ones.
3886 * config/i386/i386.c (ix86_fold_builtin): Fold shift builtins by
3888 (ix86_gimple_fold_builtin): Likewise.
3891 * config/i386/i386.c: Include tree-vector-builder.h.
3892 (ix86_vector_shift_count): New function.
3893 (ix86_fold_builtin): Fold shift builtins by scalar count.
3894 (ix86_gimple_fold_builtin): Likewise.
3896 * config/i386/avx512fintrin.h (_mm512_set_epi16, _mm512_set_epi8,
3897 _mm512_setzero): New intrinsics.
3899 2018-05-17 James Greenhalgh <james.greenhalgh@arm.com>
3900 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
3902 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify
3903 code generation for cases where splatting a value is not useful.
3904 * simplify-rtx.c (simplify_ternary_operation): Simplify
3905 vec_merge across a vec_duplicate and a paradoxical subreg forming
3906 a vector mode to a vec_concat.
3908 2018-05-17 Olga Makhotina <olga.makhotina@intel.com>
3910 * config.gcc: Support "goldmont-plus".
3911 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
3913 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
3914 PROCESSOR_GOLDMONT_PLUS.
3915 * config/i386/i386.c (m_GOLDMONT_PLUS): Define.
3916 (processor_target_table): Add "goldmont-plus".
3917 (PTA_GOLDMONT_PLUS): Define.
3918 (ix86_lea_outperforms): Add TARGET_GOLDMONT_PLUS.
3919 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT_PLUS.
3920 (fold_builtin_cpu): Add M_INTEL_GOLDMONT_PLUS.
3921 (fold_builtin_cpu): Add "goldmont-plus".
3922 (ix86_add_stmt_cost): Add TARGET_GOLDMONT_PLUS.
3923 (ix86_option_override_internal): Add "goldmont-plus".
3924 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT_PLUS.
3925 (processor_type): Add PROCESSOR_GOLDMONT_PLUS.
3926 * config/i386/x86-tune.def: Add m_GOLDMONT_PLUS.
3927 * doc/invoke.texi: Add goldmont-plus as x86 -march=/-mtune= CPU type.
3929 2018-05-17 Richard Biener <rguenther@suse.de>
3931 PR tree-optimization/85757
3932 * tree-ssa-dse.c (dse_classify_store): Record a PHI def and
3933 remove defs that only feed that PHI from further processing.
3935 2018-05-16 Jim Wilson <jimw@sifive.com>
3937 * config/riscv/riscv.md (<optab>si3_mask, <optab>si3_mask_1): Prepend
3939 (<optab>di3_mask, <optab>di3_mask_1): Likewise.
3941 2018-05-16 Mark Wielaard <mark@klomp.org>
3943 * dwarf2out.c (count_index_strings): New function.
3944 (output_indirect_strings): Call count_index_strings and generate
3945 header for dwarf_version >= 5.
3947 2018-05-16 Mark Wielaard <mark@klomp.org>
3949 * dwarf2out.c (dwarf_FORM): New function.
3950 (set_indirect_string): Use dwarf_FORM.
3951 (reset_indirect_string): Likewise.
3952 (size_of_die): Likewise.
3953 (value_format): Likewise.
3954 (output_die): Likewise.
3955 (add_skeleton_AT_string): Likewise.
3956 (output_macinfo_op): Likewise.
3957 (index_string): Likewise.
3958 (output_index_string_offset): Likewise.
3959 (output_index_string): Likewise.
3960 (count_index_strings): Likewise.
3962 2018-05-16 Carl Love <cel@us.ibm.com>
3964 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
3965 dcbt and dcbtstt with TH=16 if operands[2] is 0 and Power 8 or newer.
3967 2018-05-16 Martin Jambor <mjambor@suse.cz>
3969 * ipa-prop.c (ipa_free_all_edge_args): Remove.
3970 * ipa-prop.h (ipa_free_all_edge_args): Likewise.
3972 2018-05-16 Wilco Dijkstra <wdijkstr@arm.com>
3974 * config/aarch64/aarch64.md (fma<mode>4): Change into expand pattern.
3975 (fnma<mode>4): Likewise.
3976 (fms<mode>4): Likewise.
3977 (fnms<mode>4): Likewise.
3978 (aarch64_fma<mode>4): Rename insn, reorder accumulator operand.
3979 (aarch64_fnma<mode>4): Likewise.
3980 (aarch64_fms<mode>4): Likewise.
3981 (aarch64_fnms<mode>4): Likewise.
3982 (aarch64_fnmadd<mode>4): Likewise.
3984 2018-05-16 Jason Merrill <jason@redhat.com>
3986 * tree.c (warn_deprecated_use): Return bool. Simplify logic.
3988 2018-05-16 Richard Biener <rguenther@suse.de>
3990 * tree-vectorizer.h (struct stmt_info_for_cost): Add where member.
3991 (dump_stmt_cost): Declare.
3992 (add_stmt_cost): Dump cost we add.
3993 (add_stmt_costs): New function.
3994 (vect_model_simple_cost, vect_model_store_cost, vect_model_load_cost):
3996 (vect_analyze_stmt): Adjust prototype.
3997 (vectorizable_condition): Likewise.
3998 (vectorizable_live_operation): Likewise.
3999 (vectorizable_reduction): Likewise.
4000 (vectorizable_induction): Likewise.
4001 * tree-vect-loop.c (vect_analyze_loop_operations): Create local
4002 cost vector to pass to vectorizable_ and record afterwards.
4003 (vect_model_reduction_cost): Take cost vector argument and adjust.
4004 (vect_model_induction_cost): Likewise.
4005 (vectorizable_reduction): Likewise.
4006 (vectorizable_induction): Likewise.
4007 (vectorizable_live_operation): Likewise.
4008 * tree-vect-slp.c (vect_create_new_slp_node): Initialize
4009 SLP_TREE_NUMBER_OF_VEC_STMTS.
4010 (vect_analyze_slp_cost_1): Remove.
4011 (vect_analyze_slp_cost): Likewise.
4012 (vect_slp_analyze_node_operations): Take visited args and
4013 a target cost vector. Avoid processing already visited stmt sets.
4014 (vect_slp_analyze_operations): Use a local cost vector to gather
4015 costs and register those of non-discarded instances.
4016 (vect_bb_vectorization_profitable_p): Use add_stmt_costs.
4017 (vect_schedule_slp_instance): Remove copying of
4018 SLP_TREE_NUMBER_OF_VEC_STMTS. Instead assert that it is not
4020 * tree-vect-stmts.c (record_stmt_cost): Remove path directly
4021 adding cost. Record cost entry location.
4022 (vect_prologue_cost_for_slp_op): Function to compute cost of
4023 a constant or invariant generated for SLP vect in the prologue,
4024 split out from vect_analyze_slp_cost_1.
4025 (vect_model_simple_cost): Make static. Adjust for SLP costing.
4026 (vect_model_promotion_demotion_cost): Likewise.
4027 (vect_model_store_cost): Likewise, make static.
4028 (vect_model_load_cost): Likewise.
4029 (vectorizable_bswap): Add cost vector arg and adjust.
4030 (vectorizable_call): Likewise.
4031 (vectorizable_simd_clone_call): Likewise.
4032 (vectorizable_conversion): Likewise.
4033 (vectorizable_assignment): Likewise.
4034 (vectorizable_shift): Likewise.
4035 (vectorizable_operation): Likewise.
4036 (vectorizable_store): Likewise.
4037 (vectorizable_load): Likewise.
4038 (vectorizable_condition): Likewise.
4039 (vectorizable_comparison): Likewise.
4040 (can_vectorize_live_stmts): Likewise.
4041 (vect_analyze_stmt): Likewise.
4042 (vect_transform_stmt): Adjust calls to vectorizable_*.
4043 * tree-vectorizer.c: Include gimple-pretty-print.h.
4044 (dump_stmt_cost): New function.
4046 2018-05-16 Richard Biener <rguenther@suse.de>
4048 * params.def (PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE): New param.
4049 * doc/invoke.texi (dse-max-alias-queries-per-store): Document.
4050 * tree-ssa-dse.c: Include tree-ssa-loop.h.
4051 (check_name): New callback.
4052 (dse_classify_store): Track cycles via a visited bitmap of PHI
4053 defs and simplify handling of in-loop and across loop dead stores
4054 and properly fail for loop-variant refs. Handle byte-tracking with
4055 multiple defs. Use PARAM_DSE_MAX_ALIAS_QUERIES_PER_STORE for
4058 2018-05-16 Richard Sandiford <richard.sandiford@linaro.org>
4060 * tree-vectorizer.h (vect_get_vector_types_for_stmt): Declare.
4061 (vect_get_mask_type_for_stmt): Likewise.
4062 * tree-vect-slp.c (vect_two_operations_perm_ok_p): New function,
4064 (vect_build_slp_tree_1): ...here. Use vect_get_vector_types_for_stmt
4065 to determine the statement's vector type and the vector type that
4066 should be used for calculating nunits. Deal with cases in which
4067 the type has to be deferred.
4068 (vect_slp_analyze_node_operations): Use vect_get_vector_types_for_stmt
4069 and vect_get_mask_type_for_stmt to calculate STMT_VINFO_VECTYPE.
4070 * tree-vect-loop.c (vect_determine_vf_for_stmt_1)
4071 (vect_determine_vf_for_stmt): New functions, split out from...
4072 (vect_determine_vectorization_factor): ...here.
4073 * tree-vect-stmts.c (vect_get_vector_types_for_stmt)
4074 (vect_get_mask_type_for_stmt): New functions, split out from
4075 vect_determine_vectorization_factor.
4077 2018-05-16 Richard Biener <rguenther@suse.de>
4079 * tree-cfg.c (verify_gimple_assign_ternary): Properly
4080 verify the [VEC_]COND_EXPR embedded comparison.
4082 2018-05-15 Martin Sebor <msebor@redhat.com>
4084 PR tree-optimization/85753
4085 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Handle
4086 RECORD_TYPE in addition to ARRAY_TYPE.
4088 2018-05-15 Martin Sebor <msebor@redhat.com>
4091 * calls.c (get_attr_nonstring_decl): Handle MEM_REF.
4093 2018-05-15 Richard Biener <rguenther@suse.de>
4095 * tree-ssa-dse.c (dse_classify_store): Remove use_stmt parameter,
4096 add by_clobber_p one. Change algorithm to collect all defs
4097 representing uses we need to walk and try reducing them to
4098 a single one before failing.
4099 (dse_dom_walker::dse_optimize_stmt): Adjust.
4101 2018-05-13 Mark Wielaard <mark@klomp.org>
4103 * dwarf2out.c (dwarf_OP): Handle DW_OP_addrx and DW_OP_constx.
4104 (size_of_loc_descr): Likewise.
4105 (output_loc_operands): Likewise.
4106 (output_loc_operands_raw): Likewise.
4107 (dw_addr_op): Use dwarf_OP () for DW_OP_constx and DW_OP_addrx.
4108 (resolve_addr_in_expr): Handle DW_OP_addrx and DW_OP_constx.
4109 (hash_loc_operands): Likewise.
4110 (compare_loc_operands): Likewise.
4112 2018-05-14 Mark Wielaard <mark@klomp.org>
4114 * dwarf2out.c (count_index_addrs): New function.
4115 (dwarf2out_finish): Use count_index_addrs to calculate addrs_length.
4117 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4119 PR tree-optimization/83648
4120 * ipa-pure-const.c (malloc_candidate_p): Allow function with NULL
4121 return value as malloc candidate.
4123 2018-05-15 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
4126 * ipa-pure-const.c (warn_function_malloc): Pass value of known_finite
4127 param as true in call to suggest_attribute.
4129 2018-05-14 Segher Boessenkool <segher@kernel.crashing.org>
4131 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete @itemx for
4132 -mreadonly-in-sdata.
4134 2018-05-14 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4136 * config/aarch64/aarch64-simd.md (*aarch64_crypto_aese_fused):
4138 (aarch64_crypto_aesd_fused): Likewise.
4140 2018-05-14 Wilco Dijkstra <wdijkstr@arm.com>
4142 * config/aarch64/aarch64.md (mov<mode>): Remove '*' in alternatives.
4143 (movsi_aarch64): Likewise.
4144 (load_pairsi): Likewise.
4145 (load_pairdi): Likewise.
4146 (store_pairsi): Likewise.
4147 (store_pairdi): Likewise.
4148 (load_pairsf): Likewise.
4149 (load_pairdf): Likewise.
4150 (store_pairsf): Likewise.
4151 (store_pairdf): Likewise.
4152 (zero_extend): Likewise.
4153 (trunc): Swap alternatives.
4154 (fcvt_target): Add '?' to prefer w over r.
4156 2018-05-14 Jakub Jelinek <jakub@redhat.com>
4159 * config/i386/i386.md: Disallow non-commutative arithmetics in
4160 last twpeephole for mem {+,-,&,|,^}= x; mem != 0 after cmpelim
4161 optimization. Use COMMUTATIVE_ARITH_P test rather than != MINUS
4162 in the peephole2 before it.
4164 2018-05-14 Sebastian Peryt <sebastian.peryt@intel.com>
4166 * common/config/i386/i386-common.c (OPTION_MASK_ISA_CLDEMOTE_SET,
4167 OPTION_MASK_ISA_CLDEMOTE_UNSET): New defines.
4168 (ix86_handle_option): Handle -mcldemote.
4169 * config.gcc: New header.
4170 * config/i386/cldemoteintrin.h: New file.
4171 * config/i386/cpuid.h (bit_CLDEMOTE): New bit.
4172 * config/i386/driver-i386.c (host_detect_local_cpu): Detect
4174 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4175 OPTION_MASK_ISA_CLDEMOTE.
4176 * config/i386/i386.c (ix86_target_string): Add -mcldemote.
4177 (ix86_valid_target_attribute_inner_p): Ditto.
4178 (enum ix86_builtins): Add IX86_BUILTIN_CLDEMOTE.
4179 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_cldemote.
4180 (ix86_expand_builtin): Expand IX86_BUILTIN_CLDEMOTE.
4181 * config/i386/i386.h (TARGET_CLDEMOTE, TARGET_CLDEMOTE_P): New.
4182 * config/i386/i386.md (UNSPECV_CLDEMOTE): New.
4184 * config/i386/i386.opt: Add -mcldemote.
4185 * config/i386/x86intrin.h: New header.
4186 * doc/invoke.texi: Add -mcldemote.
4188 2018-05-14 Richard Biener <rguenther@suse.de>
4190 * doc/match-and-simplify.texi: Adjust :s documentation.
4192 2018-05-14 Alexander Monakov <amonakov@ispras.ru>
4194 * sort.cc (REORDER_23): Pass the type for the temporaries instead of
4195 intended memcpy size.
4196 (REORDER_45): Likewise.
4198 2018-05-13 Alexander Monakov <amonakov@ispras.ru>
4200 * sort.cc: New file.
4201 * system.h [!CHECKING_P] (qsort): Redirect to gcc_qsort.
4202 * vec.c (qsort_chk): Use gcc_qsort.
4203 * Makefile.in (OBJS-libcommon): Add sort.o.
4204 (build/sort.o): New target. Use it...
4205 (BUILD_RTL): ... here, and...
4206 (build/gencfn-macros): ... here, and...
4207 (build/genmatch): ... here.
4209 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4210 Chung-Ju Wu <jasonwucj@gmail.com>
4212 * config.gcc (nds32*-*-*): Check that n15 is valid to --with-cpu.
4213 * config/nds32/nds32-graywolf.md: New file.
4214 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_GRAYWOLF.
4215 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n15
4217 * config/nds32/nds32-protos.h: More declarations for n15 pipeline.
4218 * config/nds32/nds32-utils.c: More implementations for n15 pipeline.
4219 * config/nds32/nds32.md (pipeline_model): Add graywolf.
4220 * config/nds32/nds32.opt (mcpu): Support n15 pipeline cpus.
4221 * config/nds32/pipelines.md: Include n15 settings.
4223 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4224 Chung-Ju Wu <jasonwucj@gmail.com>
4226 * config.gcc (nds32*-*-*): Check that n12/n13 are valid to --with-cpu.
4227 * config/nds32/nds32-n13.md: New file.
4228 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N12 and CPU_N13.
4229 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n12/n13
4231 * config/nds32/nds32-protos.h: More declarations for n12/n13 pipeline.
4232 * config/nds32/nds32.md (pipeline_model): Add n13.
4233 * config/nds32/nds32.opt (mcpu): Support n12/n13 pipeline cpus.
4234 * config/nds32/pipelines.md: Include n13 settings.
4236 2018-05-13 Kito Cheng <kito.cheng@gmail.com>
4237 Chung-Ju Wu <jasonwucj@gmail.com>
4239 * config.gcc (nds32*-*-*): Check that n10/d10 are valid to --with-cpu.
4240 * config/nds32/nds32-n10.md: New file.
4241 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N10.
4242 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n10
4244 * config/nds32/nds32-protos.h: More declarations for n10 pipeline.
4245 * config/nds32/nds32-utils.c: More implementations for n10 pipeline.
4246 * config/nds32/nds32.md (pipeline_model): Add n10.
4247 * config/nds32/nds32.opt (mcpu): Support n10 pipeline cpus.
4248 * config/nds32/pipelines.md: Include n10 settings.
4250 2018-05-13 Monk Chiang <sh.chiang04@gmail.com>
4251 Kito Cheng <kito.cheng@gmail.com>
4252 Chung-Ju Wu <jasonwucj@gmail.com>
4254 * config.gcc (nds32be-*-*): Handle --with-ext-dsp.
4255 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
4256 Add enum values for DSP extension instructions.
4257 * config/nds32/constraints.md (Iu06, IU06, CVp5, CVs5, CVs2, CVhi):
4259 * config/nds32/iterators.md (shifts, shiftrt, sat_plus, all_plus,
4260 sat_minus, all_minus, plus_minus, extend, sumax, sumin, sumin_max):
4262 (su, zs, uk, opcode, add_rsub, add_sub): New code attributes.
4263 * config/nds32/nds32-dspext.md: New file for DSP implementation.
4264 * config/nds32/nds32-intrinsic.c: Implementation of DSP extension.
4265 * config/nds32/nds32-intrinsic.md: Likewise.
4266 * config/nds32/nds32_intrinsic.h: Likewise.
4267 * config/nds32/nds32-md-auxiliary.c: Likewise.
4268 * config/nds32/nds32-memory-manipulation.c: Consider DSP extension.
4269 * config/nds32/nds32-predicates.c (const_vector_to_hwint): New.
4270 (nds32_valid_CVp5_p, nds32_valid_CVs5_p): New.
4271 (nds32_valid_CVs2_p, nds32_valid_CVhi_p): New.
4272 * config/nds32/nds32-protos.h: New declarations for DSP extension.
4273 * config/nds32/nds32-utils.c (extract_mac_non_acc_rtx): New case
4274 TYPE_DMAC in switch statement.
4275 * config/nds32/nds32.c: New checking and implementation for DSP
4276 extension instructions.
4277 * config/nds32/nds32.h: Likewise.
4278 * config/nds32/nds32.md: Likewise.
4279 * config/nds32/nds32.opt (mhw-abs, mext-dsp): New options.
4280 * config/nds32/predicates.md: Implement new predicates for DSP
4283 2018-05-11 Michael Meissner <meissner@linux.vnet.ibm.com>
4285 * config/rs6000/rs6000.md (mov<mode>_softfloat, FMOVE32):
4286 Reformat alternatives and attributes so it is easier to identify
4287 which constraints/attributes go with which instruction.
4288 (mov<mode>_hardfloat32, FMOVE64): Likewise.
4289 (mov<mode>_softfloat32, FMOVE64): Likewise.
4290 (mov<mode>_hardfloat64, FMOVE64): Likewise.
4291 (mov<mode>_softfloat64, FMOVE64): Likewise.
4293 2018-05-11 Kelvin Nilsen <kelvin@gcc.gnu.org>
4295 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4297 (Basic PowerPC Built-in Functions): The new name of the
4298 subsection previously known as "PowerPC Built-in Functions".
4299 (Basic PowerPC Built-in Functions Available on all Configurations):
4301 (Basic PowerPC Built-in Functions Available on ISA 2.05): Likewise.
4302 (Basic PowerPC Built-in Functions Available on ISA 2.06): Likewise.
4303 (Basic PowerPC Built-in Functions Available on ISA 2.07): Likewise.
4304 (Basic PowerPC Built-in Functions Available on ISA 3.0): Likewise.
4306 2018-05-11 Martin Jambor <mjambor@suse.cz>
4309 * ipa-cp.c (intersect_with_plats): Check that the lattice contains
4312 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4315 * config/arm/arm-cpus.in (fgroup ARMv6m): Add be8 feature.
4317 2018-05-11 Sebastian Peryt <sebastian.peryt@intel.com>
4319 * common/config/i386/i386-common.c (OPTION_MASK_ISA_WAITPKG_SET,
4320 OPTION_MASK_ISA_WAITPKG_UNSET): New defines.
4321 (ix86_handle_option): Handle -mwaitpkg.
4322 * config.gcc: New header.
4323 * config/i386/cpuid.h (bit_WAITPKG): New bit.
4324 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mwaitpkg.
4325 * config/i386/i386-builtin-types.def ((UINT8, UNSIGNED, UINT64)): New
4327 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4328 OPTION_MASK_ISA_WAITPKG.
4329 * config/i386/i386.c (ix86_target_string): Add -mwaitpkg.
4330 (ix86_option_override_internal): Add PTA_WAITPKG.
4331 (ix86_valid_target_attribute_inner_p): Add -mwaitpkg.
4332 (enum ix86_builtins): Add IX86_BUILTIN_UMONITOR, IX86_BUILTIN_UMWAIT,
4333 IX86_BUILTIN_TPAUSE.
4334 (ix86_init_mmx_sse_builtins): Define __builtin_ia32_umonitor,
4335 __builtin_ia32_umwait and __builtin_ia32_tpause.
4336 (ix86_expand_builtin): Expand IX86_BUILTIN_UMONITOR,
4337 IX86_BUILTIN_UMWAIT, IX86_BUILTIN_TPAUSE.
4338 * config/i386/i386.h (TARGET_WAITPKG, TARGET_WAITPKG_P): New.
4339 * config/i386/i386.md (UNSPECV_UMWAIT, UNSPECV_UMONITOR,
4340 UNSPECV_TPAUSE): New.
4341 (umwait, umwait_rex64, umonitor_<mode>, tpause, tpause_rex64): New.
4342 * config/i386/i386.opt: Add -mwaitpkg.
4343 * config/i386/waitpkgintrin.h: New file.
4344 * config/i386/x86intrin.h: New header.
4345 * doc/invoke.texi: Add -mwaitpkg.
4347 2018-05-11 Richard Earnshaw <rearnsha@arm.com>
4350 * config/arm/arm-cpus.in: Add comment that ARMv6-m and ARMv6S-m are now
4352 (cortex-m0): Use armv6s-m isa.
4353 (cortex-m0plus): Likewise.
4354 (cortex-m1): Likewise.
4355 (cortex-m0.small-multiply): Likewise.
4356 (cortex-m0plus.small-multiply): Likewise.
4357 (cortex-m1.small-multiply): Likewise.
4359 2018-05-11 Allan Sandfeld Jensen <allan.jensen@qt.io>
4360 Jakub Jelinek <jakub@redhat.com>
4362 PR tree-optimization/85692
4363 * tree-ssa-forwprop.c (simplify_vector_constructor): Try two
4364 source permute as well.
4366 2018-05-11 Martin Liska <mliska@suse.cz>
4369 * doc/extend.texi: Document LLVM style format for no_sanitize
4372 2018-05-10 Michael Meissner <meissner@linux.vnet.ibm.com>
4374 * config/rs6000/rs6000.c (mode_supports_dq_form): Rename
4375 mode_supports_vsx_dform_quad to mode_supports_dq_form.
4376 (mode_supports_vsx_dform_quad): Likewise.
4377 (mode_supports_vmx_dform): Move these functions to be next to the
4378 other mode_supports functions.
4379 (mode_supports_dq_form): Likewise.
4380 (quad_address_p): Change calls of mode_supports_vsx_dform_quad to
4381 mode_supports_dq_form.
4382 (reg_offset_addressing_ok_p): Likewise.
4383 (offsettable_ok_by_alignment): Likewise.
4384 (rs6000_legitimate_offset_address_p): Likewise.
4385 (legitimate_lo_sum_address_p): Likewise.
4386 (rs6000_legitimize_address): Likewise.
4387 (rs6000_legitimize_reload_address): Likewise.
4388 (rs6000_secondary_reload_inner): Likewise.
4389 (rs6000_preferred_reload_class): Likewise.
4390 (rs6000_output_move_128bit): Likewise.
4392 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
4394 * config/i386/i386.c (ix86_expand_builtin) <case IX86_BUILTIN_RDPID>:
4395 Generate SImode target register for null target.
4396 <case IX86_BUILTIN_XGETBV>: Ditto.
4397 <case IX86_BUILTIN_XSETBV>: Optimize LSHIFTRT generation.
4398 * config/i386/xsaveintrin.h (_xgetbv): Add missing return.
4400 2018-05-10 Carl Love <cel@us.ibm.com>
4402 * config/rs6000/rs6000.md (prefetch): Generate ISA 2.06 instructions
4403 dcbtt and dcbtstt if operands[2] is 0.
4405 2018-05-10 Uros Bizjak <ubizjak@gmail.com>
4408 * config/i386/sse.md (usadv64qi): New expander.
4410 2018-05-10 Segher Boessenkool <segher@kernel.crashing.org>
4412 * config/rs6000/altivec.md (altivec_vmrghb, altivec_vmrghh,
4413 altivec_vmrghw, altivec_vmrglb, altivec_vmrglh, altivec_vmrglw): Remove
4414 -maltivec=be support.
4415 (vec_widen_umult_even_v16qi, vec_widen_smult_even_v16qi,
4416 vec_widen_umult_even_v8hi, vec_widen_smult_even_v8hi,
4417 vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
4418 vec_widen_umult_odd_v16qi, vec_widen_smult_odd_v16qi,
4419 vec_widen_umult_odd_v8hi, vec_widen_smult_odd_v8hi,
4420 vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si, altivec_vpkpx,
4421 altivec_vpks<VI_char>ss, altivec_vpks<VI_char>us,
4422 altivec_vpku<VI_char>us, altivec_vpku<VI_char>um, altivec_vsum2sws,
4423 altivec_vsumsws): Adjust.
4424 (altivec_vspltb *altivec_vspltb_internal, altivec_vsplth,
4425 *altivec_vsplth_internal, altivec_vspltw, *altivec_vspltw_internal,
4426 altivec_vspltsf, *altivec_vspltsf_internal): Remove -maltivec=be
4428 (altivec_vperm_<mode>, altivec_vperm_<mode>_uns,
4429 altivec_vupkhs<VU_char>, altivec_vupkls<VU_char>, altivec_vupkhpx,
4430 altivec_vupklpx, altivec_lvsl, altivec_lvsr): Adjust.
4431 (altivec_lve<VI_char>x): Delete expand.
4432 (*altivec_lve<VI_char>x_internal): Rename to...
4433 (altivec_lve<VI_char>x): ... this.
4434 (altivec_lvxl_<mode>): Delete expand.
4435 (*altivec_lvxl_<mode>_internal): Rename to ...
4436 (altivec_lvxl_<mode>): ... this.
4437 (altivec_stvxl_<mode>): Delete expand.
4438 (*altivec_stvxl_<mode>_internal): Rename to ...
4439 (altivec_stvxl_<mode>): ... this.
4440 (altivec_stve<VI_char>x): Delete expand.
4441 (*altivec_stve<VI_char>x_internal): Rename to ...
4442 (altivec_stve<VI_char>x): ... this.
4443 (doublee<mode>2, unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2,
4444 doubleh<mode>2, unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2,
4445 reduc_plus_scal_<mode>): Adjust.
4446 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Adjust
4448 (rs6000_cpu_cpp_builtins): Adjust.
4449 (altivec_resolve_overloaded_builtin): Remove -maltivec=be support.
4450 * config/rs6000/rs6000-protos.h (altivec_expand_lvx_be,
4451 altivec_expand_stvx_be, altivec_expand_stvex_be): Delete.
4452 * config/rs6000/rs6000.c (rs6000_option_override_internal): Remove
4453 -maltivec=be support.
4454 (rs6000_split_vec_extract_var): Adjust.
4455 (rs6000_split_v4si_init): Adjust.
4456 (swap_selector_for_mode): Delete.
4457 (altivec_expand_lvx_be, altivec_expand_stvx_be,
4458 altivec_expand_stvex_be): Delete.
4459 (altivec_expand_lv_builtin, altivec_expand_stv_builtin): Remove
4460 -maltivec=be support.
4461 (rs6000_gimple_fold_builtin): Ditto.
4462 (rs6000_generate_float2_double_code, rs6000_generate_float2_code):
4464 * config/rs6000/rs6000.h (VECTOR_ELT_ORDER_BIG): Delete.
4465 (TARGET_DIRECT_MOVE_64BIT): Adjust.
4466 * config/rs6000/rs6000.md (split for extendsidi2 for vectors): Adjust.
4467 * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Delete.
4468 * config/rs6000/vsx.md (floate<mode>, unsfloatev2di, floato<mode>,
4469 unsfloatov2di, vsignedo_v2df, vsignede_v2df, vunsignedo_v2df,
4470 vunsignede_v2df, vsx_extract_<mode>_p9, *vsx_extract_si,
4471 *vsx_extract_<mode>_p8, *vsx_extract_si_<uns>float_df,
4472 *vsx_extract_si_<uns>float_<mode>, vsx_set_<mode>_p9, vsx_set_v4sf_p9,
4473 *vsx_insert_extract_v4sf_p9, *vsx_insert_extract_v4sf_p9_2, and an
4474 anonymous split): Adjust.
4475 (vsx_mergel_<mode>, vsx_mergeh_<mode>): Remove -maltivec=be support.
4476 (vsx_xxspltd_<mode>, extract4b, insert4b): Adjust.
4478 2018-05-10 Eric Botcazou <ebotcazou@adacore.com>
4480 * configure.ac (gcc_gxx_include_dir_add_sysroot): Set it to 1 only
4481 when --with-gxx-include-dir is also specified.
4482 * configure: Regenerate.
4484 2018-05-09 Jim Wilson <jimw@sifive.com>
4487 * config.gcc (riscv*-*-*): Handle --with-multilib-list.
4488 * config/riscv/t-withmultilib: New.
4489 * config/riscv/withmultilib.h: New.
4490 * doc/install.texi: Document RISC-V --with-multilib-list support.
4492 2018-05-09 Richard Biener <rguenther@suse.de>
4494 * tree-vect-slp.c (vect_bb_slp_scalar_cost): Fill a cost
4496 (vect_bb_vectorization_profitable_p): Adjust. Compute
4497 actual scalar cost using the cost vector and the add_stmt_cost
4500 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4502 PR rtl-optimization/85645
4503 * config/rs6000/rs6000.c (rs6000_emit_prologue_components): Put a SET
4504 in the REG_CFA_REGISTER note for LR, don't leave it empty.
4506 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4508 PR rtl-optimization/85645
4509 * shrink-wrap.c (spread_components): Return a boolean saying if
4510 anything was changed.
4511 (try_shrink_wrapping_separate): Iterate spread_components until
4512 nothing changes anymore.
4514 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4516 PR rtl-optimization/85645
4517 * regrename.c (build_def_use): Also kill the chains that include the
4518 destination of a REG_CFA_REGISTER note.
4520 2018-05-09 Segher Boessenkool <segher@kernel.crashing.org>
4522 PR rtl-optimization/85645
4523 * regcprop.c (copyprop_hardreg_forward_1): Don't propagate into an
4524 insn that has a REG_CFA_REGISTER note.
4526 2018-05-09 Richard Sandiford <richard.sandiford@linaro.org>
4528 * cfgexpand.c (expand_clobber): New function.
4529 (expand_gimple_stmt_1): Use it.
4530 * tree-vect-stmts.c (vect_clobber_variable): New function,
4532 (vectorizable_simd_clone_call): ...here.
4533 (vectorizable_store): Emit a clobber either side of an
4534 IFN_STORE_LANES sequence.
4535 (vectorizable_load): Emit a clobber after an IFN_LOAD_LANES sequence.
4537 2018-05-09 Tom de Vries <tom@codesourcery.com>
4540 * config/nvptx/nvptx.md (define_insn "trap", define_insn "trap_if_true")
4541 (define_insn "trap_if_false"): Add exit after trap.
4543 2018-05-09 Eric Botcazou <ebotcazou@adacore.com>
4545 PR rtl-optimization/85638
4546 * bb-reorder.c: Include common/common-target.h.
4547 (create_forwarder_block): New function extracted from...
4548 (fix_up_crossing_landing_pad): ...here. Rename into...
4549 (dw2_fix_up_crossing_landing_pad): ...this.
4550 (sjlj_fix_up_crossing_landing_pad): New function.
4551 (find_rarely_executed_basic_blocks_and_crossing_edges): In SJLJ mode,
4552 call sjlj_fix_up_crossing_landing_pad if there are incoming EH edges
4553 from both partitions and exit the loop after one iteration.
4555 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
4558 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4560 (Basic PowerPC Built-in Functions): The new name of the
4561 subsection previously known as "PowerPC Built-in Functions".
4562 (Basic PowerPC Built-in Functions Available on all Configurations):
4564 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
4566 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
4568 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
4570 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
4573 2018-05-08 Jim Wilson <jimw@sifive.com>
4575 * config/riscv/linux.h (MUSL_ABI_SUFFIX): Delete unnecessary backslash.
4576 (LD_EMUL_SUFFIX): New.
4577 (LINK_SPEC): Use it.
4579 2018-05-08 Kelvin Nilsen <kelvin@gcc.gnu.org>
4581 * doc/extend.texi (PowerPC Built-in Functions): Rename this
4583 (Basic PowerPC Built-in Functions): The new name of the
4584 subsection previously known as "PowerPC Built-in Functions".
4585 (Basic PowerPC Built-in Functions Available on all Configurations):
4587 (Basic PowerPC Built-in Functions Available on ISA 2.05): New
4589 (Basic PowerPC Built-in Functions Available on ISA 2.06): New
4591 (Basic PowerPC Built-in Functions Available on ISA 2.07): New
4593 (Basic PowerPC Built-in Functions Available on ISA 3.0): New
4596 2018-05-08 Jakub Jelinek <jakub@redhat.com>
4599 * config/i386/i386.md: Add peepholes for mem {+,-,&,|,^}= x; mem != 0
4600 after cmpelim optimization.
4602 2018-05-08 Olga Makhotina <olga.makhotina@intel.com>
4604 * config.gcc: Support "goldmont".
4605 * config/i386/driver-i386.c (host_detect_local_cpu): Detect "goldmont".
4606 * config/i386/i386-c.c (ix86_target_macros_internal): Handle
4608 * config/i386/i386.c (m_GOLDMONT): Define.
4609 (processor_target_table): Add "goldmont".
4610 (PTA_GOLDMONT): Define.
4611 (ix86_lea_outperforms): Add TARGET_GOLDMONT.
4612 (get_builtin_code_for_version): Handle PROCESSOR_GOLDMONT.
4613 (fold_builtin_cpu): Add M_INTEL_GOLDMONT.
4614 (fold_builtin_cpu): Add "goldmont".
4615 (ix86_add_stmt_cost): Add TARGET_GOLDMONT.
4616 (ix86_option_override_internal): Add "goldmont".
4617 * config/i386/i386.h (processor_costs): Define TARGET_GOLDMONT.
4618 (processor_type): Add PROCESSOR_GOLDMONT.
4619 * config/i386/i386.md: Add CPU "glm".
4620 * config/i386/glm.md: New file.
4621 * config/i386/x86-tune.def: Add m_GOLDMONT.
4622 * doc/invoke.texi: Add goldmont as x86 -march=/-mtune= CPU type.
4624 2018-05-08 Jakub Jelinek <jakub@redhat.com>
4627 * config/i386/i386.c (ix86_expand_sse2_abs): Handle E_V2DImode and
4629 * config/i386/sse.md (abs<mode>2): Use VI_AVX2 iterator instead of
4630 VI1248_AVX512VL_AVX512BW. Handle V2DImode and V4DImode if not
4631 TARGET_AVX512VL using ix86_expand_sse2_abs. Formatting fixes.
4634 * config/i386/i386.c (ix86_fold_builtin): Handle
4635 IX86_BUILTIN_{,P}MOVMSK{PS,PD,B}{,128,256}.
4638 * config/i386/sse.md (ssequaterinsnmode): New mode attribute.
4639 (*<extract_type>_vinsert<shuffletype><extract_suf>_0): New pattern.
4641 2018-05-08 Richard Earnshaw <rearnsha@arm.com>
4644 * config/arm/parsecpu.awk (check_cpu): Fix operator precedence.
4645 (check_arch): Likewise.
4646 (check_fpu): Return the result rather than printing it.
4647 (end arch): Fix operator precedence.
4648 (end cpu): Likewise.
4649 (END): Print the result from check_fpu.
4651 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
4652 Alan Hayward <alan.hayward@arm.com>
4653 David Sherwood <david.sherwood@arm.com>
4655 * config/aarch64/aarch64-sve.md (*pred_cmp<cmp_op><mode>_combine)
4656 (*pred_cmp<cmp_op><mode>, *fcm<cmp_op><mode>_and_combine)
4657 (*fcmuo<mode>_and_combine, *fcm<cmp_op><mode>_and)
4658 (*fcmuo<mode>_and): New patterns.
4660 2018-05-08 Richard Sandiford <richard.sandiford@linaro.org>
4662 * config/aarch64/iterators.md (UNSPEC_COND_LO, UNSPEC_COND_LS)
4663 (UNSPEC_COND_HI, UNSPEC_COND_HS, UNSPEC_COND_UO): Delete.
4664 (SVE_INT_CMP, SVE_FP_CMP): New code iterators.
4665 (cmp_op, sve_imm_con): New code attributes.
4666 (SVE_COND_INT_CMP, imm_con): Delete.
4667 (cmp_op): Remove above unspecs from int attribute.
4668 * config/aarch64/aarch64-sve.md (*vec_cmp<cmp_op>_<mode>): Rename
4670 (*cmp<cmp_op><mode>): ...this. Use UNSPEC_MERGE_PTRUE instead of
4671 comparison-specific unspecs.
4672 (*vec_cmp<cmp_op>_<mode>_ptest): Rename to...
4673 (*cmp<cmp_op><mode>_ptest): ...this and adjust likewise.
4674 (*vec_cmp<cmp_op>_<mode>_cc): Rename to...
4675 (*cmp<cmp_op><mode>_cc): ...this and adjust likewise.
4676 (*vec_fcm<cmp_op><mode>): Rename to...
4677 (*fcm<cmp_op><mode>): ...this and adjust likewise.
4678 (*vec_fcmuo<mode>): Rename to...
4679 (*fcmuo<mode>): ...this and adjust likewise.
4680 (*pred_fcm<cmp_op><mode>): New pattern.
4681 * config/aarch64/aarch64.c (aarch64_emit_unop, aarch64_emit_binop)
4682 (aarch64_emit_sve_ptrue_op, aarch64_emit_sve_ptrue_op_cc): New
4684 (aarch64_unspec_cond_code): Remove handling of LTU, GTU, LEU, GEU
4686 (aarch64_gen_unspec_cond, aarch64_emit_unspec_cond): Delete.
4687 (aarch64_emit_sve_predicated_cond): New function.
4688 (aarch64_expand_sve_vec_cmp_int): Use aarch64_emit_sve_ptrue_op_cc.
4689 (aarch64_emit_unspec_cond_or): Replace with...
4690 (aarch64_emit_sve_or_conds): ...this new function. Use
4691 aarch64_emit_sve_ptrue_op for the individual comparisons and
4692 aarch64_emit_binop to OR them together.
4693 (aarch64_emit_inverted_unspec_cond): Replace with...
4694 (aarch64_emit_sve_inverted_cond): ...this new function. Use
4695 aarch64_emit_sve_ptrue_op for the comparison and
4696 aarch64_emit_unop to invert the result.
4697 (aarch64_expand_sve_vec_cmp_float): Update after the above
4698 changes. Use aarch64_emit_sve_ptrue_op for native comparisons.
4700 2018-05-07 Nathan Sidwell <nathan@acm.org>
4702 * doc/invoke.texi (C++ Dialect Options): Remove -ffor-scope.
4703 * doc/extend.texi (Deprecated Features): Remove -fno-for-scope
4704 (Backwards Compatibility): Likewise.
4706 2018-05-07 Luis Machado <luis.machado@linaro.org>
4710 2018-05-07 Luis Machado <luis.machado@linaro.org>
4712 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4713 <prefetch_dynamic_strides>: New const bool field.
4714 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4715 prefetch_dynamic_strides.
4716 (exynosm1_prefetch_tune): Likewise.
4717 (thunderxt88_prefetch_tune): Likewise.
4718 (thunderx_prefetch_tune): Likewise.
4719 (thunderx2t99_prefetch_tune): Likewise.
4720 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
4722 (aarch64_override_options_internal): Update to set
4723 PARAM_PREFETCH_DYNAMIC_STRIDES.
4724 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4725 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4726 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4727 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4728 prefetch-dynamic-strides setting.
4730 2018-05-07 Luis Machado <luis.machado@linaro.org>
4732 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4733 <minimum_stride>: New const int field.
4734 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4735 minimum_stride field.
4736 (exynosm1_prefetch_tune): Likewise.
4737 (thunderxt88_prefetch_tune): Likewise.
4738 (thunderx_prefetch_tune): Likewise.
4739 (thunderx2t99_prefetch_tune): Likewise.
4740 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
4741 (aarch64_override_options_internal): Update to set
4742 PARAM_PREFETCH_MINIMUM_STRIDE.
4743 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4744 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4745 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4746 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4747 stride is constant and is below the minimum stride threshold.
4749 2018-05-07 Luis Machado <luis.machado@linaro.org>
4751 * config/aarch64/aarch64.c (qdf24xx_prefetch_tune) <l2_cache_size>: Set
4754 2018-05-07 Luis Machado <luis.machado@linaro.org>
4756 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4757 <prefetch_dynamic_strides>: New const bool field.
4758 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4759 prefetch_dynamic_strides.
4760 (exynosm1_prefetch_tune): Likewise.
4761 (thunderxt88_prefetch_tune): Likewise.
4762 (thunderx_prefetch_tune): Likewise.
4763 (thunderx2t99_prefetch_tune): Likewise.
4764 (qdf24xx_prefetch_tune): Likewise. Set prefetch_dynamic_strides
4766 (aarch64_override_options_internal): Update to set
4767 PARAM_PREFETCH_DYNAMIC_STRIDES.
4768 * doc/invoke.texi (prefetch-dynamic-strides): Document new option.
4769 * params.def (PARAM_PREFETCH_DYNAMIC_STRIDES): New.
4770 * params.h (PARAM_PREFETCH_DYNAMIC_STRIDES): Define.
4771 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Account for
4772 prefetch-dynamic-strides setting.
4774 2018-05-07 Luis Machado <luis.machado@linaro.org>
4776 * config/aarch64/aarch64-protos.h (cpu_prefetch_tune)
4777 <minimum_stride>: New const int field.
4778 * config/aarch64/aarch64.c (generic_prefetch_tune): Update to include
4779 minimum_stride field.
4780 (exynosm1_prefetch_tune): Likewise.
4781 (thunderxt88_prefetch_tune): Likewise.
4782 (thunderx_prefetch_tune): Likewise.
4783 (thunderx2t99_prefetch_tune): Likewise.
4784 (qdf24xx_prefetch_tune): Likewise. Set minimum_stride to 2048.
4785 (aarch64_override_options_internal): Update to set
4786 PARAM_PREFETCH_MINIMUM_STRIDE.
4787 * doc/invoke.texi (prefetch-minimum-stride): Document new option.
4788 * params.def (PARAM_PREFETCH_MINIMUM_STRIDE): New.
4789 * params.h (PARAM_PREFETCH_MINIMUM_STRIDE): Define.
4790 * tree-ssa-loop-prefetch.c (should_issue_prefetch_p): Return false if
4791 stride is constant and is below the minimum stride threshold.
4793 2018-05-06 Jakub Jelinek <jakub@redhat.com>
4796 * cfgexpand.c (expand_asm_stmt): Don't create a temporary if
4797 the type is addressable. Don't force op into register if it has
4800 2018-05-05 Roland McGrath <mcgrathr@google.com>
4803 * varasm.c (default_section_type_flags): Set SECTION_NOTYPE for
4804 any section for which we don't know a specific type it should have,
4805 regardless of name. Previously this was done only for the exact
4806 names ".init_array", ".fini_array", and ".preinit_array".
4807 (default_elf_asm_named_section): Add comment about
4808 relationship with default_section_type_flags and SECTION_NOTYPE.
4809 (get_section): Don't consider it a type conflict if one side has
4810 SECTION_NOTYPE and the other doesn't, as long as neither has the
4811 SECTION_BSS et al used in the default_section_type_flags logic.
4813 2018-05-05 Tom de Vries <tom@codesourcery.com>
4816 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_3): Define.
4817 (workaround_barsyncs): New function.
4818 (nvptx_reorg): Use workaround_barsyncs.
4819 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
4820 (define_expand "nvptx_membar_cta"): New define_expand.
4821 (define_insn "*nvptx_membar_cta"): New insn.
4823 2018-05-04 Pekka Jääskeläinen <pekka.jaaskelainen@parmance.com>
4825 * brig-builtins.def: Add consts to ptrs etc. in BRIG builtin defs.
4826 To improve optimization opportunities.
4827 * builtin-types.def: The new needed builtin types for the above.
4829 2018-05-04 Richard Biener <rguenther@suse.de>
4831 * bb-reorder.c (sanitize_hot_paths): Release hot_bbs_to_check.
4832 * gimple-ssa-store-merging.c
4833 (imm_store_chain_info::output_merged_store): Remove redundant create,
4834 release split_store vector contents on failure.
4835 * tree-vect-slp.c (vect_schedule_slp_instance): Avoid leaking
4836 scalar stmt vector on cache hit.
4838 2018-05-04 Segher Boessenkool <segher@kernel.crashing.org>
4840 * common/config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
4842 * config.gcc (powerpc-xilinx-eabi*): Remove.
4843 * config/rs6000/predicates.md (easy_fp_constant): Remove Xilinx FP
4845 (fusion_addis_mem_combo_load): Ditto.
4846 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove Xilinx
4848 (rs6000_cpu_cpp_builtins): Ditto.
4849 * config/rs6000/rs6000-linux.c
4850 (rs6000_linux_float_exceptions_rounding_supported_p): Ditto.
4851 * config/rs6000/rs6000-opts.h (enum fpu_type_t): Delete.
4852 * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove Xilinx FP
4854 (rs6000_setup_reg_addr_masks): Ditto.
4855 (rs6000_init_hard_regno_mode_ok): Ditto.
4856 (rs6000_option_override_internal): Ditto.
4857 (legitimate_lo_sum_address_p): Ditto.
4858 (rs6000_legitimize_address): Ditto.
4859 (rs6000_legitimize_reload_address): Ditto.
4860 (rs6000_legitimate_address_p): Ditto.
4861 (abi_v4_pass_in_fpr): Ditto.
4862 (setup_incoming_varargs): Ditto.
4863 (rs6000_gimplify_va_arg): Ditto.
4864 (rs6000_split_multireg_move): Ditto.
4865 (rs6000_savres_strategy): Ditto.
4866 (rs6000_emit_prologue_components): Ditto.
4867 (rs6000_emit_epilogue_components): Ditto.
4868 (rs6000_emit_prologue): Ditto.
4869 (rs6000_emit_epilogue): Ditto.
4870 (rs6000_elf_file_end): Ditto.
4871 (rs6000_function_value): Ditto.
4872 (rs6000_libcall_value): Ditto.
4873 * config/rs6000/rs6000.h: Ditto.
4874 (TARGET_MINMAX_SF, TARGET_MINMAX_DF): Delete, merge to ...
4875 (TARGET_MINMAX): ... this. New.
4876 (TARGET_SF_FPR, TARGET_DF_FPR, TARGET_SF_INSN, TARGET_DF_INSN): Delete.
4877 * config/rs6000/rs6000.md: Remove Xilinx FP support.
4878 (*movsi_internal1_single): Delete.
4879 * config/rs6000/rs6000.opt (msingle-float, mdouble-float, msimple-fpu,
4880 mfpu=, mxilinx-fpu): Delete.
4881 * config/rs6000/singlefp.h: Delete.
4882 * config/rs6000/sysv4.h: Remove Xilinx FP support.
4883 * config/rs6000/t-rs6000: Ditto.
4884 * config/rs6000/t-xilinx: Delete.
4885 * gcc/config/rs6000/titan.md: Adjust for fp_type removal.
4886 * gcc/config/rs6000/vsx.md: Remove Xilinx FP support.
4887 (VStype_simple): Delete.
4888 (VSfptype_simple, VSfptype_mul, VSfptype_div, VSfptype_sqrt): Delete.
4889 * config/rs6000/xfpu.h: Delete.
4890 * config/rs6000/xfpu.md: Delete.
4891 * config/rs6000/xilinx.h: Delete.
4892 * config/rs6000/xilinx.opt: Delete.
4893 * gcc/doc/invoke.texi (RS/6000 and PowerPC Options): Remove
4894 -msingle-float, -mdouble-float, -msimple-fpu, -mfpu=, and -mxilinx-fpu.
4896 2018-05-04 Tom de Vries <tom@codesourcery.com>
4899 * builtins.c (expand_builtin_goacc_parlevel_id_size): Handle null target
4902 2018-05-04 Richard Biener <rguenther@suse.de>
4905 * tree-complex.c (update_complex_assignment): We are always in SSA form.
4906 (expand_complex_div_wide): Likewise.
4907 (expand_complex_operations_1): Likewise.
4908 (expand_complex_libcall): Preserve EH info of the original stmt.
4909 (tree_lower_complex): Handle removed blocks.
4910 * tree.c (build_common_builtin_nodes): Do not set ECF_NOTRHOW
4911 on complex multiplication and division libcall builtins.
4913 2018-05-04 Richard Biener <rguenther@suse.de>
4916 * fold-const.c (negate_expr_p): Restrict negation of operand
4917 zero of a division to when we know that can happen without
4919 (fold_negate_expr_1): Likewise.
4921 2018-05-04 Jakub Jelinek <jakub@redhat.com>
4924 * real.h (real_nextafter): Declare.
4925 * real.c (real_nextafter): New function.
4926 * fold-const-call.c (fold_const_nextafter): New function.
4927 (fold_const_call_sss): Call it for CASE_CFN_NEXTAFTER and
4928 CASE_CFN_NEXTTOWARD.
4929 (fold_const_call_1): For CASE_CFN_NEXTTOWARD call fold_const_call_sss
4930 even when arg1_mode is different from arg0_mode.
4932 2018-05-03 Nathan Sidwell <nathan@acm.org>
4934 * doc/extend.texi (Deprecated Features): Remove
4936 (Backwards Compatibility): Likewise.
4937 * doc/invoke.texi (C++ Language Options): Likewise.
4938 (C++ Dialect Options): Likewise.
4940 2018-05-03 Jakub Jelinek <jakub@redhat.com>
4943 * config/i386/avx512fintrin.h (_mm512_mullox_epi64,
4944 _mm512_mask_mullox_epi64): New intrinsics.
4946 2018-05-03 Tom de Vries <tom@codesourcery.com>
4949 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
4950 dump files): Add offload-tree.
4952 2018-05-03 Richard Biener <rguenther@suse.de>
4954 PR tree-optimization/85615
4955 * tree-ssa-threadupdate.c (thread_block_1): Only allow exits
4956 to loops not nested in BBs loop father to avoid creating multi-entry
4959 2018-05-03 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
4961 PR tree-optimization/70291
4962 * tree-complex.c (expand_complex_libcall): Add type, inplace_p
4963 arguments. Change return type to tree. Emit libcall as a new
4964 statement rather than replacing existing one when inplace_p is true.
4965 (expand_complex_multiplication_components): New function.
4966 (expand_complex_multiplication): Expand floating-point complex
4967 multiplication using the above.
4968 (expand_complex_division): Rename inner_type parameter to type.
4969 Update expand_complex_libcall call-site.
4970 (expand_complex_operations_1): Update expand_complex_multiplication
4971 and expand_complex_division call-sites.
4973 2018-05-02 Jakub Jelinek <jakub@redhat.com>
4976 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
4977 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
4978 *<shift_insn><dwi>3_doubleword_mask_1): In condition require that
4979 the highest significant bit of the shift count mask is clear. In
4980 check whether and[sq]i3 is needed verify that all significant bits
4981 of the shift count other than the highest are set.
4983 2018-05-02 Tom de Vries <tom@codesourcery.com>
4986 * builtins.def (DEF_GOACC_BUILTIN_ONLY): Define.
4987 * omp-builtins.def (BUILT_IN_GOACC_PARLEVEL_ID)
4988 (BUILT_IN_GOACC_PARLEVEL_SIZE): New builtin.
4989 * builtins.c (expand_builtin_goacc_parlevel_id_size): New function.
4990 (expand_builtin): Call expand_builtin_goacc_parlevel_id_size.
4991 * doc/extend.texi (Other Builtins): Add __builtin_goacc_parlevel_id and
4992 __builtin_goacc_parlevel_size.
4994 2018-05-02 Richard Biener <rguenther@suse.de>
4996 PR tree-optimization/85597
4997 * tree-vect-stmts.c (vectorizable_operation): For ternary SLP
4998 do not use split vect_get_vec_defs call but call vect_get_slp_defs
5001 2018-05-02 Tom de Vries <tom@codesourcery.com>
5004 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5005 dump files): Add ltrans-tree.
5007 2018-05-02 Tom de Vries <tom@codesourcery.com>
5010 * doc/sourcebuild.texi (Commands for use in dg-final, Scan optimization
5011 dump files): Add wpa-ipa.
5013 2018-05-02 Segher Boessenkool <segher@kernel.crashing.org>
5015 * config.gcc (powerpc*-*-*): Remove paired.h. Unsupport the
5016 powerpc*-*-linux*paired* target.
5017 * config/rs6000/750cl.h: Delete.
5018 * config/rs6000/paired.h: Delete.
5019 * config/rs6000/paired.md: Delete.
5020 * config/rs6000/predicates.md (easy_vector_constant): Remove paired
5022 * config/rs6000/rs6000-builtin.def: Remove paired float support.
5023 * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
5024 comment. Remove paired float support.
5025 * config/rs6000/rs6000-modes.def: Remove V2SF and V2SI.
5026 * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete
5028 * config/rs6000/rs6000-protos.h (paired_expand_vector_init,
5029 paired_emit_vector_cond_expr, paired_expand_vector_move): Delete
5031 * config/rs6000/rs6000.c: Remove paired float support.
5032 (paired_expand_vector_init, paired_expand_vector_move,
5033 paired_emit_vector_compare, paired_emit_vector_cond_expr,
5034 (paired_expand_lv_builtin, paired_expand_stv_builtin,
5035 paired_expand_builtin, paired_expand_predicate_builtin,
5036 paired_init_builtins): Delete.
5037 * config/rs6000/rs6000.h: Remove paired float support.
5038 * config/rs6000/rs6000.md: Remove paired float support.
5039 (move_from_CR_ov_bit): Delete.
5040 * config/rs6000/rs6000.opt (mpaired): Delete.
5041 * config/rs6000/t-rs6000: Remove paired.md from MD_INCLUDES.
5042 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mpaired.
5044 2018-05-02 Richard Biener <rguenther@suse.de>
5047 * gimplify.c (gimplify_save_expr): When in SSA form allow
5048 SAVE_EXPRs to compute to SSA vars.
5050 2018-05-02 Jakub Jelinek <jakub@redhat.com>
5053 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask,
5054 *ashl<dwi>3_doubleword_mask_1, *<shift_insn><dwi>3_doubleword_mask,
5055 *<shift_insn><dwi>3_doubleword_mask_1): If and[sq]i3 is needed, don't
5056 clobber operands[2], instead use a new pseudo. Formatting fixes.
5058 2018-05-02 Richard Sandiford <richard.sandiford@linaro.org>
5060 PR tree-optimization/85586
5061 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence): Only
5062 exit early for statements in the same group if the accesses are
5065 2018-05-02 Tom de Vries <tom@codesourcery.com>
5068 * lto-wrapper.c (compile_offload_image): Add "could not find mkoffload"
5071 2018-05-01 Marc Glisse <marc.glisse@inria.fr>
5073 PR tree-optimization/85143
5074 * match.pd (A<B&A<C): Extend to BIT_IOR_EXPR.
5076 2018-05-01 Tom de Vries <tom@codesourcery.com>
5079 * config/nvptx/mkoffload.c (main): Suggest using -B in "offload compiler
5080 not found" error message.
5082 2018-05-01 Tom de Vries <tom@codesourcery.com>
5085 * vec.h (VEC_ORDERED_REMOVE_IF, VEC_ORDERED_REMOVE_IF_FROM_TO): Define.
5086 * vec.c (test_ordered_remove_if): New function.
5087 (vec_c_tests): Call test_ordered_remove_if.
5088 * dwarf2cfi.c (connect_traces): Use VEC_ORDERED_REMOVE_IF_FROM_TO.
5089 * lto-streamer-out.c (prune_offload_funcs): Use VEC_ORDERED_REMOVE_IF.
5090 * tree-vect-patterns.c (vect_pattern_recog_1): Use
5091 VEC_ORDERED_REMOVE_IF.
5093 2018-05-01 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
5095 PR tree-optimization/82665
5096 * vr-values.c (vr_values::extract_range_from_binary_expr): Handle
5097 pointer subtraction where arguments come from a memchr call.
5099 2018-05-01 Jakub Jelinek <jakub@redhat.com>
5101 * configure.ac (LD_AS_NEEDED_OPTION, LD_NO_AS_NEEDED_OPTION): Use
5102 --push-state --as-needed and --pop-state instead of --as-needed and
5103 --no-as-needed if ld supports it.
5104 * configure: Regenerated.
5107 * doc/install.texi2html: Replace _002d with - and _002a with * in
5108 generated html files using sed.
5110 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5113 * gcc-rich-location.c (blank_line_before_p): New function.
5114 (use_new_line): New function.
5115 (gcc_rich_location::add_fixit_insert_formatted): New function.
5116 * gcc-rich-location.h
5117 (gcc_rich_location::add_fixit_insert_formatted): New function.
5119 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5121 * selftest.c (assert_streq): Rename "expected" and "actual" to
5122 "val1" and "val2". Extend NULL-handling to cover both inputs
5123 symmetrically, while still requiring both to be non-NULL for a pass.
5124 * selftest.h (assert_streq): Rename "expected" and "actual" to
5126 (ASSERT_EQ): Likewise.
5127 (ASSERT_EQ_AT): Likewise.
5128 (ASSERT_KNOWN_EQ): Likewise.
5129 (ASSERT_KNOWN_EQ_AT): Likewise.
5130 (ASSERT_NE): Likewise.
5131 (ASSERT_MAYBE_NE): Likewise.
5132 (ASSERT_MAYBE_NE_AT): Likewise.
5133 (ASSERT_STREQ): Likewise. Clarify that both must be non-NULL for
5134 the assertion to pass.
5135 (ASSERT_STREQ_AT): Likewise.
5137 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
5139 * doc/invoke.texi (-fpie, -fPIE): Fix grammar and clarify
5140 interaction with -pie.
5142 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5144 * selftest.h: Fix alphabetization of per-source-file selftest
5147 2018-04-30 Jason Merrill <jason@redhat.com>
5149 PR c++/61982 - dead stores to destroyed objects.
5150 * gimplify.c (gimplify_modify_expr): Simplify complex lvalue on LHS
5153 2018-04-30 Jason Merrill <jason@redhat.com>
5155 * tree.c (build_clobber): New.
5156 * tree.h: Declare it.
5157 * gimplify.c (gimplify_bind_expr, gimplify_target_expr): Use it.
5159 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5161 * diagnostic-show-locus.c (layout::layout): Update for
5162 location_get_source_line returning a char_span.
5163 (struct char_span): Move to input.h.
5164 (struct correction): Update for fields in char_span becoming
5166 (struct source_line): Update for location_get_source_line
5167 returning a char_span.
5168 (layout::print_line): Likewise.
5169 * edit-context.c (edited_file::print_content): Likewise.
5170 (edited_file::print_diff_hunk): Likewise.
5171 (edited_file::print_run_of_changed_lines): Likewise.
5172 (edited_file::get_num_lines): Likewise.
5173 (edited_line::edited_line): Likewise.
5174 * final.c (asm_show_source): Likewise.
5175 * input.c (location_get_source_line): Convert return type
5176 from const char * to char_span, losing the final "line_len"
5178 (dump_location_info): Update for the above.
5179 (get_substring_ranges_for_loc): Likewise. Use a char_span
5180 when handling the literal within the line.
5181 (test_reading_source_line): Update for location_get_source_line
5182 returning a char_span.
5183 * input.h (class char_span): Move here from
5184 diagnostic-show-locus.c, converting from a struct to a class.
5185 Make data members private.
5186 (char_span::operator bool): New.
5187 (char_span::length): New.
5188 (char_span::get_buffer): New.
5189 (char_span::operator[]): New.
5190 (char_span::subspan): Make const.
5191 (char_span::xstrdup): New.
5192 (location_get_source_line): Convert return type from const char *
5193 to char_span, losing the final "line_size" param.
5195 2018-04-30 Jan Hubicka <jh@suse.cz>
5197 * lto-wrapper.c (ltrans_priorities): New static var.
5198 (cmp_priority): New.
5199 (run_gcc): Read priorities and if doing parallel build order
5200 the Makefile by them.
5202 2018-04-30 David Malcolm <dmalcolm@redhat.com>
5204 * input.h (builtins_location_check): Convert to a STATIC_ASSERT.
5206 2018-04-30 Richard Biener <rguenther@suse.de>
5208 * tree-cfg.c (verify_address): Remove base argument, add
5209 flag whether to check TREE_ADDRESSABLE and do that.
5210 (verify_expr): Remove.
5211 (verify_types_in_gimple_reference): Add pieces from verify_expr.
5212 (verify_gimple_assign_single): Likewise.
5213 (verify_gimple_switch): Likewise.
5214 (verify_expr_location_1): Dereference tp once. Add (disabled)
5215 piece from verify_expr.
5216 (verify_gimple_in_cfg): Do not call verify_expr on all ops.
5218 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5220 * config/arc/linux.h (CLEAR_INSN_CACHE): Define.
5222 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5224 * config/arc/arc-protos.h (prepare_extend_operands): Remove.
5225 (small_data_pattern): Likewise.
5226 (arc_rewrite_small_data): Likewise.
5227 * config/arc/arc.c (LEGITIMATE_SMALL_DATA_OFFSET_P): Remove.
5228 (LEGITIMATE_SMALL_DATA_ADDRESS_P): Likewise.
5229 (get_symbol_alignment): New function.
5230 (legitimate_small_data_address_p): Likewise.
5231 (legitimate_scaled_address): Update, call
5232 legitimate_small_data_address_p.
5233 (output_sdata): New static variable.
5234 (arc_print_operand): Update how we handle small data operands.
5235 (arc_print_operand_address): Likewise.
5236 (arc_legitimate_address_p): Update, use
5237 legitimate_small_data_address_p.
5238 (arc_rewrite_small_data_p): Remove.
5239 (arc_rewrite_small_data_1): Likewise.
5240 (arc_rewrite_small_data): Likewise.
5241 (small_data_pattern): Likewise.
5242 (compact_sda_memory_operand): Update to use
5243 legitimate_small_data_address_p and get_symbol_alignment.
5244 (prepare_move_operands): Don't rewite sdata pattern.
5245 (prepare_extend_operands): Remove.
5246 * config/arc/arc.md (zero_extendqihi2): Don't rewrite sdata
5248 (zero_extendqisi2): Likewise.
5249 (zero_extendhisi2): Likewise.
5250 (extendqihi2): Likewise.
5251 (extendqisi2): Likewise.
5252 (extendhisi2): Likewise.
5256 * config/arc/constraints.md (Usd): Change it to memory constraint.
5258 2018-04-30 Claudiu Zissulescu <claziss@synopsys.com>
5260 * config/arc/arc.c (arc_split_move): Allow signed 6-bit constants
5261 as source of std instructions.
5262 * config/arc/arc.md (movsi_insn): Update pattern predicate to
5263 allow 6-bit constants as source for store instructions.
5264 (movdi_insn): Update instruction pattern to allow 6-bit constants
5265 as source for store instructions.
5267 2018-04-30 Jonathan Wakely <jwakely@redhat.com>
5269 * doc/invoke.texi (-fdebug-types-section): Fix grammar.
5271 2018-04-30 Nathan Sidwell <nathan@acm.org>
5272 Sandra Loosemore <sandra@codesourcery.com>
5274 * dumpfile.c (dump_open): Allow '-' for stdout.
5275 * doc/invoke.texi (Developer Options): Document dump filename
5276 determination early. Document stdin/stdout selection.
5278 2018-04-30 Andrew Sadek <andrew.sadek.se@gmail.com>
5280 Microblaze Target: PIC data text relative
5282 * config/microblaze/microblaze.opt: add new option -mpic-data-text-rel.
5283 * config/microblaze/microblaze-protos.h (microblaze_constant_address_p):
5285 * gcc/config/microblaze/microblaze.h (microblaze_constant_address_p):
5286 CONSTANT_ADDRESS_P definition to microblaze_constant_address_p.
5287 * config/microblaze/microblaze.c (TARGET_PIC_DATA_TEXT_REL):
5288 New addressing mode for data-text relative position indepenedent code.
5289 (microblaze_classify_unspec): add 'UNSPEC_TEXT' case ->
5290 'ADDRESS_SYMBOLIC_TXT_REL'.
5291 (microblaze_classify_address): Add handling for UNSPEC + CONST_INT.
5292 (microblaze_legitimate_pic_operand): Exclude function calls from
5293 pic operands in case of TARGET_PIC_DATA_TEXT_REL option.
5294 (microblaze_legitimize_address): Generate 'UNSPEC_TEXT' for all possible
5296 (microblaze_address_insns): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
5297 (print_operand): Add 'ADDRESS_SYMBOLIC_TXT_REL' case.
5298 (print_operand_address): Add 'ADDRESS_SYMBOLIC_TXT_REL' case + handling
5299 for 'address + offset'.
5300 (microblaze_expand_prologue): Add new function prologue call for
5302 (microblaze_asm_generate_pic_addr_dif_vec): Override new target hook
5303 'TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC' to disable address diff vector
5304 table in case of TARGET_PIC_DATA_TEXT_REL.
5305 (expand_pic_symbol_ref): Add handling for 'UNSPEC_TEXT'.
5306 * gcc/config/microblaze/microblaze.md (TARGET_PIC_DATA_TEXT_REL):
5307 Add new macros 'UNSPEC_TEXT',
5308 'UNSPEC_SET_TEXT' + add rule for setting r20 in function prologue
5309 + exclude function calls from 'UNSPEC_PLT' in case of data text
5311 * doc/tm.texi.in (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
5312 new target hook for generating address diff vector tables in case of
5314 * doc/tm.texi : Regenerate.
5315 * stmt.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Append new condition
5316 'targetm.asm_out.generate_pic_addr_diff_vec' to flag_pic in case
5317 of addr diff vector generation.
5318 * target.def (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC): Add
5319 target hook definition.
5320 * targhooks.h, gcc/targhooks.c (TARGET_ASM_GENERATE_PIC_ADDR_DIFF_VEC):
5321 Add default function for generate_pic_addr_diff_vec -> flag_pic.
5322 * doc/invoke.texi (Add new pic option): Add new microblaze pic
5323 option for data text relative.
5325 2018-04-30 Richard Biener <rguenther@suse.de>
5327 * tree-chrec.h (evolution_function_is_constant_p): Remove
5329 * tree-cfg.c (tree_node_can_be_shared): Re-order checks.
5331 2018-04-30 Richard Biener <rguenther@suse.de>
5334 * dwarf2out.c (gen_producer_string): Ignore -fchecking[=].
5336 2018-04-30 Richard Biener <rguenther@suse.de>
5338 PR tree-optimization/28364
5339 PR tree-optimization/85275
5340 * tree-ssa-loop-ch.c (ch_base::copy_headers): Stop after
5341 copying first exit test.
5343 2018-04-28 Mark Wielaard <mark@klomp.org>
5345 * dwarf2out.c (dwarf2out_finish): Add .debug_addr table header for
5347 (dwarf_AT): Handle DW_AT_addr_base.
5348 (add_top_level_skeleton_die_attrs): Use dwarf_AT for DW_AT_addr_base.
5350 2018-04-28 Uros Bizjak <ubizjak@gmail.com>
5353 * config/i386/i386.md (*ashl<dwi>3_doubleword_mask): New pattern.
5354 (*ashl<dwi>3_doubleword_mask_1): Ditto.
5355 (*<shift_insn><dwi>3_doubleword_mask): Ditto.
5356 (*<shift_insn><dwi>3_doubleword_mask_1): Ditto.
5358 2018-04-28 Richard Biener <rguenther@suse.de>
5360 * tree-cfg.c (verify_gimple_phi): Take a gphi * argument.
5361 (verify_gimple_in_cfg): Rename visited_stmts to visited_throwing_stmts
5362 to reflect use. Only add interesting stmts.
5364 2018-04-27 Martin Jambor <mjambor@suse.cz>
5367 * ipa-cp.c (find_aggregate_values_for_callers_subset): Make sure
5368 the jump function allows for passing through aggregate values.
5370 2018-04-27 David Malcolm <dmalcolm@redhat.com>
5372 * input.h (in_system_header_at): Convert from macro to inline
5374 (from_macro_expansion_at): Likewise.
5375 (from_macro_definition_at): Likewise.
5377 2018-04-27 Jeff Law <law@redhat.com>
5379 * config.gcc: Mark tile* targets as deprecated/obsolete.
5381 2018-04-27 Richard Biener <rguenther@suse.de>
5383 * config/aarch64/aarch64.c: Simplify ap.__stack advance and
5386 2018-04-27 Richard Biener <rguenther@suse.de>
5388 * tree-cfg.c (verify_expr): Make dead code hit gcc_unreachable.
5390 2018-04-27 Uros Bizjak <ubizjak@gmail.com>
5392 * config/i386/i386.md (*movti_internal): Substitute Ye constraint
5393 with Yd constraint. Set "preferred_for_speed" attribute from
5394 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for alternatives
5396 (*movdi_internal): Ditto.
5397 (movti_interunit splitters): Remove
5398 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC from insn condition.
5399 (movdi_interunit splitters): Ditto.
5400 * config/i386/constraints.md (Ye): Remove.
5401 (Yd): Do not depend on TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC.
5403 2018-04-27 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5406 * config/aarch64/constraints.md (Usg): Limit to 31.
5409 2018-04-27 Jakub Jelinek <jakub@redhat.com>
5411 PR tree-optimization/85529
5412 * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Add FIRST_BB
5413 argument. Don't call get_nonzero_bits if opcode is ERROR_MARK_NODE,
5414 rhs2 def stmt's bb is dominated by first_bb and it isn't an obvious
5415 zero extension or masking of the MSB bit.
5416 (optimize_range_tests): Add FIRST_BB argument, pass it through
5417 to optimize_range_tests_var_bound.
5418 (maybe_optimize_range_tests, reassociate_bb): Adjust
5419 optimize_range_tests callers.
5421 2018-04-26 Richard Biener <rguenther@suse.de>
5422 Jakub Jelinek <jakub@redhat.com>
5424 * cgraph.h (symbol_table): Just declare debug method here.
5425 * symtab.c (symbol_table::debug): Define.
5427 2018-04-26 Eric Botcazou <ebotcazou@adacore.com>
5429 * loop-invariant.c (may_assign_reg_p): Return false for frame pointer.
5431 2018-04-26 Uros Bizjak <ubizjak@gmail.com>
5433 * config/i386/i386.md ("isa" attribute): Add x64_sse2.
5434 ("enabled" attribute): Handle x64_sse2 "isa" attribute.
5435 (*movdi_internal): Substitute Yi and Yj constraint with x
5436 and Ym and Yn constraint with y constraint. Update "isa"
5437 attribute and set "preferred_for_speed" attribute from
5438 TARGET_INTER_UNIT_MOVES_{FROM,TO}_VEC for updated alternatives.
5439 (*movsi_internal): Ditto.
5440 (*movdf_internal): Ditto.
5441 (*movsf_internal): Ditto.
5442 (*zero_extendsidi2): Ditto.
5443 * config/i386/sse.md (vec_set<mode>_0): Ditto.
5444 (sse2_loadld): Ditto.
5445 (*vec_extract<ssevecmodelower>_0): Ditto.
5446 (*vec_extractv4si_0_zext_sse4): Ditto.
5447 (vec_concatv2di): Ditto.
5448 (*vec_dup<mode>): Ditto.
5449 * config/i386/mmx.md (*mov<mode>_internal): Ditto.
5450 * config/i386/constraints.md (Yi): Remove.
5455 2018-04-26 Nathan Sidwell <nathan@acm.org>
5457 * dumpfile.c (dump_open): New.
5458 (dump_open_alternate_stream, dump_start, dump_begin): Call it.
5459 (dump_finish): Detect stdio/stderr by value not name.
5461 2018-04-26 Jonathan Wakely <jwakely@redhat.com>
5463 * doc/invoke.texi (-Wreturn-type): Document default status for C++.
5465 2018-04-26 Tom de Vries <tom@codesourcery.com>
5468 * config/nvptx/nvptx.c (verify_neutering_jumps)
5469 (verify_neutering_labels): New function
5470 (nvptx_single): Use verify_neutering_jumps and verify_neutering_labels.
5472 2018-04-26 Tom de Vries <tom@codesourcery.com>
5475 * config/nvptx/nvptx.c (needs_neutering_p): New function.
5476 (nvptx_single): Use needs_neutering_p to skip over insns that do not
5479 2018-04-26 Richard Biener <rguenther@suse.de>
5480 Tom de Vries <tom@codesourcery.com>
5483 * lto-streamer-out.c (output_function): Fixup loops if required to match
5484 discovery done in the reader.
5486 2018-04-26 Richard Biener <rguenther@suse.de>
5488 PR tree-optimization/85116
5489 * tree-ssa-loop-ch.c (do_while_loop_p): A do-while loop should
5490 have a loop exit from the single latch predecessor. Remove
5491 case of header with just condition.
5492 (ch_base::copy_headers): Exclude infinite loops from any
5494 (pass_ch::execute): Record exits.
5496 2018-04-26 Richard Biener <rguenther@suse.de>
5498 * tree-vect-data-refs.c (vect_get_data_access_cost): Get
5499 prologue cost vector and pass it to vect_get_load_cost.
5500 (vect_get_peeling_costs_all_drs): Likewise.
5501 (vect_peeling_hash_get_lowest_cost): Likewise.
5502 (vect_enhance_data_refs_alignment): Likewise.
5504 2018-04-26 Richard Biener <rguenther@suse.de>
5507 * tree-cfg.c (verify_gimple_assign_unary): Restore proper
5508 checking of integer<->pointer conversions.
5509 * omp-expand.c (expand_omp_for_static_nochunk): Avoid
5510 sign-/zero-extending pointer types.
5511 (expand_omp_for_static_chunk): Likewise.
5513 2018-03-22 Hans-Peter Nilsson <hp@axis.com>
5514 Jean Lee <xiaoyur347@gmail.com>
5516 * config/mips/mips.c (mips_asan_shadow_offset): New function.
5517 (TARGET_ASAN_SHADOW_OFFSET): Define.
5518 * config/mips/mips.h (FRAME_GROWS_DOWNWARD): Augment to also be
5519 true for -fsanitize=address.
5521 2018-04-25 Mark Wielaard <mark@klomp.org>
5523 * dwarf2out.c (file_info_cmp): Sort longer dir prefixes before
5526 2018-04-25 Jakub Jelinek <jakub@redhat.com>
5528 * config/i386/i386.md (*x86_mov<mode>cc_0_m1): Use type "alu1" rather
5529 than "alu", remove explicit "memory" and "imm_disp" attributes.
5530 (*x86_mov<mode>cc_0_m1_se, *x86_mov<mode>cc_0_m1_neg): Likewise.
5533 * simplify-rtx.c (simplify_unary_operation_1) <case SIGN_EXTEND,
5534 case ZERO_EXTEND>: Pass SUBREG_REG (op) rather than op to
5535 gen_lowpart_no_emit.
5537 2018-04-25 Sebastian Peryt <sebastian.peryt@intel.com>
5540 * config/i386/i386.c (ix86_expand_builtin): Change memory
5541 operand to XI, extend p0 to Pmode.
5542 * config/i386/i386.md: Change unspec volatile and operand
5543 1 mode to XI, change operand 0 mode to P.
5545 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5547 * config/nds32/nds32-predicates.c (nds32_can_use_bclr_p): Mask with
5548 GET_MODE_MASK before any checking.
5549 (nds32_can_use_bset_p): Likewise.
5550 (nds32_can_use_btgl_p): Likewise.
5552 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5554 * config/nds32/nds32-doubleword.md: New define_split pattern for
5555 illegal register number.
5557 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5559 * config/nds32/nds32.c (nds32_print_operand): Set op_value ealier.
5561 2018-04-25 Chung-Ju Wu <jasonwucj@gmail.com>
5563 * config/nds32/nds32.h (ASM_APP_ON): Add missing newline character.
5565 2018-04-25 Richard Biener <rguenther@suse.de>
5567 * lto-streamer.h (LTO_major_version): Bump to 8.
5569 2018-04-25 Jakub Jelinek <jakub@redhat.com>
5571 * BASE-VER: Set to 9.0.0.
5573 2018-04-24 Segher Boessenkool <segher@kernel.crashing.org>
5575 * config/rs6000/rs6000.c (init_float128_ieee): Fix spelling mistakes
5576 in __abskf2 and __powikf2.
5578 2018-04-24 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
5581 * config/aarch64/constraints.md (Usg, Usj): New constraints.
5582 * config/aarch64/iterators.md (cmode_simd): New mode attribute.
5583 * config/aarch64/aarch64.md (*aarch64_ashr_sisd_or_int_<mode>3):
5584 Use the above on operand 2. Reindent.
5585 (*aarch64_lshr_sisd_or_int_<mode>3): Likewise.
5587 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
5590 * common/config/i386/i386-common.c (ix86_handle_option): Don't
5592 * config/i386/i386.opt (mcet): Removed.
5593 * doc/install.texi: Remove -mcet documentation.
5594 * doc/invoke.texi: Likewise.
5596 2018-04-24 H.J. Lu <hongjiu.lu@intel.com>
5599 * doc/install.texi: Remove -mcet from bootstrap-cet.
5601 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5604 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Don't define
5605 __builtin_ia32_readeflags_u32 and __builtin_ia32_writeeflags_u32
5609 * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Punt if
5610 const_vector is not CONST_VECTOR or SYMBOL_REF for a constant pool
5611 containing a CONST_VECTOR.
5613 2018-04-24 Cesar Philippidis <cesar@codesourcery.com>
5615 * doc/install.texi: Update newlib dependency for nvptx.
5617 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5620 * config/i386/i386.c (ix86_expand_vector_init_one_var): Use UINTVAL
5621 instead of INTVAL when shifting x left.
5623 2018-04-24 Andreas Krebbel <krebbel@linux.ibm.com>
5625 PR tree-optimization/85478
5626 * tree-vect-loop.c (vect_analyze_loop_2): Do not call
5627 vect_grouped_store_supported for single element vectors.
5629 2018-04-24 Richard Biener <rguenther@suse.de>
5632 * config/i386/i386.c (ix86_add_stmt_cost): Restrict strided
5633 load cost increase to the case of non-constant step.
5635 2018-04-24 Jakub Jelinek <jakub@redhat.com>
5638 * reg-stack.c (move_for_stack_reg): Don't fail assertion about dead
5639 destination if any_malformed_asm.
5641 2018-04-23 Eric Botcazou <ebotcazou@adacore.com>
5644 * expr.c (store_field): In the bitfield case, if the value comes from
5645 a function call and is returned in registers by means of a PARALLEL,
5646 do not change the mode of the temporary unless BLKmode and VOIDmode.
5648 2018-04-23 Andrey Belevantsev <abel@ispras.ru>
5650 PR rtl-optimization/85423
5651 * sel-sched-ir.c (has_dependence_note_mem_dep): Only discard
5652 dependencies to debug insns when the previous insn is non-debug.
5654 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
5656 * config/arc/arc.md ("vunspec"): Delete it, unify all the unspec
5657 enums into a single definition.
5658 (fls): Fix predicates and printing.
5661 2018-04-23 Claudiu Zissulescu <claziss@synopsys.com>
5663 * config/arc/arc-protos.h (check_if_valid_sleep_operand): Remove.
5664 * config/arc/arc.c (arc_expand_builtin): Sleep accepts registers
5665 and short u6 immediate.
5666 (check_if_valid_sleep_operand): Remove.
5667 * config/arc/arc.md (Sleep): Accepts registers and u6 immediates.
5669 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5671 * config/nds32/nds32.c (nds32_compute_stack_frame): Consider
5672 flag_always_save_lp condition.
5673 * config/nds32/nds32.opt (malways-save-lp): New option.
5675 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5677 * config/nds32/nds32-protos.h (nds32_use_load_post_increment): Declare.
5678 * config/nds32/nds32.c (nds32_use_load_post_increment): New.
5679 * config/nds32/nds32.h
5680 (USE_LOAD_POST_INCREMENT, USE_LOAD_POST_DECREMENT): Define.
5681 (USE_STORE_POST_INCREMENT, USE_STORE_POST_DECREMENT): Define.
5683 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5685 * config/nds32/nds32-protos.h (nds32_ls_333_p): Remove.
5686 * config/nds32/nds32.c (nds32_ls_333_p): Remove.
5688 2018-04-22 Shiva Chen <shiva0217@gmail.com>
5689 Chung-Ju Wu <jasonwucj@gmail.com>
5691 * config/nds32/nds32-protos.h (nds32_case_vector_shorten_mode):
5693 * config/nds32/nds32.c (nds32_case_vector_shorten_mode): New function.
5694 * config/nds32/nds32.h (CASE_VECTOR_SHORTEN_MODE): Modify.
5696 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5698 * config/nds32/nds32.c (nds32_compute_stack_frame): Fix wrong value.
5700 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5702 * config/nds32/nds32-protos.h (nds32_data_alignment,
5703 nds32_local_alignment): Declare.
5704 * config/nds32/nds32.c (nds32_data_alignment, nds32_constant_alignment,
5705 nds32_local_alignment): New functions.
5706 (TARGET_CONSTANT_ALIGNMENT): Define.
5707 * config/nds32/nds32.h (DATA_ALIGNMENT, LOCAL_ALIGNMENT): Define.
5709 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5711 * config/nds32/nds32.c
5712 (TARGET_HARD_REGNO_MODE_OK): Move to the bottom of file.
5713 (TARGET_MODES_TIEABLE_P): Likewise.
5715 2018-04-22 Chung-Ju Wu <jasonwucj@gmail.com>
5717 * config/nds32/nds32.c (nds32_asm_file_start): Display optimization
5720 2018-04-22 Monk Chiang <sh.chiang04@gmail.com>
5721 Chung-Ju Wu <jasonwucj@gmail.com>
5723 * config/nds32/constants.md (unspec_volatile_element): Add enum values
5724 for unaligned access.
5725 * config/nds32/nds32-intrinsic.c: Implementation of expanding
5727 * config/nds32/nds32-intrinsic.md: Likewise.
5728 * config/nds32/nds32_intrinsic.h: Likewise.
5729 * config/nds32/nds32.h (nds32_builtins): Likewise.
5730 * config/nds32/nds32.opt (munaligned-access): New option.
5731 * config/nds32/nds32.c (nds32_asm_file_start): Display
5732 flag_unaligned_access status.
5734 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
5736 * config/riscv/elf.h (LINK_SPEC): Pass --no-relax if
5737 -mno-relax is present.
5738 * config/riscv/linux.h (LINK_SPEC): Ditto.
5740 2018-04-20 Martin Sebor <msebor@redhat.com>
5743 * gimple-fold.c (gimple_fold_builtin_strcpy): Suppress -Wrestrict
5745 (gimple_fold_builtin_stxcpy_chk): Same.
5746 * gimple-ssa-warn-restrict.c (check_bounds_or_overlap): Same.
5748 2018-04-20 Michael Meissner <meissner@linux.ibm.com>
5751 * config/rs6000/rs6000.c (init_float128_ieee): Add support to call
5752 __powikf2 when long double is IEEE 128-bit.
5754 2018-04-20 Kito Cheng <kito.cheng@gmail.com>
5756 * config/riscv/riscv.c (riscv_first_stack_step): Round up min
5757 step to make sure stack always aligned.
5759 2018-04-20 Carl Love <cel@us.ibm.com>
5762 * config/rs6000/rs6000-c.c (rs6000_gimple_fold_builtin): Add
5763 size check for arg0.
5765 2018-04-20 Nathan Sidwell <nathan@codesourcery.com>
5766 Tom de Vries <tom@codesourcery.com>
5769 * config/nvptx/nvptx.c (nvptx_emit_forking, nvptx_emit_joining):
5770 Emit insns for calls too.
5771 (nvptx_find_par): Always look for worker-level predecessor insn.
5772 (nvptx_propagate): Add is_call parm, return bool. Copy frame for
5774 (nvptx_vpropagate, nvptx_wpropagate): Adjust.
5775 (nvptx_process_pars): Propagate frames for calls.
5777 2018-04-20 H.J. Lu <hongjiu.lu@intel.com>
5780 * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET):
5782 (OPTION_MASK_ISA_IBT_UNSET): Likewise.
5783 (ix86_handle_option): Don't handle OPT_mibt.
5784 * config/i386/cet.h: Check __CET__ instead of __IBT__ and
5786 * config/i386/driver-i386.c (host_detect_local_cpu): Remove
5788 * config/i386/i386-c.c (ix86_target_macros_internal): Don't
5789 check OPTION_MASK_ISA_IBT nor flag_cf_protection.
5790 (ix86_target_macros): Define __CET__ with flag_cf_protection
5791 for -fcf-protection.
5792 * config/i386/i386.c (isa2_opts): Remove -mibt.
5793 * config/i386/i386.h (TARGET_IBT): Removed.
5794 (TARGET_IBT_P): Likewise.
5795 (ix86_valid_target_attribute_inner_p): Don't check OPT_mibt.
5796 * config/i386/i386.md (nop_endbr): Don't check TARGET_IBT.
5797 * config/i386/i386.opt (mcet): Update help message.
5800 * doc/invoke.texi: Remove -mibt. Document __CET__. Document
5801 -mcet as an alias for -mshstk.
5803 2018-04-20 Richard Biener <rguenther@suse.de>
5806 * match.pd ((X * CST) * Y -> (X * Y) * CST): Avoid exponential
5807 complexity by forcing a single use of the multiply operand.
5809 2018-04-20 Martin Jambor <mjambor@suse.cz>
5812 * ipa-cp.c (cgraph_edge_brings_value_p): Move check for self-feeding
5813 recursion dependency to only apply to non-clones.
5815 2018-04-20 Martin Jambor <mjambor@suse.cz>
5818 * ipa-cp.c (create_specialized_node): Check that clones of
5819 self-recursive edges exist during IPA-CP.
5821 2018-04-19 Toon Moene <toon@moene.org>
5823 * doc/invoke.texi: Add -floop-unroll-and-jam to options enabled
5826 2018-04-19 Jakub Jelinek <jakub@redhat.com>
5828 PR tree-optimization/85467
5829 * fold-const.c (fold_ternary_loc) <case BIT_FIELD_REF>: Use
5830 VECTOR_TYPE_P macro. If type is vector type, VIEW_CONVERT_EXPR the
5831 VECTOR_CST element to type.
5833 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
5836 * config/i386/i386.h (STACK_SAVEAREA_MODE): New.
5837 * config/i386/i386.md (builtin_setjmp_setup): Removed.
5838 (builtin_longjmp): Likewise.
5839 (save_stack_nonlocal): New pattern.
5840 (restore_stack_nonlocal): Likewise.
5842 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
5845 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
5846 Replace ASM_OUTPUT_LABEL with fprintf.
5848 2018-04-19 H.J. Lu <hongjiu.lu@intel.com>
5851 * config/i386/cet.c (file_end_indicate_exec_stack_and_cet):
5852 Check flag_cf_protection instead of TARGET_IBT and TARGET_SHSTK.
5853 * config/i386/i386-c.c (ix86_target_macros_internal): Also
5854 define __IBT__ and __SHSTK__ for -fcf-protection.
5855 * config/i386/i386.c (pass_insert_endbranch::gate): Don't check
5857 (ix86_trampoline_init): Likewise.
5858 (x86_output_mi_thunk): Likewise.
5859 (ix86_notrack_prefixed_insn_p): Likewise.
5860 (ix86_option_override_internal): Don't disallow -fcf-protection.
5861 * config/i386/i386.md (rdssp<mode>): Also enable for
5863 (incssp<mode>): Likewise.
5864 (nop_endbr): Likewise.
5865 * config/i386/i386.opt (mcet): Change help message to built-in
5869 * doc/invoke.texi: Remove -mcet, -mibt and -mshstk condition
5870 on -fcf-protection. Change -mcet, -mibt and -mshstk to only
5871 enable CET built-in functions.
5873 2018-04-19 Sebastian Peryt <sebastian.peryt@intel.com>
5875 * common/config/i386/i386-common.c
5876 (OPTION_MASK_ISA_MOVDIRI_SET, OPTION_MASK_ISA_MOVDIR64B_SET,
5877 OPTION_MASK_ISA_MOVDIRI_UNSET,
5878 OPTION_MASK_ISA_MOVDIR64B_UNSET): New defines.
5879 (ix86_handle_option): Handle -mmovdiri and -mmovdir64b.
5880 * config.gcc (movdirintrin.h): New header.
5881 * config/i386/cpuid.h (bit_MOVDIRI,
5882 bit_MOVDIR64B): New bits.
5883 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mmovdiri
5885 * config/i386/i386-builtin-types.def ((VOID, PUNSIGNED, UNSIGNED),
5886 (VOID, PVOID, PCVOID)): New function types.
5887 * config/i386/i386-builtin.def (__builtin_ia32_directstoreu_u32,
5888 __builtin_ia32_directstoreu_u64,
5889 __builtin_ia32_movdir64b): New builtins.
5890 * config/i386/i386-c.c (__MOVDIRI__, __MOVDIR64B__): New.
5891 * config/i386/i386.c (ix86_target_string): Added -mmovdir64b
5893 (ix86_valid_target_attribute_inner_p): Ditto.
5894 (ix86_expand_special_args_builtin): Added VOID_FTYPE_PUNSIGNED_UNSIGNED
5895 and VOID_FTYPE_PUNSIGNED_UNSIGNED.
5896 (ix86_expand_builtin): Expand IX86_BUILTIN_MOVDIR64B.
5897 * config/i386/i386.h (TARGET_MOVDIRI, TARGET_MOVDIRI_P,
5898 TARGET_MOVDIR64B, TARGET_MOVDIR64B_P): New.
5899 * config/i386/i386.md (UNSPECV_MOVDIRI, UNSPECV_MOVDIR64B): New.
5900 (movdiri<mode>, movdir64b_<mode>): New.
5901 * config/i386/i386.opt: Add -mmovdiri and -mmovdir64b.
5902 * config/i386/immintrin.h: Include movdirintrin.h.
5903 * config/i386/movdirintrin.h: New file.
5904 * doc/invoke.texi: Added -mmovdiri and -mmovdir64b.
5906 2018-04-19 Richard Biener <rguenther@suse.de>
5909 * cfg.c (clear_bb_flags): When loop state says we have
5910 marked irreducible regions also preserve BB_IRREDUCIBLE_LOOP.
5912 2018-04-19 Richard Biener <rguenther@suse.de>
5914 PR tree-optimization/84737
5915 * tree-vect-data-refs.c (vect_copy_ref_info): New function
5916 copying restrict info.
5917 (vect_setup_realignment): Use it.
5918 * tree-vectorizer.h (vect_copy_ref_info): Declare.
5919 * tree-vect-stmts.c (vectorizable_store): Copy ref info from
5920 the first DR to all generated stores.
5921 (vectorizable_load): Likewise for loads.
5923 2018-04-19 Jakub Jelinek <jakub@redhat.com>
5925 PR tree-optimization/85446
5926 * match.pd ((intptr_t) x eq/ne CST to x eq/ne (typeof x) cst): Require
5927 the integral and pointer types to have the same precision.
5929 * doc/install.texi: Document --disable-cet being the default and
5932 2018-04-18 Martin Liska <mliska@suse.cz>
5934 * ipa-devirt.c (odr_subtypes_equivalent_p): Fix GNU coding
5937 2018-04-18 Martin Liska <mliska@suse.cz>
5940 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
5943 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
5944 arguments if they are comparable.
5946 2018-04-18 Martin Liska <mliska@suse.cz>
5949 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
5952 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
5955 2018-04-18 H.J. Lu <hongjiu.lu@intel.com>
5958 * config/i386/i386.c (ix86_expand_split_stack_prologue): Insert
5959 ENDBR after calling __morestack.
5961 2018-04-18 David Malcolm <dmalcolm@redhat.com>
5964 * configure.ac (gcc-driver-name.h): Honor --with-gcc-major-version
5965 by using gcc_base_ver to generate a gcc_driver_version, and use
5966 it when generating GCC_DRIVER_NAME.
5967 * configure: Regenerate.
5969 2018-04-18 Jakub Jelinek <jakub@redhat.com>
5972 * config.gcc: Obsolete powerpc*-*-*spe*.
5974 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5977 * dbxout.c (dbxout_int): Perform negation in unsigned int type.
5978 (stabstr_D): Change type of unum from unsigned int to
5979 unsigned HOST_WIDE_INT. Perform negation in unsigned HOST_WIDE_INT
5982 2018-04-17 Jim Wilson <jimw@sifive.com>
5985 * config/riscv/riscv.c (riscv_compute_frame_info): Add calls to
5986 RISCV_STACK_ALIGN when using outgoing_args_size and pretend_args_size.
5987 Set arg_pointer_offset after using pretend_args_size.
5989 2018-04-17 Jakub Jelinek <jakub@redhat.com>
5991 PR rtl-optimization/85431
5992 * dse.c (record_store): Ignore zero width stores.
5995 * asan.c (handle_builtin_stack_restore): Adjust comment. Emit
5996 __asan_allocas_unpoison call and last_alloca_addr = new_sp before
5997 __builtin_stack_restore rather than after it.
5998 * builtins.c (expand_asan_emit_allocas_unpoison): Pass
5999 arg1 + (virtual_dynamic_stack_rtx - stack_pointer_rtx) as second
6000 argument instead of virtual_dynamic_stack_rtx.
6002 2018-04-17 Kelvin Nilsen <kelvin@gcc.gnu.org>
6004 * config/rs6000/rs6000-protos.h (rs6000_builtin_is_supported_p):
6006 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
6007 Add note to error message to explain internal mapping of overloaded
6008 built-in function name to non-overloaded built-in function name.
6009 * config/rs6000/rs6000.c (rs6000_builtin_is_supported_p): New
6012 2018-04-17 Michael Meissner <meissner@linux.vnet.ibm.com>
6015 * config/rs6000/rs6000.md (pack<mode>): Do not try handle a pack
6016 where the inputs overlap with the output.
6018 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6020 * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): Add
6021 (=v, v) alternative and explicit "memory" attribute.
6022 (vec_extract_lo_<mode><mask_name>): Likewise. Also add
6023 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6025 (vec_extract_lo_<mode><mask_name>): Add (=v, v) alternative and use
6026 "sselog1" type instead of "sselog".
6027 (vec_extract_hi_<mode><mask_name>): Use "sselog1" type instead of
6028 "sselog". Remove explicit "memory" attribute.
6029 (vec_extract_lo_v32hi): Add (=v, v) alternative and explicit "memory",
6030 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6032 (vec_extract_hi_v32hi): Merge all alternatives into one, use
6033 "sselog1" type instead of "sselog". Remove explicit "memory"
6035 (vec_extract_hi_v16hi): Merge each pair of alternatives into one,
6036 use "sselog1" type instead of "sselog". Remove explicit "memory"
6038 (vec_extract_lo_v64qi): Add (=v, v) alternative and explicit "memory",
6039 "type", "prefix", "prefix_extra", "length_immediate" and "mode"
6041 (vec_extract_hi_v64qi): Merge all alternatives into one, use
6042 "sselog1" type instead of "sselog". Remove explicit "memory"
6044 (vec_extract_hi_v32qi): Merge each pair of alternatives into one,
6045 use "sselog1" type instead of "sselog". Remove explicit "memory"
6049 * config/i386/i386.md (*ashlqi3_1_slp): Use alu1 type instead of alu.
6052 * rtlhooks.c (gen_lowpart_if_possible): Don't call gen_lowpart_SUBREG
6055 2018-04-17 Martin Jambor <mjambor@suse.cz>
6058 * ipa-cp.c (create_specialized_node): Call
6059 expand_all_artificial_thunks if necessary.
6061 2018-04-17 Martin Liska <mliska@suse.cz>
6064 * ipa-devirt.c (odr_types_equivalent_p): Remove trailing
6065 in message, remote space in between '_G' and '('.
6067 2018-04-17 Jakub Jelinek <jakub@redhat.com>
6070 * config/i386/sse.md (reduces<mode><mask_scalar_name>,
6071 avx512f_vmcmp<mode>3<round_saeonly_name>,
6072 avx512f_vmcmp<mode>3_mask<round_saeonly_name>,
6073 avx512f_sgetexp<mode><mask_scalar_name><round_saeonly_scalar_name>,
6074 avx512f_rndscale<mode><round_saeonly_name>,
6075 avx512dq_ranges<mode><mask_scalar_name><round_saeonly_scalar_name>,
6076 avx512f_vgetmant<mode><mask_scalar_name><round_saeonly_scalar_name>):
6077 Use %<iptr>2 instead of %2 for -masm=intel.
6078 (avx512f_vcvtss2usi<round_name>, avx512f_vcvtss2usiq<round_name>,
6079 avx512f_vcvttss2usi<round_saeonly_name>,
6080 avx512f_vcvttss2usiq<round_saeonly_name>): Use %k1 instead of %1 for
6082 (avx512f_vcvtsd2usi<round_name>, avx512f_vcvtsd2usiq<round_name>,
6083 avx512f_vcvttsd2usi<round_saeonly_name>,
6084 avx512f_vcvttsd2usiq<round_saeonly_name>, ufloatv2siv2df2<mask_name>):
6085 Use %q1 instead of %1 for -masm=intel.
6086 (avx512f_sfixupimm<mode><sd_maskz_name><round_saeonly_name>,
6087 avx512f_sfixupimm<mode>_mask<round_saeonly_name>): Use %<iptr>3 instead
6088 of %3 for -masm=intel.
6089 (sse2_shufpd_v2df_mask): Fix a typo, change %{6%} to %{%6%} for
6091 (*avx512vl_<code>v2div2qi2_store): Use %w0 instead of %0 for
6093 (*avx512vl_<code><mode>v4qi2_store): Use %k0 instead of %0 for
6095 (avx512vl_<code><mode>v4qi2_mask_store): Use a single pattern with
6096 %k0 and %1 for -masm=intel rather than two patterns, one with %0 and
6098 (*avx512vl_<code><mode>v8qi2_store): Use %q0 instead of %0 for
6100 (avx512vl_<code><mode>v8qi2_mask_store): Use a single pattern with
6101 %q0 and %1 for -masm=intel rather than two patterns, one with %0 and
6102 %g1 and one with %0 and %1.
6103 (avx512er_vmrcp28<mode><round_saeonly_name>,
6104 avx512er_vmrsqrt28<mode><round_saeonly_name>): Use %<iptr>1 instead of
6106 (avx5124fmaddps_4fmaddps_mask, avx5124fmaddps_4fmaddss_mask,
6107 avx5124fmaddps_4fnmaddps_mask, avx5124fmaddps_4fnmaddss_mask,
6108 avx5124vnniw_vp4dpwssd_mask, avx5124vnniw_vp4dpwssds_mask): Swap order
6109 of %0 and %{%4%} for -masm=intel.
6110 (avx5124fmaddps_4fmaddps_maskz, avx5124fmaddps_4fmaddss_maskz,
6111 avx5124fmaddps_4fnmaddps_maskz, avx5124fmaddps_4fnmaddss_maskz,
6112 avx5124vnniw_vp4dpwssd_maskz, avx5124vnniw_vp4dpwssds_maskz): Swap
6113 order of %0 and %{%5%}%{z%} for -masm=intel.
6115 2018-04-17 Jan Hubicka <jh@suse.cz>
6118 * ipa-devirt.c (odr_types_equivalent_p): Handle bit fields.
6120 2018-04-17 Martin Liska <mliska@suse.cz>
6123 * multiple_target.c (create_dispatcher_calls): Set apostrophes
6124 for target_clone error message. Make default implementation
6125 clone to be a local declaration.
6126 (separate_attrs): Add new argument and check for an empty
6128 (expand_target_clones): Handle it.
6129 (ipa_target_clone): Make redirection just for target_clones
6132 2018-04-16 Cesar Philippidis <cesar@codesourcery.com>
6133 Tom de Vries <tom@codesourcery.com>
6136 * omp-expand.c (expand_oacc_for): Add dummy false branch for
6137 tiled basic blocks without omp continue statements.
6139 2018-04-16 Aaron Sawdey <acsawdey@linux.ibm.com>
6142 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Mark
6143 vec_extract expression as having side effects to make sure it gets
6146 2018-04-16 H.J. Lu <hongjiu.lu@intel.com>
6149 * config/i386/i386.c (get_builtin_code_for_version): Check
6152 2018-04-16 Olga Makhotina <olga.makhotina@intel.com>
6155 * gcc/config.gcc: Support "skylake".
6156 * gcc/config/i386/i386-c.c (ix86_target_macros_internal): Handle
6158 * gcc/config/i386/i386.c (m_SKYLAKE): Define.
6159 (processor_target_table): Add "skylake".
6160 (ix86_option_override_internal): Add "skylake".
6161 (get_builtin_code_for_version): Handle PROCESSOR_SKYLAKE,
6162 PROCESSOR_CANNONLAKE.
6163 (get_builtin_code_for_version): Fix priority for
6164 PROCESSOR_ICELAKE_CLIENT, PROCESSOR_ICELAKE_SERVER,
6165 PROCESSOR_SKYLAKE-AVX512.
6166 * gcc/config/i386/i386.h (processor_costs): Define TARGET_SKYLAKE.
6167 (processor_type): Add PROCESSOR_SKYLAKE.
6169 2018-04-16 Paolo Carlini <paolo.carlini@oracle.com>
6170 Jason Merrill <jason@redhat.com>
6173 * convert.c (convert_to_integer_1): Use direct recursion for
6174 enumeral types and types with a precision less than the number
6175 of bits in their mode.
6177 2018-04-16 Julia Koval <julia.koval@intel.com>
6180 * config/i386/x86-tune.def (X86_TUNE_SSE_UNALIGNED_LOAD_OPTIMAL,
6181 X86_TUNE_SSE_UNALIGNED_STORE_OPTIMAL): Add m_SKYLAKE_AVX512
6183 2018-04-14 Segher Boessenkool <segher@kernel.crashing.org>
6186 * config/rs6000/rs6000.opt (mdirect-move): Make deprecated.
6187 * doc/invoke.texi (RS/6000 and PowerPC Options): Remove -mdirect-move
6188 and -mno-direct-move.
6190 2018-04-13 Paul A. Clarke <pc@us.ibm.com>
6193 * config/rs6000/emmintrin.h (_mm_slli_epi{16,32,64}):
6194 Ensure that vec_splat_s32 is only called with 0 <= shift < 16.
6195 Ensure negative shifts result in {0}.
6197 2018-04-13 Vladimir Makarov <vmakarov@redhat.com>
6199 PR rtl-optimization/79916
6200 * config/rs6000/rs6000.c (rs6000_emit_move): Use assigned hard
6201 regs (if any) to define how to gnerate SD moves when LRA is in
6204 2018-04-13 Jakub Jelinek <jakub@redhat.com>
6206 PR rtl-optimization/85393
6207 * except.h (expand_dw2_landing_pad_for_region): Remove declaration.
6208 * except.c (expand_dw2_landing_pad_for_region): Make static.
6209 * bb-reorder.c (fix_up_crossing_landing_pad): In new_bb emit just
6210 a label and unconditional jump to old_bb, rather than
6211 expand_dw2_landing_pad_for_region insn(s) and jump to single_succ
6214 PR rtl-optimization/85376
6215 * simplify-rtx.c (simplify_const_unary_operation): For CLZ and CTZ and
6216 zero op0, if C?Z_DEFINED_VALUE_AT_ZERO is false, return NULL_RTX
6217 instead of a specific value.
6219 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
6220 Bin Cheng <bin.cheng@arm.com>
6222 PR tree-optimization/82965
6223 PR tree-optimization/83991
6224 * cfgloopanal.c (expected_loop_iterations_unbounded): Add
6225 by_profile_only parameter.
6226 * cfgloopmanip.c (scale_loop_profile): Further scale loop's profile
6227 information if the loop was predicted to iterate too many times.
6228 * cfgloop.h (expected_loop_iterations_unbounded): Update prototype
6230 2018-04-13 Jan Hubicka <hubicka@ucw.cz>
6233 * config/i386/i386.c (ix86_can_inline_p): Allow safe transitions for
6236 2018-04-13 Martin Liska <mliska@suse.cz>
6237 Jakub Jelinek <jakub@redhat.com>
6240 * expr.h (enum block_op_methods): Add BLOCK_OP_NO_LIBCALL_RET.
6241 * expr.c (emit_block_move_hints): Handle BLOCK_OP_NO_LIBCALL_RET.
6242 * builtins.c (expand_builtin_memory_copy_args): Use
6243 BLOCK_OP_NO_LIBCALL_RET method for mempcpy with non-ignored target,
6244 handle dest_addr == pc_rtx.
6246 2018-04-12 Segher Boessenkool <segher@kernel.crashing.org>
6249 * config/rs6000/rs6000.md (fix_trunc<mode>si2): Use legacy code if
6250 asked to not generate direct moves.
6251 (fix_trunc<mode>si2_stfiwx): Similar.
6252 (fix_trunc<mode>si2_internal): Similar.
6254 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6257 * var-tracking.c (add_stores): Handle STRICT_LOW_PART SET_DEST.
6258 * cselib.c (cselib_record_sets): For STRICT_LOW_PART dest,
6259 lookup if dest in some wider mode is known to be const0_rtx and
6260 if so, record permanent equivalence for it to be ZERO_EXTEND of
6261 the narrower mode destination.
6263 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
6265 * lto-streamer-out.c (output_function): Revert 259346.
6266 * omp-expand.c (expand_oacc_for): Likewise.
6268 2018-04-12 Alexander Monakov <amonakov@ispras.ru>
6270 PR rtl-optimization/85354
6271 * sel-sched-ir.c (sel_init_pipelining): Move cfg_cleanup call...
6272 * sel-sched.c (sel_global_init): ... here.
6274 2018-04-12 Eric Botcazou <ebotcazou@adacore.com>
6277 * lto-wrapper.c (debug_objcopy): Open the files in binary mode.
6278 * dwarf2out.c (dwarf2out_early_finish): Do not generate assembly in LTO
6279 mode for PE-COFF targets.
6280 * config/i386/i386-protos.h (i386_pe_asm_lto_start): Declare.
6281 (i386_pe_asm_lto_end): Likewise.
6282 * config/i386/cygming.h (TARGET_ASM_LTO_START): Define.
6283 (TARGET_ASM_LTO_END): Likewise.
6284 * config/i386/winnt.c (saved_debug_info_level): New static variable.
6285 (i386_pe_asm_lto_start): New function.
6286 (i386_pe_asm_lto_end): Likewise.
6288 2018-04-12 Cesar Philippidis <cesar@codesourcery.com>
6289 Richard Biener <rguenther@suse.de>
6292 * lto-streamer-out.c (output_function): Fix CFG loop state before
6294 * omp-expand.c (expand_oacc_for): Handle calls to internal
6295 functions like regular functions.
6297 2018-04-12 Richard Biener <rguenther@suse.de>
6300 * dwarf2out.c (init_sections_and_labels): Use debug_line_section[_label]
6301 for the early LTO debug to properly generate references to it
6302 during DIE emission. Do not re-use that for the skeleton for
6304 (dwarf2out_early_finish): Likewise.
6306 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6309 * config/i386/sse.md
6310 (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name> split,
6311 <mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name> split,
6312 vec_extract_lo_<mode><mask_name> split, vec_extract_lo_v32hi,
6313 vec_extract_lo_v64qi): For non-AVX512VL if input is xmm16+ reg
6314 and output is a reg, avoid creating invalid lowpart subreg, but
6315 instead split into a 512-bit move. Don't split if not AVX512VL,
6316 input is xmm16+ reg and output is a mem.
6317 (vec_extract_lo_<mode><mask_name>, vec_extract_lo_v32hi,
6318 vec_extract_lo_v64qi): Don't require split if not AVX512VL, input is
6319 xmm16+ reg and output is a mem.
6321 2018-04-12 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6323 * config/s390/s390.c (s390_output_indirect_thunk_function): Check
6324 also for flag_dwarf2_cfi_asm.
6326 2018-04-12 Jakub Jelinek <jakub@redhat.com>
6328 PR rtl-optimization/85342
6329 * regcprop.c (copyprop_hardreg_forward_1): Remove replaced array, use
6330 a bool scalar var inside of the loop instead. Don't try to update
6331 recog_data.operand after failed apply_change_group.
6333 2018-04-12 Tom de Vries <tom@codesourcery.com>
6336 * config/nvptx/nvptx.c (flexible_array_member_type_p): New function.
6337 (nvptx_assemble_decl_begin): Add undefined param. Declare undefined
6338 array with flexible array member as array without given dimension.
6339 (nvptx_assemble_undefined_decl): Set nvptx_assemble_decl_begin call
6340 argument for undefined param to true.
6342 2018-04-11 Aaron Sawdey <acsawdey@linux.ibm.com>
6345 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
6346 -mcall- and -mtraceback=. Remove options -mabi=spe and -mabi=no-spe
6347 from PowerPC section.
6348 * config/rs6000/sysv4.opt (mcall-): Improve help text.
6349 * config/rs6000/rs6000.opt (mblock-compare-inline-limit=): Trim
6350 help text that is too long.
6351 * config/rs6000/rs6000.opt (mblock-compare-inline-loop-limit=): Trim
6352 help text that is too long.
6353 * config/rs6000/rs6000.opt (mstring-compare-inline-limit=): Trim
6354 help text that is too long.
6356 2018-04-11 Uros Bizjak <ubizjak@gmail.com>
6358 * config/alpha/alpha.md (stack_probe_internal): Rename
6359 from "probe_stack". Update all callers.
6361 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6363 PR rtl-optimization/84566
6364 * sched-deps.c (sched_analyze_insn): Check deps->readonly when invoking
6365 sched_macro_fuse_insns.
6367 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6370 * sched-rgn.c (add_branch_dependences): Move sel_sched_p check here...
6371 (compute_block_dependences): ... from here.
6373 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6375 PR tree-optimization/85331
6376 * vec-perm-indices.h (vec_perm_indices::clamp): Change input type
6377 from int to HOST_WIDE_INT.
6379 2018-04-11 Martin Jambor <mjambor@suse.cz>
6382 * ipa-cp.c (propagate_vals_across_pass_through): Expand comment.
6383 (cgraph_edge_brings_value_p): New parameter dest_val, check if it is
6384 not the same as the source val.
6385 (cgraph_edge_brings_value_p): New parameter.
6386 (gather_edges_for_value): Pass destination value to
6387 cgraph_edge_brings_value_p.
6388 (perhaps_add_new_callers): Likewise.
6389 (get_info_about_necessary_edges): Likewise and exclude values brought
6390 only by self-recursive edges.
6391 (create_specialized_node): Redirect only clones of self-calling edges.
6392 (+self_recursive_pass_through_p): New function.
6393 (find_more_scalar_values_for_callers_subset): Use it.
6394 (find_aggregate_values_for_callers_subset): Likewise.
6395 (known_aggs_to_agg_replacement_list): Removed.
6396 (decide_whether_version_node): Re-calculate known constants for all
6397 remaining context clones.
6399 2018-04-11 Richard Biener <rguenther@suse.de>
6402 * dwarf2out.c (dwarf2out_finish): Remove DW_AT_stmt_list attribute
6403 from early DWARF output.
6404 (dwarf2out_early_finish): Output line info unconditionally into
6405 early DWARF and add reference to it.
6407 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6410 * config/i386/sse.md (iptr): Add V16SFmode and V8DFmode cases.
6411 (<avx512>_vec_dup<mode><mask_name>): Use a single pattern for modes
6412 other than V2DFmode using iptr mode attribute.
6413 (<avx512>_vec_dup<mode><mask_name>): Use iptr mode attribute.
6415 2018-04-11 Alexander Monakov <amonakov@ispras.ru>
6417 PR rtl-optimization/84659
6418 * sel-sched-ir.c (sel_init_pipelining): Invoke cleanup_cfg.
6420 2018-04-11 Jakub Jelinek <jakub@redhat.com>
6423 * dwarf2out.c (skip_loc_list_entry): Don't call size_of_locs if
6425 (output_loc_list): Pass address of a dummy size variable even in the
6426 locview handling loop.
6427 (index_location_lists): Add comment on why skip_loc_list_entry can't
6430 2018-04-11 Thomas Preud'homme <thomas.preudhomme@arm.com>
6433 * config/arm/arm-builtins.c (arm_expand_builtin): Force input operand
6436 2018-04-10 Aaron Sawdey <acsawdey@linux.ibm.com>
6439 * doc/invoke.texi (RS/6000 and PowerPC Options): Document options
6440 -mblock-compare-inline-limit, -mblock-compare-inline-loop-limit,
6441 and -mstring-compare-inline-limit.
6443 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6446 * gcc/config/rs6000/rs6000.md (allocate_stack): Put the residual size
6447 for stack clash protection in a register whenever we need it to be in
6450 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6452 * common/config/rs6000/rs6000-common.c (rs6000_option_init_struct):
6453 Enable -fasynchronous-unwind-tables by default if OBJECT_FORMAT_ELF.
6455 2018-04-10 Segher Boessenkool <segher@kernel.crashing.org>
6458 * config/rs6000/rs6000.opt (mtraceback=): Show the allowed values in
6460 (mlong-double-): Ditto.
6461 * config/rs6000/sysv4.opt (msdata=): Ditto.
6462 (mtls-size=): Ditto.
6464 2018-04-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
6466 * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Remove
6467 erroneous entries for
6468 "vector int vec_ldl (int, long int *)", and
6469 "vector unsigned int vec_ldl (int, unsigned long int *)".
6470 Add comments and entries for
6471 "vector bool char vec_ldl (int, bool char *)",
6472 "vector bool short vec_ldl (int, bool short *)",
6473 "vector bool int vec_ldl (int, bool int *)",
6474 "vector bool long long vec_ldl (int, bool long long *)",
6475 "vector pixel vec_ldl (int, pixel *)",
6476 "vector long long vec_ldl (int, long long *)",
6477 "vector unsigned long long vec_ldl (int, unsigned long long *)".
6478 * config/rs6000/rs6000.c (rs6000_init_builtins): Initialize new
6479 type tree bool_long_long_type_node and correct definition of
6480 bool_V2DI_type_node to make reference to this new type tree.
6481 (rs6000_mangle_type): Replace erroneous reference to
6482 bool_long_type_node with bool_long_long_type_node.
6483 * config/rs6000/rs6000.h (enum rs6000_builtin_type_index): Add
6484 comments to emphasize sign distinctions for char and int types and
6485 replace RS6000_BTI_bool_long constant with
6486 RS6000_BTI_bool_long_long constant. Also add comment to restrict
6487 use of RS6000_BTI_pixel.
6488 (bool_long_type_node): Remove this macro definition.
6489 (bool_long_long_type_node): New macro definition
6491 2018-04-10 Jakub Jelinek <jakub@redhat.com>
6493 PR rtl-optimization/85300
6494 * combine.c (subst): Handle subst of CONST_SCALAR_INT_P new_rtx also
6495 into FLOAT and UNSIGNED_FLOAT like ZERO_EXTEND, return a CLOBBER if
6496 simplify_unary_operation fails.
6498 2018-04-10 Martin Liska <mliska@suse.cz>
6500 * gdbhooks.py: Add pretty-printers for varpool_node, symtab_node,
6501 cgraph_edge and ipa_ref.
6503 2018-04-10 Jakub Jelinek <jakub@redhat.com>
6507 * config/i386/sse.md
6508 (<extract_type>_vinsert<shuffletype><extract_suf>_mask): Fix
6509 computation of the VEC_MERGE selector from mask.
6510 (<extract_type>_vinsert<shuffletype><extract_suf>_1<mask_name>):
6511 Fix decoding of the VEC_MERGE selector into mask.
6513 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
6515 PR tree-optimization/85286
6516 * tree-vect-data-refs.c (vect_get_smallest_scalar_type):
6518 2018-04-10 Richard Sandiford <richard.sandiford@linaro.org>
6520 * final.c (final_1): Set insn_last_address as well as
6521 insn_current_address.
6523 2018-04-10 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
6526 * explow.c (emit_stack_probe): Call validize_mem on memory location
6527 before passing it to gen_probe_stack. Create address operand and
6528 legitimize it for the probe_stack_address case.
6530 2018-04-09 Jan Hubicka <jh@suse.cz>
6533 * ipa-devirt.c (rebuild_type_inheritance-hash): New.
6534 * ipa-utils.h (rebuild_type_inheritance-hash): Declare.
6535 * tree.c (free_lang_data_in_type): Fix handling of binfos;
6537 (free_lang_data): Rebuild type inheritance graph.
6539 2018-04-09 Martin Sebor <msebor@redhat.com>
6541 * invoke.texi (-finline-small-functions): Mention other optimization
6543 (-findirect-inlining, -fpartial-inlining): Same.
6544 (-finline-functions-called-once): Same.
6545 (-freorder-blocks-and-partition): Same.
6547 2018-04-09 Jan Hubicka <jh@suse.cz>
6550 * cfgcleanup.c (try_forward_edges): Do not give up on crossing
6551 jumps; choose last target that matches the criteria (i.e.
6552 no partition changes for non-crossing jumps).
6553 * cfgrtl.c (cfg_layout_redirect_edge_and_branch): Add basic
6554 support for redirecting crossing jumps to non-crossing.
6556 2018-04-09 Alexey Brodkin <abrodkin@synopsys.com>
6558 * config/arc/arc.c (arc_expand_prologue): Set stack usage info
6559 also for naked functions.
6561 2018-04-09 Claudiu Zissulescu <claziss@synopsys.com>
6563 * config/arc/arc.md (add_shift): New pattern.
6564 (add_shift2): Likewise.
6565 (sub_shift): Likewise.
6566 (sub_shift_cmp0_noout): Likewise.
6567 (compare_si_ashiftsi): Likewise.
6568 (xbfu_cmp0_noout): New combine pattern.
6569 (xbfu_cmp0"): Likewise.
6570 (movsi_set_cc_insn): Place the predicable variant first.
6571 (commutative_binary_cmp0_noout): Remove clobber.
6572 (commutative_binary_cmp0): New pattern.
6573 (noncommutative_binary_cmp0): Likewise.
6574 (noncommutative_binary_cmp0_noout): Likewise.
6575 (noncommutative_binary_comparison_result_used): Removed.
6576 (rsub_cmp0): New pattern.
6577 (rsub_cmp0_noout): Likewise.
6578 (extzvsi): Changed, keep only meaningful variants.
6579 (SQH, SEZ): New iterators.
6580 (SQH_postfix): New mode attribute.
6581 (SEZ_prefix): New code attribute.
6582 (<SEZ_prefix>xt<SQH_postfix>_cmp0_noout): New instruction pattern.
6583 (<SEZ_prefix>xt<SQH_postfix>_cmp0): Likewise.
6584 * config/arc/predicates.md (cc_set_register): Use CC_REG instead
6586 (noncommutative_operator): Check the availability of barrel
6589 2018-04-09 Richard Biener <rguenther@suse.de>
6591 PR tree-optimization/85284
6592 * tree-ssa-loop-niter.c (number_of_iterations_exit_assumptions):
6593 Only use the niter constraining form of simple_iv when the exit
6596 2018-04-09 Tom de Vries <tom@codesourcery.com>
6599 * config/nvptx/nvptx.md (define_c_enum "unspecv"): Add UNSPECV_MEMBAR.
6600 (define_expand "*memory_barrier"): New define_expand.
6601 (define_insn "memory_barrier"): New insn.
6603 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6605 PR rtl-optimization/80463
6606 PR rtl-optimization/83972
6607 PR rtl-optimization/83480
6609 * sel-sched-ir.c (has_dependence_note_mem_dep): Take into account the
6610 correct producer for the insn.
6611 (tidy_control_flow): Fixup seqnos in case of debug insns.
6613 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6615 PR rtl-optimization/83913
6617 * sel-sched-ir.c (merge_expr_data): Choose the middle between two
6618 different sched-times when merging exprs.
6620 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6622 PR rtl-optimization/83962
6624 * sel-sched-ir.c (tidy_control_flow): Correct the order in which we call
6625 tidy_fallthru_edge and tidy_control_flow.
6627 2018-04-09 Andrey Belevantsev <abel@ispras.ru>
6629 PR rtl-optimization/83530
6631 * sel-sched.c (force_next_insn): New global variable.
6632 (remove_insn_for_debug): When force_next_insn is true, also leave only
6633 next insn in the ready list.
6634 (sel_sched_region): When the region wasn't scheduled, make another pass
6635 over it with force_next_insn set to 1.
6637 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
6639 * config.gcc (nds32le-*-*, nds32be-*-*): Add nds32/nds32_intrinsic.h
6641 * config/nds32/constants.md (unspec_volatile_element): Add enum values
6642 for interrupt control.
6643 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
6644 functions for interrupt control.
6645 * config/nds32/nds32-intrinsic.md: Likewise.
6646 * config/nds32/nds32_intrinsic.h: Likewise.
6647 * config/nds32/nds32.h (nds32_builtins): Likewise.
6649 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
6651 * config/nds32/nds32.c (nds32_init_machine_status,
6652 nds32_legitimate_index_p, nds32_legitimate_address_p): Consider
6653 strict_aligned_p field.
6654 (nds32_expand_to_rtl_hook): New function.
6655 (TARGET_EXPAND_TO_RTL_HOOK): Define.
6656 * config/nds32/nds32.h (machine_function): Add strict_aligned_p field.
6658 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6659 Chung-Ju Wu <jasonwucj@gmail.com>
6661 * config.gcc (nds32*-*-*): Check that n7 is valid to --with-cpu.
6662 * config/nds32/nds32-n7.md: New file.
6663 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N7.
6664 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n7
6666 * config/nds32/nds32-protos.h: More declarations for n7 pipeline.
6667 * config/nds32/nds32.md (pipeline_model): Add n7.
6668 * config/nds32/nds32.opt (mcpu): Support n7 pipeline cpus.
6669 * config/nds32/pipelines.md: Include n7 settings.
6671 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6672 Chung-Ju Wu <jasonwucj@gmail.com>
6674 * config.gcc (nds32*-*-*): Check that e8 is valid to --with-cpu.
6675 * config/nds32/nds32-e8.md: New file.
6676 * config/nds32/nds32-opts.h (nds32-cpu_type): Add CPU_E8.
6677 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for e8
6679 * config/nds32/nds32-protos.h: More declarations for e8 pipeline.
6680 * config/nds32/nds32.md (pipeline_model): Add e8.
6681 * config/nds32/nds32.opt (mcpu): Support e8 pipeline cpus.
6682 * config/nds32/pipelines.md: Include e8 settings.
6684 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6685 Chung-Ju Wu <jasonwucj@gmail.com>
6687 * config.gcc (nds32*-*-*): Check that n6/n8/s8 are valid to --with-cpu.
6688 * config/nds32/nds32-n8.md: New file.
6689 * config/nds32/nds32-opts.h (nds32_cpu_type): Add CPU_N6 and CPU_N8.
6690 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n8
6692 * config/nds32/nds32-protos.h: More declarations for n8 pipeline.
6693 * config/nds32/nds32-utils.c: More implementations for n8 pipeline.
6694 * config/nds32/nds32.md (pipeline_model): Add n8.
6695 * config/nds32/nds32.opt (mcpu): Support n8 pipeline cpus.
6696 * config/nds32/pipelines.md: Include n8 settings.
6698 2018-04-08 Kito Cheng <kito.cheng@gmail.com>
6699 Chung-Ju Wu <jasonwucj@gmail.com>
6701 * config.gcc (nds32*): Add nds32-utils.o into extra_objs.
6702 * config/nds32/nds32-n9-2r1w.md: New file.
6703 * config/nds32/nds32-n9-3r2w.md: New file.
6704 * config/nds32/nds32-opts.h (nds32_cpu_type, nds32_mul_type,
6705 nds32_register_ports): New or modify for cpu n9.
6706 * config/nds32/nds32-pipelines-auxiliary.c: Implementation for n9
6708 * config/nds32/nds32-protos.h: More declarations for n9 pipeline.
6709 * config/nds32/nds32-utils.c: New file.
6710 * config/nds32/nds32.h (TARGET_PIPELINE_N9, TARGET_PIPELINE_SIMPLE,
6711 TARGET_MUL_SLOW): Define.
6712 * config/nds32/nds32.md (pipeline_model): New attribute.
6713 * config/nds32/nds32.opt (mcpu, mconfig-mul, mconfig-register-ports):
6714 New options that support cpu n9.
6715 * config/nds32/pipelines.md: Include n9 settings.
6716 * config/nds32/t-nds32 (nds32-utils.o): Add dependency.
6718 2018-04-08 Chung-Ju Wu <jasonwucj@gmail.com>
6720 * config/nds32/nds32-md-auxiliary.c (output_cond_branch): Output align
6721 information if necessary.
6722 (output_cond_branch_compare_zero): Likewise.
6723 * config/nds32/nds32.c (nds32_adjust_insn_length): Consider align case.
6724 (nds32_target_alignment): Refine for alignment.
6725 * config/nds32/nds32.h (NDS32_ALIGN_P): Define.
6726 (FUNCTION_BOUNDARY): Modify.
6727 * config/nds32/nds32.md (call_internal, call_value_internal): Consider
6729 * config/nds32/nds32.opt (malways-align, malign-functions): New.
6731 2018-04-08 Monk Chiang <sh.chiang04@gmail.com>
6733 * config/nds32/constants.md (unspec_volatile_element): Add values for
6734 TLB operation and data prefetch.
6735 * config/nds32/nds32-intrinsic.c: Implementation of intrinsic
6736 functions for TLB operation and data prefetch.
6737 * config/nds32/nds32-intrinsic.md: Likewise.
6738 * config/nds32/nds32_intrinsic.h: Likewise.
6739 * config/nds32/nds32.c (nds32_dpref_names): Likewise.
6740 (nds32_print_operand): Likewise.
6741 * config/nds32/nds32.h (nds32_builtins): Likewise.
6743 2018-04-07 Thomas Koenig <tkoenig@gcc.gnu.org>
6744 Andrew Pinski <pinsika@gcc.gnu.org>
6747 * match.pd: Use constant_boolean_node of correct type instead of
6748 boolean_true_node or boolean_false_node for simplifying
6749 pointer comparisons to zero.
6751 2018-04-07 Jakub Jelinek <jakub@redhat.com>
6753 PR tree-optimization/80021
6754 * tree.c (verify_type_variant): Make error call in verify_variant_match
6755 translatable and remove final full stop.
6757 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6759 * config/nds32/constants.md (unspec_volatile_element): Add
6760 UNSPEC_VOLATILE_EH_RETURN.
6761 * config/nds32/nds32-md-auxiliary.c (nds32_output_stack_push,
6762 nds32_output_stack_pop): Support dwarf exception handling process.
6763 * config/nds32/nds32-protos.h (nds32_dynamic_chain_address): Declare.
6764 * config/nds32/nds32.c (nds32_init_machine_status): Support dwarf
6765 exception handling process.
6766 (nds32_compute_stack_frame): Likewise.
6767 (nds32_return_addr_rtx): Likewise.
6768 (nds32_initial_elimination_offset): Likewise.
6769 (nds32_expand_prologue): Likewise.
6770 (nds32_expand_epilogue): Likewise.
6771 (nds32_dynamic_chain_address): New function.
6772 * config/nds32/nds32.h (machine_function): Add fields for dwarf
6774 (DYNAMIC_CHAIN_ADDRESS): Define.
6775 (EH_RETURN_DATA_REGNO): Define.
6776 (EH_RETURN_STACKADJ_RTX): Define.
6777 * config/nds32/nds32.md (eh_return, nds32_eh_return): Implement
6778 patterns for dwarf exception handling.
6780 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6782 * config/nds32/nds32.h: Clean up obsolete macros.
6784 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6786 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6787 Add enum values for particular instructions.
6788 * config/nds32/nds32-intrinsic.c: Implementation of expanding
6789 particular intrinsic functions.
6790 * config/nds32/nds32-intrinsic.md: Likewise.
6791 * config/nds32/nds32_intrinsic.h: Likewise.
6792 * config/nds32/nds32.h (nds32_builtins): Likewise.
6793 * config/nds32/nds32.md (type): Add pbsad and pbsada.
6794 (btst, ave): New patterns for particular instructions.
6796 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6798 * config/nds32/constants.md (unspec_element, unspec_volatile_element):
6799 Add enum values for atomic load/store and memory sync.
6800 * config/nds32/nds32-intrinsic.c: Implementation for atomic load/store
6802 * config/nds32/nds32-intrinsic.md: Likewise.
6803 * config/nds32/nds32_intrinsic.h: Likewise.
6804 * config/nds32/nds32.h (nds32_builtins): Likewise.
6806 2018-04-07 Jakub Jelinek <jakub@redhat.com>
6808 PR tree-optimization/85257
6809 * fold-const.c (native_encode_vector): If not all elts could fit
6810 and off is -1, return 0 rather than offset.
6811 * tree-ssa-sccvn.c (vn_reference_lookup_3): Pass
6812 (offseti - offset2) / BITS_PER_UNIT as 4th argument to
6813 native_encode_expr. Verify len * BITS_PER_UNIT >= maxsizei. Don't
6814 adjust buffer in native_interpret_expr call.
6816 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6818 * config/nds32/constants.md (unspec_volatile_element): Add cache
6819 control enum values.
6820 * config/nds32/nds32-intrinsic.c: Add cache control expand functions.
6821 * config/nds32/nds32-intrinsic.md: Add cache control patterns.
6822 * config/nds32/nds32.c (nds32_cctl_names): New.
6823 (nds32_print_operand): Handle cache control register names.
6824 * config/nds32/nds32.h (nds32_builtins): New enum values.
6825 * config/nds32/nds32_intrinsic.h: Add cache control enum types and
6827 * config/nds32/nds32.md (type): Add mmu.
6828 * config/nds32/pipelines.md (simple_insn): Add mmu.
6830 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6832 * config/nds32/nds32.md (type): Remove call.
6833 * config/nds32/pipelines.md (simple_insn): Likewise.
6835 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6837 * config/nds32/constants.md (unspec_volatile_element): Add
6838 UNSPEC_VOLATILE_FMFCSR, UNSPEC_VOLATILE_FMTCSR and
6839 UNSPEC_VOLATILE_FMFCFG.
6840 * config/nds32/nds32-intrinsic.c (bdesc_noarg): New builtin
6841 description for fmfcfg and fmfcsr.
6842 (bdesc_1arg): Add fmtcsr.
6843 (bdesc_2arg): Add fcpynss, fcpyss, fcpynsd and fcpysd.
6844 (nds32_expand_builtin_impl): Deal with FPU intrinsic functions.
6845 * config/nds32/nds32-intrinsic.md (unspec_fcpynsd, unspec_fcpysd,
6846 unspec_fcpynss, unspec_fcpysd, unspec_fcpyss, unspec_fmfcsr,
6847 unspec_fmfcfg): New patterns.
6848 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FMFCFG,
6849 NDS32_BUILTIN_FMFCSR, NDS32_BUILTIN_FMTCSR, NDS32_BUILTIN_FCPYNSS,
6850 NDS32_BUILTIN_FCPYSS,NDS32_BUILTIN_FCPYNSD and NDS32_BUILTIN_FCPYSD.
6851 * config/nds32/nds32_intrinsic.h (__nds32__fcpynsd, __nds32__fcpynss,
6852 __nds32__fcpysd, __nds32__fcpyss, __nds32__fmfcsr, __nds32__fmtcsr,
6853 __nds32__fmfcfg): Define.
6855 2018-04-07 Monk Chiang <sh.chiang04@gmail.com>
6857 * config/nds32/nds32.c (nds32_intrinsic_register_names): Add more
6858 intrinsic register names.
6859 * config/nds32/nds32_intrinsic.h (nds32_intrinsic_registers): Add more
6860 intrinsic register enum values and macros.
6862 2018-04-07 Chung-Ju Wu <jasonwucj@gmail.com>
6864 * config/nds32/nds32.c (nds32_legitimate_index_p): Modify condition
6865 for load/store addressing form.
6866 (nds32_print_operand_address): Likewise.
6868 2018-04-06 Eric Botcazou <ebotcazou@adacore.com>
6871 * config/sparc/sparc.c (sparc_expand_move): Deal with symbolic operands
6872 based on LABEL_REF. Remove useless assertion.
6873 (pic_address_needs_scratch): Fix formatting.
6874 (sparc_legitimize_pic_address): Minor tweaks.
6875 (sparc_delegitimize_address): Adjust assertion accordingly.
6876 * config/sparc/sparc.md (movsi_pic_label_ref): Change label_ref_operand
6877 into symbolic_operand.
6878 (movsi_high_pic_label_ref): Likewise.
6879 (movsi_lo_sum_pic_label_ref): Likewise.
6880 (movdi_pic_label_ref): Likewise.
6881 (movdi_high_pic_label_ref): Likewise.
6882 (movdi_lo_sum_pic_label_ref): Likewise.
6884 2018-04-06 Amaan Cheval <amaan.cheval@gmail.com>
6886 * config.gcc (x86_64-*-rtems*): Add rtems.h to tm_file for
6887 custom LIB_SPEC setup.
6889 2018-04-06 Ruslan Bukin <br@bsdpad.com>
6890 Kito Cheng <kito.cheng@gmail.com>
6892 * config.gcc (riscv*-*-freebsd*): Add RISC-V FreeBSD support.
6893 * config/riscv/freebsd.h: New.
6895 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6897 * config/nds32/nds32.c (nds32_adjust_insn_length): Refine.
6898 * config/nds32/nds32.h (ADJUST_INSN_LENGTH): Change the location in
6901 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6902 Kito Cheng <kito.cheng@gmail.com>
6904 * config/nds32/nds32-md-auxiliary.c (nds32_output_return,
6905 nds32_output_call, nds32_symbol_binds_local_p): New functions.
6906 * config/nds32/nds32-protos.h (nds32_output_call,
6907 nds32_output_return): Declare.
6908 * config/nds32/nds32.md: Refine all the call and return patterns.
6910 2018-04-06 Jakub Jelinek <jakub@redhat.com>
6913 * dwarf2out.c (rtl_for_decl_init): For STRING_CST initializer only
6914 build CONST_STRING if TYPE_MAX_VALUE is non-NULL and is INTEGER_CST.
6916 PR rtl-optimization/84872
6917 * cfgloopmanip.c (create_preheader): Use make_forwarder_block even if
6918 nentry == 1 when CP_FALLTHRU_PREHEADERS and single_entry is
6921 2018-04-06 Tamar Christina <tamar.christina@arm.com>
6923 * expr.c (copy_blkmode_to_reg): Revert 254862.
6924 * doc/sourcebuild.texi (word_mode_no_slow_unalign): Likewise.
6926 2018-04-06 Richard Biener <rguenther@suse.de>
6929 * tree-dfa.c (get_ref_base_and_extent): Reset seen_variable_array_ref
6930 after seeing a component reference with an adjacent field. Treat
6931 refs to arrays at struct end of external decls similar to
6932 refs to unconstrained commons.
6934 2018-04-06 Jakub Jelinek <jakub@redhat.com>
6937 * fold-const.c (twoval_comparison_p): Remove SAVE_P argument and don't
6938 look through SAVE_EXPRs with non-side-effects argument. Adjust
6940 (fold_comparison): Adjust twoval_comparison_p caller, don't handle
6943 2018-04-06 Richard Biener <rguenther@suse.de>
6946 * alias.c (find_base_term): New wrapper around find_base_term
6947 unwinding CSELIB_VAL_PTR changes.
6948 (find_base_term): Do not restore CSELIB_VAL_PTR during the
6951 2018-04-06 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
6953 * config/s390/s390.c (s390_z10_optimize_cmp): Expand dedicated NOP
6955 * config/s390/s390.md (UNSPECV_NOP_LR_0, UNSPECV_NOP_LR_1): New
6956 constant definitions.
6957 ("nop"): lr 0,0 -> nopr r0
6958 ("nop_lr0", "nop_lr1"): New insn definitions.
6960 2018-04-06 Chung-Ju Wu <jasonwucj@gmail.com>
6962 * config/nds32/nds32.md (*stack_push, *stack_pop): Use
6963 NDS32_V3PUSH_AVAILABLE_P macro.
6965 2018-04-06 Monk Chiang <sh.chiang04@gmail.com>
6966 Chung-Ju Wu <jasonwucj@gmail.com>
6968 * config.gcc (nds32*-*-*): Add v2j v3f v3s checking.
6969 (nds32*-*-*): Add float and fpu_config into supported_defaults.
6970 * common/config/nds32/nds32-common.c (TARGET_DEFAULT_TARGET_FLAGS):
6971 Include TARGET_DEFAULT_FPU_ISA and TARGET_DEFAULT_FPU_FMA.
6972 * config/nds32/constants.md (unspec_element): Add UNSPEC_COPYSIGN,
6973 UNSPEC_FCPYNSD, UNSPEC_FCPYNSS, UNSPEC_FCPYSD and UNSPEC_FCPYSS.
6974 * config/nds32/constraints.md: New constraints and checking for hard
6975 float configuration.
6976 * config/nds32/iterators.md: New mode iterator and attribute for hard
6977 float configuration.
6978 * config/nds32/nds32-doubleword.md: Use hard float alternatives and
6980 * config/nds32/nds32-fpu.md: New file.
6981 * config/nds32/nds32-md-auxiliary.c: New functions and checkings to
6982 deal with hard float code generation.
6983 * config/nds32/nds32-opts.h (nds32_arch_type): Add ARCH_V3F and
6985 (abi_type, float_reg_number): New enum type.
6986 * config/nds32/nds32-predicates.c: New predicates for hard float.
6987 * config/nds32/nds32-protos.h: Declare functions for hard float.
6988 * config/nds32/nds32.c: Implementation for hard float configuration.
6989 * config/nds32/nds32.h: Definitions for hard float configuration.
6990 * config/nds32/nds32.md: Include hard float machine description and
6991 modify patterns for hard float configuration.
6992 * config/nds32/nds32.opt: New options for hard float configuration.
6993 * config/nds32/predicates.md: New predicates for hard float
6996 2018-04-06 Kuan-Lin Chen <kuanlinchentw@gmail.com>
6998 * common/config/nds32/nds32-common.c
6999 (nds32_option_optimization_table): Enable -mreleax-hint by default.
7001 2018-04-05 Jakub Jelinek <jakub@redhat.com>
7004 * match.pd (BIT_FIELD_REF CONSTRUCTOR@0 @1 @2): Use view_convert around
7005 CONSTRUCTOR_ELT (ctor, ...)->value.
7007 2018-04-05 Uros Bizjak <ubizjak@gmail.com>
7010 * config/i386/i386.md (define_attr "memory"): Handle rotate1 type.
7012 2018-04-05 Tom de Vries <tom@codesourcery.com>
7015 * config/nvptx/nvptx.c (nvptx_single): Fix neutering of bb with only
7018 2018-04-05 Shiva Chen <shiva0217@gmail.com>
7019 Kito Cheng <kito.cheng@gmail.com>
7021 * config/nds32/constraints.md (U33): Fine-tune checking condition.
7022 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Ditto.
7023 * config/nds32/nds32.h (nds32_16bit_address_type): Add
7024 ADDRESS_POST_MODIFY_LO_REG_IMM3U.
7026 2018-04-05 Shiva Chen <shiva0217@gmail.com>
7027 Kito Cheng <kito.cheng@gmail.com>
7029 * config/nds32/constraints.md (Ufe): New memory constraint.
7030 * config/nds32/nds32-md-auxiliary.c (nds32_mem_format,
7031 nds32_output_16bit_load): Consider r8 register for lwi45.fe format.
7032 * config/nds32/nds32.c (nds32_print_operand): Output lwi45.fe
7034 * config/nds32/nds32.h (nds32_16bit_address_type): Add ADDRESS_R8_IMM7U.
7035 * config/nds32/nds32.md (*mov<mode>): Adjust pattern.
7037 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7039 * config/nds32/nds32.md: Use optimize_size in the condition for
7040 alu-shift instructions.
7042 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7044 * config/nds32/nds32.md (divsi4, udivsi4): New patterns.
7046 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7048 * config/nds32/nds32.md (negsi2): Refine pattern.
7050 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
7051 Chung-Ju Wu <jasonwucj@gmail.com>
7053 * config/nds32/iterators.md (shift_rotate): New code iterator.
7054 (shift): New code attribute.
7055 * config/nds32/nds32-md-auxiliary.c (nds32_expand_constant): New.
7056 * config/nds32/nds32-protos.h (nds32_expand_constant): Declare.
7057 * config/nds32/nds32.c (nds32_print_operand): Deal with more cases.
7058 * config/nds32/nds32.md (addsi3, *add_srli): Refine implementation for
7059 bit-wise operations.
7060 (andsi3, *andsi3): Ditto.
7061 (iorsi3, *iorsi3, *or_slli, *or_srli): Ditto.
7062 (xorsi3, *xorsi3, *xor_slli, *xor_srli): Ditto.
7063 (<shift>si3, *ashlsi3, *ashrsi3, *lshrsi3, *rotrsi3): Ditto.
7064 * config/nds32/predicates.md (nds32_rimm5u_operand, nds32_and_operand,
7065 nds32_ior_operand, nds32_xor_operand): New predicates.
7067 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7069 * config/nds32/nds32.md (add<mode>3, sub<mode>3): Rename to ...
7070 (addsi3, subsi3): ... this.
7072 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7074 * config/nds32/nds32.md (*sub_srli, *and_slli): Fine-tune predicator.
7076 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7078 * config/nds32/nds32.md: Adjust indention.
7080 2018-04-05 Kito Cheng <kito.cheng@gmail.com>
7082 * config/nds32/nds32.md (feature): New attribute.
7084 2018-04-05 Chung-Ju Wu <jasonwucj@gmail.com>
7086 * config/nds32/nds32.md (subtype): New attribute.
7088 2018-04-04 Thomas Preud'homme <thomas.preudhomme@arm.com>
7091 * config/arm/arm-builtins.c (arm_expand_builtin): Change
7092 expansion to perform a bitwise AND of the argument followed by a
7093 boolean negation of the result.
7095 2018-04-04 Peter Bergner <bergner@vnet.ibm.com>
7097 PR rtl-optimization/84878
7098 * ddg.c (add_cross_iteration_register_deps): Use DF_REF_BB to determine
7099 the basic block. Assert the use reference is not artificial and that
7100 it has an associated insn.
7102 2018-04-04 Michael Matz <matz@suse.de>
7104 * builtins.c (compute_objsize): Pass correct operand
7105 to array_at_struct_end_p.
7107 2018-04-04 Richard Biener <rguenther@suse.de>
7110 * dwarf2out.c (dwarf2out_register_external_die): Peel namespaces
7111 from contexts for DINFO_LEVEL_TERSE and below.
7113 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7115 * config/nds32/nds32-doubleword.md (move_<mode>): Require
7116 resiter_operand condition.
7117 * config/nds32/nds32.md (*move<mode>): Ditto.
7119 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7120 Monk Chiang <sh.chiang04@gmail.com>
7122 * config/nds32/nds32.md (movmisalign<mode>): New pattern.
7124 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7126 * config/nds32/nds32.md (movqi, movhi): Merge into mov<mode>.
7128 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7129 Kito Cheng <kito.cheng@gmail.com>
7131 * config/nds32/nds32-md-auxiliary.c (nds32_inverse_cond_code,
7132 nds32_cond_code_str, output_cond_branch,
7133 output_cond_branch_compare_zero, nds32_expand_cbranch,
7134 nds32_expand_cstore, nds32_expand_movcc,
7135 nds32_output_cbranchsi4_equality_zero,
7136 nds32_output_cbranchsi4_equality_reg,
7137 nds32_output_cbranchsi4_equality_reg_or_const_int,
7138 nds32_output_cbranchsi4_greater_less_zero: New functions.
7139 * config/nds32/nds32-protos.h (nds32_expand_cbranch,
7140 nds32_expand_cstore, nds32_expand_movcc,
7141 nds32_output_cbranchsi4_equality_zero,
7142 nds32_output_cbranchsi4_equality_reg,
7143 nds32_output_cbranchsi4_equality_reg_or_const_int,
7144 nds32_output_cbranchsi4_greater_less_zero): Declare.
7145 * config/nds32/predicates.md (nds32_movecc_comparison_operator,
7146 nds32_rimm11s_operand): New predicates.
7147 * config/nds32/nds32.h (nds32_expand_result_type): New enum type.
7148 * config/nds32/nds32.md: Rewrite all the branch and conditional move
7151 2018-04-04 Kito Cheng <kito.cheng@gmail.com>
7153 * config/nds32/nds32-doubleword.md: Refine all the instruction type.
7154 * config/nds32/nds32.md: Ditto.
7155 * config/nds32/pipelines.md: Ditto.
7157 2018-04-04 Richard Biener <rguenther@suse.de>
7159 PR tree-optimization/85168
7160 * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address): Avoid
7161 propagating abnormals.
7163 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7165 * config/nds32/nds32.md (enabled): Use yes/no for this attribute.
7167 2018-04-04 Chung-Ju Wu <jasonwucj@gmail.com>
7168 Kito Cheng <kito.cheng@gmail.com>
7170 * config/nds32/nds32-md-auxiliary.c (nds32_long_call_p): New function.
7171 * config/nds32/nds32-protos.h (nds32_long_call_p): Declare.
7172 * config/nds32/nds32.c (nds32_function_ok_for_sibcall): New function.
7173 (TARGET_FUNCTION_OK_FOR_SIBCALL): Define.
7174 * config/nds32/nds32.md (sibcall_internal): New.
7175 (sibcall_register): Remove.
7176 (sibcall_immediate): Remove.
7177 (sibcall_value_internal): New.
7178 (sibcall_value_register): Remove.
7179 (sibcall_value_immediate): Remove.
7180 * config/nds32/predicates.md (nds32_general_register_operand): New.
7181 (nds32_call_address_operand): New.
7183 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7185 PR rtl-optimization/85167
7186 * shrink-wrap.c (move_insn_for_shrink_wrap): Don't set bb_uses and
7187 bb_defs if *split_p, instead preinitialize it to NULL.
7189 PR tree-optimization/85156
7190 * builtins.c (fold_builtin_expect): Use save_expr on arg1 to avoid
7191 evaluating the argument multiple times.
7193 2018-04-03 Bill Schmidt <wschmidt@linux.ibm.com>
7195 * config/rs6000/emmintrin.h (_mm_cvtpd_epi32): Use __vector rather
7197 (_mm_cvtpd_ps): Likewise.
7198 (_mm_cvttpd_epi32): Likewise.
7199 * config/rs6000/mmintrin.h (_mm_unpacklo_pi8): Likewise.
7200 * config/rs6000/xmmintrin.h: For strict-ANSI C++ or C11, undefine
7201 vector, pixel, and bool following altivec.h include.
7203 2018-04-03 Martin Sebor <msebor@redhat.com>
7205 * doc/extend.texi (Common Function Attributes): Clarify.
7206 (const attribute): Likewise.
7207 (pure attribute): Likewise.
7209 2018-04-03 Jakub Jelinek <jakub@redhat.com>
7212 * config/i386/i386.c (ix86_expand_vector_set): Use
7213 HOST_WIDE_INT_1U << elt instead of 1 << elt. Formatting fix.
7215 2018-04-03 Uros Bizjak <ubizjak@gmail.com>
7217 * config/i386/i386.c (emit_i387_cw_initialization): Always use logic
7218 instructions when changing rounding bits to preserve precision bits
7219 in the x87 control word.
7221 2018-04-03 Martin Liska <mliska@suse.cz>
7223 PR tree-optimization/82491
7224 * rtl.h (strip_offset_and_add): Replace += suboffset with
7225 poly_uint64 () + suboffset.
7227 2018-03-29 Martin Liska <mliska@suse.cz>
7228 Martin Jambor <mjambor@suse.cz>
7231 * ipa-cp.c (propagate_bits_across_jump_function): Bail out if
7232 param_type is not an integral or pointer type.
7234 2018-04-03 Richard Biener <rguenther@suse.de>
7236 * sese.h (recompute_all_dominators): Remove.
7238 2018-04-02 Martin Sebor <msebor@redhat.com>
7240 * doc/invoke.texi (-Wrestrict): Fix typos.
7242 2018-04-02 Jim Wilson <jimw@sifive.com>
7244 * config/riscv/riscv.h (SHIFT_COUNT_TRUNCATED): Set to zero.
7245 * config/riscv/riscv.md (<optab>si3): Use QImode shift count.
7246 (<optab>di3, <optab>si3_extend): Likewise.
7247 (<optab>si3_mask, <optab>si3_mask_1): New.
7248 (<optab>di3_mask, <optab>di3_mask_1): New.
7249 (<optab>si3_extend_mask, <optab>si3_extend_mask_1): New.
7250 (lshrsi3_zero_extend_1): Use VOIDmode shift count.
7251 * config/riscv/sync.md (atomic_test_and_set): Emit QImode shift count.
7253 2018-04-02 Gerald Pfeifer <gerald@pfeifer.com>
7255 * doc/cpp.texi (Variadic Macros): Fix line continuation in an
7258 2018-04-02 Chung-Ju Wu <jasonwucj@gmail.com>
7260 * config/nds32/nds32.c (TARGET_CANONICALIZE_COMPARISON): Define.
7261 (nds32_canonicalize_comparison): New function.
7263 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7264 Kito Cheng <kito.cheng@gmail.com>
7265 Kuan-Lin Chen <kuanlinchentw@gmail.com>
7267 * config.gcc (nds32): Add nds32-relax-opt.o into extra_objs.
7268 * config/nds32/constants.md (unspec_volatile_element): Add
7269 UNSPEC_VOLATILE_RELAX_GROUP.
7270 * config/nds32/nds32-relax-opt.c: New file.
7271 * config/nds32/nds32-predicates.c
7272 (nds32_symbol_load_store_p): New function.
7273 * config/nds32/nds32-protos.h
7274 (nds32_symbol_load_store_p): Declare function.
7275 (make_pass_nds32_relax_opt): Declare new rtl pass function.
7276 * config/nds32/nds32.c
7277 (nds32_register_pass): New function to register pass.
7278 (nds32_register_passes): New function to register passes.
7279 * config/nds32/nds32.md (relax_group): New pattern.
7280 * config/nds32/nds32.opt (mrelax-hint): New option.
7281 * config/nds32/t-nds32 (nds32-relax-opt.o): New dependency.
7283 2018-04-01 Kito Cheng <kito.cheng@gmail.com>
7285 * config/nds32/t-nds32: Modify files dependency.
7287 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7289 * config/nds32/nds32.h (FUNCTION_PROFILER): Output newline character.
7290 (PROFILE_HOOK): Define its implementation.
7292 2018-04-01 Chung-Ju Wu <jasonwucj@gmail.com>
7294 * config/nds32/nds32.h (WCHAR_TYPE, WCHAR_TYPE_SIZE): Use unsigned int
7295 type and 32-bit size.
7297 2018-04-01 Jakub Jelinek <jakub@redhat.com>
7300 * config/i386/sse.md (V): Add V64QI and V32HI for TARGET_AVX512F.
7301 (V_128_256): New mode iterator.
7302 (*avx512dq_vextract<shuffletype>64x2_1 splitter): New define_split.
7303 (*avx512f_vextract<shuffletype>32x4_1 splitter): Likewise.
7304 (xop_pcmov_<mode><avxsizesuffix>): Use V_128_256 mode iterator instead
7306 * config/i386/i386.c (ix86_expand_vector_set): Improve V32HImode and
7307 V64QImode expansion for !TARGET_AVX512BW && TARGET_AVX512F.
7309 2018-03-31 Segher Boessenkool <segher@kernel.crashing.org>
7312 * config/rs6000/xmmintrin.h (_mm_set_ps, _mm_max_ps): Handle (quiet)
7313 NaN inputs correctly.
7315 2018-03-30 Peter Bergner <bergner@vnet.ibm.com>
7318 * config/rs6000/vsx.md (??r): New mode attribute.
7319 (*vsx_mov<mode>_64bit): Use it.
7320 (*vsx_mov<mode>_32bit): Likewise.
7322 2018-03-30 Martin Sebor <msebor@redhat.com>
7324 PR tree-optimization/84818
7325 * builtins.c (check_access): Use warning_n.
7327 2018-03-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
7330 * config/rs6000/rs6000-string.c (expand_compare_loop): Fix redundant
7332 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Fix redundant
7335 2018-03-30 Julia Koval <julia.koval@intel.com>
7338 * x86-tune.def (movx, partial_reg_dependency): Enable for
7341 2018-03-29 Vladimir Makarov <vmakarov@redhat.com>
7344 * lra-constraints.c (process_alt_operands): Move setting
7345 this_alternative_matches below.
7347 2018-03-29 Martin Liska <mliska@suse.cz>
7350 * doc/invoke.texi: Document how LTO works with debug info.
7351 Describe auto-load support of binutils. Mention 'x86-64'
7352 as valid option value of -march option.
7354 2018-03-29 Jakub Jelinek <jakub@redhat.com>
7356 * config/i386/sse.md (<avx512>_blendm<mode>): Use <sseintprefix>.
7359 * fold-const.c (operand_equal_p): Handle DEBUG_BEGIN_STMT.
7360 For STATEMENT_LIST, pass down OEP_LEXICOGRAPHIC and maybe
7361 OEP_NO_HASH_CHECK for recursive call, to avoid exponential
7364 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
7367 * config/rs6000/rs6000.h (RS6000_BTM_POWERPC64): New define.
7368 (RS6000_BTM_COMMON): Add RS6000_BTM_POWERPC64.
7369 * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add support
7370 for RS6000_BTM_POWERPC64.
7371 (rs6000_invalid_builtin): Add handling for RS6000_BTM_POWERPC64
7372 (rs6000_builtin_mask_names): Add RS6000_BTM_POWERPC64.
7373 * config/rs6000/rs6000-builtin.def (BU_P7_POWERPC64_MISC_2): New macro
7378 2018-03-28 Carl Love <cel@us.ibm.com>
7381 2017-09-27 Carl Love <cel@us.ibm.com>
7383 * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
7384 (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
7385 * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
7388 2018-03-28 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
7390 * config/rs6000/xmmintrin.h (_mm_max_pi16): Use __vector __bool
7391 instead of __vector bool.
7392 (_mm_max_pu8): Likewise.
7393 (_mm_min_pi16): Likewise.
7395 2018-03-28 Peter Bergner <bergner@vnet.ibm.com>
7398 * config/rs6000/rs6000-builtin.def (DIVWEO): Delete macro expansion.
7399 (DIVWEUO): Likewise.
7401 (DIVDEUO): Likewise.
7402 * config/rs6000/rs6000.c (builtin_function_type): Remove support for
7403 DIVWEUO and DIVDEUO.
7404 * config/rs6000/rs6000.md (UNSPEC_DIVEO, UNSPEC_DIVEUO): Delete unspecs.
7405 (UNSPEC_DIV_EXTEND): Remove deleted unspecs.
7406 (div_extend): Likewise.
7407 * doc/extend.texi (__builtin_divweo): Remove documentation for deleted
7409 (__builtin_divweuo): Likewise.
7410 (__builtin_divdeo): Likewise.
7411 (__builtin_divdeuo): Likewise.
7413 2018-03-28 Jakub Jelinek <jakub@redhat.com>
7416 * config/i386/i386.md (*add<mode>3_carry_0, *addsi3_carry_zext_0,
7417 *sub<mode>3_carry_0, *subsi3_carry_zext_0): New patterns.
7419 PR tree-optimization/82004
7420 * gimple-match-head.c (optimize_pow_to_exp): New function.
7421 * match.pd (pow(C,x) -> exp(log(C)*x)): Wrap with #if GIMPLE.
7422 Don't fold to exp if optimize_pow_to_exp is false.
7424 2018-03-28 Martin Liska <mliska@suse.cz>
7427 * calls.c (initialize_argument_information): Fix trailing space.
7428 * common.opt: Fix typo and provide better explanation for
7429 -fsanitize-coverage option.
7430 * config/i386/i386.opt: Fix typo.
7432 2018-03-28 Jakub Jelinek <jakub@redhat.com>
7433 Martin Liska <mliska@suse.cz>
7436 * gimplify.c (asan_poison_variable): Don't do the check for
7437 gimplify_omp_ctxp here.
7438 (gimplify_decl_expr): Do it here.
7439 (gimplify_target_expr): Likewise.
7441 2018-03-28 Martin Liska <mliska@suse.cz>
7444 * config/i386/i386.c (ix86_function_arg_advance): Do not call
7445 chkp_type_bounds_count if MPX is not enabled.
7447 2018-03-27 Chung-Ju Wu <jasonwucj@gmail.com>
7449 * config/nds32/nds32.h (BRANCH_COST): Adjust cost.
7451 2018-03-27 Michael Meissner <meissner@linux.vnet.ibm.com>
7454 * config/rs6000/rs6000.c (create_complex_muldiv): New helper
7455 function to create the function decl for complex long double
7456 multiply and divide for -mabi=ieeelongdouble.
7457 (init_float128_ieee): Call it.
7459 2018-03-27 H.J. Lu <hongjiu.lu@intel.com>
7462 * config/i386/i386.c (ix86_trampoline_init): Insert ENDBR for
7463 -fcf-protection=branch -mibt.
7464 * config/i386/i386.h (TRAMPOLINE_SIZE): Increased by 4 bytes.
7466 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7469 * config/arm/arm.c (arm_valid_symbolic_address): Handle
7470 arm_word_relocations.
7472 2018-03-27 Cesar Philippidis <cesar@codesourcery.com>
7475 * config/nvptx/nvptx.c (nvptx_assemble_decl_begin): Add '[]' to
7476 extern array declarations.
7478 2018-03-27 Richard Biener <rguenther@suse.de>
7481 * match.pd ((A * C) +- (B * C) -> (A+-B) * C): Guard with
7482 explicit single_use checks.
7484 2018-03-27 Richard Biener <rguenther@suse.de>
7486 PR tree-optimization/85082
7487 * tree-ssa-sccvn.c (vn_reference_lookup_or_insert_for_pieces):
7490 2018-03-27 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
7492 * config.gcc (aarch64*-*-linux*): New TARGET_DEFAULT_ASYNC_UNWIND_TABLES
7493 * common/config/aarch64/aarch64-common.c (aarch64_optimization_table[]):
7494 Turn on fasynchronous-unwind-tables and funwind-tables.
7496 2018-03-26 Uros Bizjak <ubizjak@gmail.com>
7499 * config/i386/i386.md (*bmi_blsr_<mode>_cmp): New insn pattern.
7500 (*bmi_blsr_<mode>_ccz): Ditto.
7502 2018-03-26 Tom de Vries <tom@codesourcery.com>
7504 PR tree-optimization/85063
7505 * omp-general.c (offloading_function_p): New function. Factor out
7507 * omp-offload.c (pass_omp_target_link::gate): ... here.
7508 * omp-general.h (offloading_function_p): Declare.
7509 * tree-switch-conversion.c (build_one_array): Mark CSWTCH.x variable
7510 with attribute omp declare target for offloading functions.
7512 2018-03-24 Richard Sandiford <richard.sandiford@linaro.org>
7514 PR tree-optimization/84005
7515 * tree-data-ref.h (get_base_for_alignment): Declare.
7516 * tree-data-ref.c (get_base_for_alignment_1): New function.
7517 (get_base_for_alignment): Likewise.
7518 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
7519 get_base_for_alignment to find a suitable base object, instead
7520 of always using drb->base_address.
7522 2018-03-23 Jakub Jelinek <jakub@redhat.com>
7525 * emit-rtl.c (init_emit_regs): Indicate that VOIDmode MEMs don't have
7526 known size by default.
7528 2018-03-23 Vladimir Makarov <vmakarov@redhat.com>
7531 * lra-constraints.c (process_alt_operands): Don't match BLKmode
7532 and non BLKmode operands.
7534 2018-03-23 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7537 * config/arm/arm.md (unaligned_loadhis): Remove first alternative.
7538 Clean up attributes.
7540 2018-03-23 Richard Biener <rguenther@suse.de>
7543 * dwarf2out.c (rtl_for_decl_location): Do not generate RTL early when
7544 we are going to emit early debug for LTO.
7546 2018-03-23 Jakub Jelinek <jakub@redhat.com>
7549 * function.c (match_asm_constraints_1): Don't optimize if input
7550 doesn't satisfy general_operand predicate for output's mode.
7553 * alias.c (write_dependence_p): Don't require for x_canonicalized
7554 non-VOIDmode if x has VOIDmode.
7557 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): If DECL_REGISTER (base),
7558 just don't try to optimize it rather than assert it never happens.
7560 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
7562 * config/rs6000/rs6000-builtin.def: Remove various BU_ALTIVEC_X
7563 macro expansions for definition of ST_INTERNAL_<mode> and
7564 LD_INTERNAL_<mode> builtins.
7565 * config/rs6000/rs6000-protos.h (rs6000_address_for_altivec):
7567 * config/rs6000/rs6000.c (altivec_expand_ld_builtin): Delete this
7569 (altivec_expand_st_builtin): Likewise.
7570 (altivec_expand_builtin): Remove calls to deleted functions.
7571 (rs6000_address_for_altivec): Delete this function.
7572 * config/rs6000/vector.md: Remove expands for
7573 vector_altivec_load_<mode> and vector_altivec_store_<mode>.
7575 2018-03-22 Sudakshina Das <sudi.das@arm.com>
7578 * config/arm/arm.h (machine_function): Add static_chain_stack_bytes.
7579 * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Avoid
7580 re-computing once computed.
7581 (arm_expand_prologue): Compute machine->static_chain_stack_bytes.
7582 (arm_init_machine_status): Initialize
7583 machine->static_chain_stack_bytes.
7585 2018-03-22 Kelvin Nilsen <kelvin@gcc.gnu.org>
7588 * doc/extend.texi: Add four new prototypes for vec_ld.
7589 * config/rs6000/rs6000-builtin.def (LVX_V1TI): Reorder symbol
7590 definitions for more logical presentation.
7591 * config/rs6000/rs6000-c.c: (altivec_overloaded_builtins): Add
7592 entries for V1TI variants of __builtin_altivec_ld builtin.
7593 * config/rs6000/rs6000.c: (altivec_expand_lv_builtin): Add test and
7594 handling of V1TI variant of LVX icode pattern.
7595 (altivec_expand_builtin): Add case for ALTIVEC_BUILTIN_LVX_V1TI.
7596 (rs6000_gimple_fold_builtin): Likewise.
7597 (altivec_init_builtins): Add code to define
7598 __builtin_altivec_lvx_v1ti function.
7600 2018-03-22 Jakub Jelinek <jakub@redhat.com>
7603 * function.c (match_asm_constraints_1): Don't do the optimization
7604 if input isn't a REG, SUBREG, MEM or constant.
7606 2018-03-22 Tom de Vries <tom@codesourcery.com>
7608 PR tree-optimization/84956
7609 * tree-ssa-tail-merge.c (find_clusters_1): Skip bbs with
7610 bb_has_abnormal_pred.
7612 2018-03-22 Jakub Jelinek <jakub@redhat.com>
7615 * dwarf2asm.c (dw2_output_indirect_constant_1): Set
7616 DECL_INITIAL (decl) to decl at the end.
7617 * varasm.c (use_blocks_for_decl_p): Revert the 2018-03-20 change,
7620 2018-03-21 Joseph Myers <joseph@codesourcery.com>
7622 * doc/extend.texi (__builtin_tgmath): Document when complex
7623 integer types are treated as _Complex _Float64.
7625 2018-03-21 Tom de Vries <tom@codesourcery.com>
7627 * doc/extend.texi (__builtin_extend_pointer): Remove pasto.
7629 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7631 PR tree-optimization/84960
7632 * tree-cfg.c (remove_bb): Don't move forced labels into bb->prev_bb
7633 if it is ENTRY block, move them into single succ of ENTRY in that case.
7635 2018-03-21 Richard Sandiford <richard.sandiford@linaro.org>
7637 PR tree-optimization/84811
7638 * poly-int.h (poly_span_traits): Remove the T3 parameter and
7639 promote HOST_WIDE_INT T2 - T1 results to unsigned HOST_WIDE_INT.
7640 (maybe_in_range_p, known_in_range_p, ranges_known_overlap_p):
7641 (known_subrange_p): Update accordingly. Cast each value involved
7642 in the size comparison, rather than casting the result of the
7645 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7647 PR tree-optimization/84982
7648 * gimple-ssa-store-merging.c (invert_op): Handle boolean inversion
7649 by flipping the least significant bit rather than all bits from
7650 bitpos to bitpos + bitsize - 1.
7652 2018-03-21 Nathan Sidwell <nathan@acm.org>
7654 * doc/extend.texi (Deprecated Features): Remove mention of
7655 long-deleted deprecations.
7657 2018-03-21 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
7660 * configure.ac (gcc_cv_ld_soname) <*-*-solaris2*>: Set.
7661 * configure: Regenerate.
7663 2018-03-21 Tom de Vries <tom@codesourcery.com>
7665 PR tree-optimization/83126
7666 * tree-parloops.c (num_phis): New function.
7667 (gen_parallel_loop): Detect and handle canonicalize_loop_ivs failure.
7669 2018-03-21 Nathan Sidwell <nathan@acm.org>
7671 * doc/extend.texi (Deprecated Features): Update deprecated flags,
7672 mention anon-struct/union members and trailing attributes.
7674 2018-03-21 Bin Cheng <bin.cheng@arm.com>
7676 PR tree-optimization/84969
7677 * tree-loop-distribution.c (fuse_memset_builtins): Don't reorder
7678 builtin memset partitions if they set different rhs values.
7680 2018-03-21 Jakub Jelinek <jakub@redhat.com>
7682 PR rtl-optimization/84989
7683 * simplify-rtx.c (simplify_unary_operation_1): Don't try to simplify
7684 VEC_DUPLICATE with scalar result mode.
7686 2018-03-21 Martin Liska <mliska@suse.cz>
7689 * ipa-icf.c (sem_item_optimizer::fixup_points_to_sets): Remove
7690 not intended return statement.
7692 2018-03-21 Martin Liska <mliska@suse.cz>
7695 * tree-chkp.c (CHKP_ARRAY_MAX_CHECK_STEPS): Define a new macro.
7696 (chkp_find_bound_slots_1): Limit number of iterations.
7698 2018-03-20 David H. Gutteridge <dhgutteridge@sympatico.ca>
7701 * Minor grammar fixes for x86 options.
7703 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7706 * dce.c (delete_unmarked_insns): Don't remove frame related noop moves
7707 holding REG_CFA_RESTORE notes, instead turn them into a USE.
7709 2018-03-20 Peter Bergner <bergner@vnet.ibm.com>
7712 * config/rs6000/altivec.md (altivec_lvx_<mode>_2op): Delete define_insn.
7713 (altivec_lvx_<mode>_1op): Likewise.
7714 (altivec_stvx_<mode>_2op): Likewise.
7715 (altivec_stvx_<mode>_1op): Likewise.
7716 (altivec_lvx_<VM2:mode>): New define_expand.
7717 (altivec_stvx_<VM2:mode>): Likewise.
7718 (altivec_lvx_<VM2:mode>_2op_<P:mptrsize>): New define_insn.
7719 (altivec_lvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
7720 (altivec_stvx_<VM2:mode>_2op_<P:mptrsize>): Likewise.
7721 (altivec_stvx_<VM2:mode>_1op_<P:mptrsize>): Likewise.
7722 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Use new expanders.
7723 (rs6000_gen_lvx): Likewise.
7724 * config/rs6000/rs6000.c (altivec_expand_lv_builtin): Likewise.
7725 (altivec_expand_stv_builtin): Likewise.
7726 (altivec_expand_builtin): Likewise.
7727 * config/rs6000/vector.md: Likewise.
7729 2018-03-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
7732 * config/arm/arm.c (arm_array_mode_supported_p): Return false for
7735 2018-03-20 Richard Biener <rguenther@suse.de>
7738 * config/i386/i386.c (ix86_add_stmt_cost): Only cost
7739 sign-conversions as zero, fall back to standard scalar_stmt
7742 2018-03-20 Martin Liska <mliska@suse.cz>
7745 * predict.c (rebuild_frequencies): Handle case when we have
7746 PROFILE_ABSENT, but flag_guess_branch_prob is false.
7748 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7751 * dwarf2asm.c (dw2_output_indirect_constant_1): Temporarily turn off
7752 flag_section_anchors.
7753 * varasm.c (use_blocks_for_decl_p): Remove hack for
7754 dw2_force_const_mem.
7757 * config/aarch64/aarch64.md (*aarch64_reg_<mode>3_neg_mask2): Rename
7759 (*aarch64_<optab>_reg_<mode>3_neg_mask2): ... this. If pseudos can't
7760 be created, use lowpart_subreg of operands[0] rather than operands[0]
7762 (*aarch64_reg_<mode>3_minus_mask): Rename to ...
7763 (*aarch64_ashl_reg_<mode>3_minus_mask): ... this.
7764 (*aarch64_<optab>_reg_di3_mask2): Use const_int_operand predicate
7765 and n constraint instead of aarch64_shift_imm_di and Usd.
7766 (*aarch64_reg_<optab>_minus<mode>3): Rename to ...
7767 (*aarch64_<optab>_reg_minus<mode>3): ... this.
7769 2018-03-20 Sudakshina Das <sudi.das@arm.com>
7772 * config/arm/neon.md (ashldi3_neon): Update ?s for constraints
7773 to favor GPR over NEON registers.
7774 (<shift>di3_neon): Likewise.
7776 2018-03-20 Tom de Vries <tom@codesourcery.com>
7779 * config/nvptx/nvptx.c (nvptx_single): Don't neuter bar.sync.
7780 (nvptx_process_pars): Emit bar.sync asap and alap.
7782 2018-03-20 Tom de Vries <tom@codesourcery.com>
7785 * config/nvptx/nvptx.c (prevent_branch_around_nothing): Also update
7786 seen_label if seen_label is already set.
7788 2018-03-20 Jakub Jelinek <jakub@redhat.com>
7791 * config/i386/i386.c (fold_builtin_cpu): For features above 31
7792 use __cpu_features2 variable instead of __cpu_model.__cpu_features[0].
7793 Use 1U instead of 1. Formatting fixes.
7796 * builtins.c (fold_builtin_strpbrk): For strpbrk(x, "") use type
7797 instead of TREE_TYPE (s1) for the return value.
7799 2018-03-19 Jakub Jelinek <jakub@redhat.com>
7801 PR tree-optimization/84946
7802 * gimple-ssa-store-merging.c (mem_valid_for_store_merging): Compute
7803 bitsize + bitsize in poly_uint64 rather than poly_int64.
7806 * dwarf2asm.c: Include fold-const.c.
7807 (dw2_output_indirect_constant_1): Set DECL_INITIAL (decl) to ADDR_EXPR
7808 of decl rather than decl itself.
7810 PR rtl-optimization/84643
7811 * memmodel.h (enum memmodel): Add MEMMODEL_MAX enumerator.
7813 2018-03-19 Maxim Ostapenko <m.ostapenko@samsung.com>
7816 * dwarf2asm.c (dw2_output_indirect_constant_1): Disable ASan before
7817 calling assemble_variable.
7819 2018-03-19 Sudakshina Das <sudi.das@arm.com>
7822 * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Modify
7823 instructions for UNLT, UNLE, UNGT, UNGE, UNEQ, UNORDERED and ORDERED.
7825 2018-03-19 Jim Wilson <jimw@sifive.com>
7828 * config/riscv/riscv.c (riscv_function_arg_boundary): Use
7829 PREFERRED_STACK_BOUNDARY instead of STACK_BOUNDARY.
7830 (riscv_first_stack_step): Likewise.
7831 (riscv_option_override): Use STACK_BOUNDARY instead of
7833 * config/riscv/riscv.h (STACK_BOUNDARY): Renamed from
7835 (BIGGEST_ALIGNMENT): Set to 128.
7836 (PREFERRED_STACK_BOUNDARY): Renamed from STACK_BOUNDARY.
7837 (RISCV_STACK_ALIGN): Use PREFERRED_STACK_BOUNDARY instead of
7840 2018-03-19 Richard Biener <rguenther@suse.de>
7842 PR tree-optimization/84933
7843 * tree-vrp.c (set_and_canonicalize_value_range): Treat out-of-bound
7844 values as -INF/INF when canonicalizing an ANTI_RANGE to a RANGE.
7846 2018-03-19 Richard Biener <rguenther@suse.de>
7848 PR tree-optimization/84859
7849 * tree-ssa-phiopt.c (single_trailing_store_in_bb): New function.
7850 (cond_if_else_store_replacement): Perform sinking operation on
7851 single-store BBs regardless of MAX_STORES_TO_SINK setting.
7852 Generalize what a BB with a single eligible store is.
7854 2018-03-19 Richard Biener <rguenther@suse.de>
7856 PR tree-optimization/84929
7857 * tree-data-ref.c (analyze_siv_subscript_cst_affine): Guard
7858 chrec_is_positive against non-chrec arg.
7860 2018-03-19 Tamar Christina <tamar.christina@arm.com>
7863 * config/arm/arm.c (arm_can_change_mode_class): revert r258554.
7865 2018-03-18 Martin Liska <mliska@suse.cz>
7867 PR rtl-optimization/84635
7868 * regrename.c (build_def_use): Use matches_mode only when
7871 2018-03-18 Richard Sandiford <richard.sandiford@linaro.org>
7873 PR tree-optimization/84913
7874 * tree-vect-loop.c (vectorizable_reduction): Don't try to
7875 vectorize chains of COND_EXPRs.
7877 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
7879 * config/nds32/nds32.h (MAX_REGS_PER_ADDRESS): Fix the value.
7881 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
7883 * config/nds32/nds32.h (LOGICAL_OP_NON_SHORT_CIRCUIT): Define.
7885 2018-03-18 Chung-Ju Wu <jasonwucj@gmail.com>
7887 * config/nds32/nds32.h (CLZ_DEFINED_VALUE_AT_ZERO): Define.
7889 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
7890 Kito Cheng <kito.cheng@gmail.com>
7892 * config/nds32/nds32-protos.h (nds32_adjust_reg_alloc_order): Declare.
7893 * config/nds32/nds32.c (nds32_reg_alloc_order_for_speed): New array.
7894 (nds32_adjust_reg_alloc_order): New function.
7895 * config/nds32/nds32.h (ADJUST_REG_ALLOC_ORDER): Define.
7897 2018-03-17 Kito Cheng <kito.cheng@gmail.com>
7899 * config/nds32/nds32.c (nds32_asm_output_mi_thunk,
7900 nds32_print_operand, nds32_print_operand_address): Use
7901 HOST_WIDE_INT_PRINT_DEC instead.
7903 2018-03-17 Chung-Ju Wu <jasonwucj@gmail.com>
7905 * config/nds32/nds32.c (nds32_register_priority): Modify cost.
7907 2018-03-17 Jakub Jelinek <jakub@redhat.com>
7910 * config/i386/i386.c (initial_ix86_tune_features,
7911 initial_ix86_arch_features): Use unsigned HOST_WIDE_INT rather than
7913 (set_ix86_tune_features): Change ix86_tune_mask from unsigned int
7914 to unsigned HOST_WIDE_INT, initialize to HOST_WIDE_INT_1U << ix86_tune
7915 rather than 1u << ix86_tune. Formatting fix.
7916 (ix86_option_override_internal): Change ix86_arch_mask from
7917 unsigned int to unsigned HOST_WIDE_INT, initialize to
7918 HOST_WIDE_INT_1U << ix86_arch rather than 1u << ix86_arch.
7919 (ix86_function_specific_restore): Likewise.
7921 2018-03-16 Jakub Jelinek <jakub@redhat.com>
7924 * postreload.c (reload_combine_recognize_pattern): Perform
7925 INTVAL addition in unsigned HOST_WIDE_INT type to avoid UB and
7926 truncate_int_for_mode the result for the destination's mode.
7929 * hsa-gen.c (mem_type_for_type): Fix comment typo.
7930 * tree-vect-loop-manip.c (vect_create_cond_for_niters_checks):
7932 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
7935 2018-03-16 Vladimir Makarov <vmakarov@redhat.com>
7938 * lra-assigns.c (lra_split_hard_reg_for): Don't use
7939 regno_allocno_class_array and sorted_pseudos.
7940 * lra-constraints.c (spill_hard_reg_in_range): Ignore hard regs in
7941 insns where regno is used.
7943 2018-03-16 Martin Liska <mliska@suse.cz>
7946 * multiple_target.c (create_dispatcher_calls): Redirect
7947 reference in the symbol table.
7949 2018-03-16 Martin Liska <mliska@suse.cz>
7952 * multiple_target.c (create_dispatcher_calls): Redirect also
7955 2018-03-16 Jakub Jelinek <jakub@redhat.com>
7959 * tree.h (TARGET_EXPR_NO_ELIDE): Define.
7960 * gimplify.c (gimplify_modify_expr_rhs): Don't elide TARGET_EXPRs with
7961 TARGET_EXPR_NO_ELIDE flag set unless *expr_p is INIT_EXPR.
7963 2018-03-16 Julia Koval <julia.koval@intel.com>
7965 * doc/invoke.texi (Skylake Server): Add CLWB.
7966 Cannonlake): Remove CLWB.
7968 2018-03-16 Jakub Jelinek <jakub@redhat.com>
7970 PR tree-optimization/84841
7971 * tree-ssa-reassoc.c (INTEGER_CONST_TYPE): Change to 1 << 4 from
7973 (FLOAT_ONE_CONST_TYPE): Define.
7974 (constant_type): Return FLOAT_ONE_CONST_TYPE for -1.0 and 1.0.
7975 (sort_by_operand_rank): Put entries with higher constant_type last
7976 rather than first to match comments.
7978 2018-03-15 Sandra Loosemore <sandra@codesourcery.com>
7980 * config/nios2/nios2.md (movsi_internal): Fix thinko in
7983 2018-03-15 Jakub Jelinek <jakub@redhat.com>
7986 * calls.c (expand_call): For TREE_ADDRESSABLE rettype ignore alignment
7987 check and use address of target always.
7989 2018-03-15 H.J. Lu <hongjiu.lu@intel.com>
7992 * config/i386/i386.c (indirect_thunk_needed): Update comments.
7993 (indirect_thunk_bnd_needed): Likewise.
7994 (indirect_thunks_used): Likewise.
7995 (indirect_thunks_bnd_used): Likewise.
7996 (indirect_return_needed): New.
7997 (indirect_return_bnd_needed): Likewise.
7998 (output_indirect_thunk_function): Add a bool argument for
8000 (output_indirect_thunk_function): Don't generate alias for
8001 function return thunk.
8002 (ix86_code_end): Call output_indirect_thunk_function to generate
8003 function return thunks.
8004 (ix86_output_function_return): Set indirect_return_bnd_needed
8005 and indirect_return_needed instead of indirect_thunk_bnd_needed
8006 and indirect_thunk_needed.
8008 2018-03-15 Olga Makhotina <olga.makhotina@intel.com>
8010 * config/i386/sgxintrin.h (_enclv_u32): New intrinsic.
8011 (__enclv_bc, __enclv_cd, __enclv_generic): New definitions.
8012 (ERDINFO, ETRACKC, ELDBC, ELDUC): New leaves.
8014 2018-03-15 David Malcolm <dmalcolm@redhat.com>
8015 Paul Hua <paul.hua.gm@gmail.com>
8018 * gcc.dg/fixits-pr84852-1.c: Fix filename in dg-regexp.
8020 2018-03-15 Segher Boessenkool <segher@kernel.crashing.org>
8022 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add back the
8023 TARGET_DOUBLE_FLOAT and TARGET_SINGLE_FLOAT conditions on the DFmode
8026 2018-03-15 Tamar Christina <tamar.christina@arm.com>
8029 * config/arm/arm.c (arm_can_change_mode_class): Use GET_MODE_UNIT_SIZE
8030 instead of GET_MODE_SIZE when comparing Units.
8032 2018-03-15 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
8035 * varasm.c (hash_section): Return an unchangeble hash value
8036 * config/aarch64/aarch64.c (aarch64_use_blocks_for_constant_p):
8037 Return !aarch64_can_use_per_function_literal_pools_p ().
8039 2018-03-15 Jakub Jelinek <jakub@redhat.com>
8042 * optabs.c (emit_conditional_move): Pass address of cmode's copy
8043 rather than address of cmode as last argument to prepare_cmp_insn.
8045 2018-03-15 Julia Koval <julia.koval@intel.com>
8047 * config/i386/i386.c (F_AVX512VBMI2, F_GFNI, F_VPCLMULQDQ,
8048 F_AVX512VNNI, F_AVX512BITALG): New.
8050 2018-03-14 John David Anglin <danglin@gcc.gnu.org>
8053 * config/pa/pa.c (pa_emit_move_sequence): Always emit secondary reload
8054 insn for floating-point loads and stores.
8056 2018-03-14 Carl Love <cel@us.ibm.com>
8058 * config/rs6000/rs6000-c.c: Add macro definitions for
8059 ALTIVEC_BUILTIN_VEC_PERMXOR.
8060 * config/rs6000/rs6000.h: Add #define for vec_permxor builtin.
8061 * config/rs6000/rs6000-builtin.def: Add macro expansions for VPERMXOR.
8062 * config/rs6000/altivec.md (altivec_vpermxor): New define expand.
8063 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Add case
8065 * config/doc/extend.texi: Add prototypes for vec_permxor.
8067 2018-03-14 David Malcolm <dmalcolm@redhat.com>
8070 * diagnostic-show-locus.c (class layout_point): Convert m_line
8071 from int to linenum_type.
8072 (line_span::comparator): Use linenum "compare" function when
8073 comparing line numbers.
8074 (test_line_span): New function.
8075 (layout_range::contains_point): Convert param "row" from int to
8077 (layout_range::intersects_line_p): Likewise.
8078 (layout::will_show_line_p): Likewise.
8079 (layout::print_source_line): Likewise.
8080 (layout::should_print_annotation_line_p): Likewise.
8081 (layout::print_annotation_line): Likewise.
8082 (layout::print_leading_fixits): Likewise.
8083 (layout::annotation_line_showed_range_p): Likewise.
8084 (struct line_corrections): Likewise for field m_row.
8085 (line_corrections::line_corrections): Likewise for param "row".
8086 (layout::print_trailing_fixits): Likewise.
8087 (layout::get_state_at_point): Likewise.
8088 (layout::get_x_bound_for_row): Likewise.
8089 (layout::print_line): Likewise.
8090 (diagnostic_show_locus): Likewise for locals "last_line" and "row".
8091 (selftest::diagnostic_show_locus_c_tests): Call test_line_span.
8092 * input.c (selftest::test_linenum_comparisons): New function.
8093 (selftest::input_c_tests): Call it.
8094 * selftest.c (selftest::test_assertions): Test ASSERT_GT,
8095 ASSERT_GT_AT, ASSERT_LT, and ASSERT_LT_AT.
8096 * selftest.h (ASSERT_GT): New macro.
8097 (ASSERT_GT_AT): New macro.
8098 (ASSERT_LT): New macro.
8099 (ASSERT_LT_AT): New macro.
8101 2018-03-14 Segher Boessenkool <segher@kernel.crashing.org>
8103 PR rtl-optimization/84780
8104 * combine.c (distribute_links): Don't make a link based on pc_rtx.
8106 2018-03-14 Martin Liska <mliska@suse.cz>
8108 * tree.c (record_node_allocation_statistics): Use
8109 get_stats_node_kind.
8110 (get_stats_node_kind): New function extracted from
8111 record_node_allocation_statistics.
8112 (free_node): Use get_stats_node_kind.
8114 2018-03-14 Richard Biener <rguenther@suse.de>
8116 * tree-ssa-pre.c (compute_antic_aux): Remove code that asserts
8117 that the value-set of ANTIC_IN doesn't grow.
8120 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
8122 (BB_VISITED_WITH_VISITED_SUCCS): New define.
8123 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
8125 2018-03-14 Julia Koval <julia.koval@intel.com>
8127 * config.gcc (icelake-client, icelake-server): New.
8129 * config/i386/i386.c (initial_ix86_tune_features): Extend to 64 bit.
8130 (initial_ix86_arch_features): Ditto.
8131 (PTA_SKYLAKE): Add SGX.
8132 (PTA_ICELAKE): Remove.
8133 (PTA_ICELAKE_CLIENT): New.
8134 (PTA_ICELAKE_SERVER): New.
8135 (ix86_option_override_internal): Split up icelake on icelake client and
8137 (get_builtin_code_for_version): Ditto.
8138 (fold_builtin_cpu): Ditto.
8139 * config/i386/driver-i386.c (config/i386/driver-i386.c): Ditto.
8140 * config/i386/i386-c.c (ix86_target_macros_internal): Ditto
8141 * config/i386/i386.h (processor_type): Ditto.
8142 * doc/invoke.texi: Ditto.
8144 2018-03-14 Jakub Jelinek <jakub@redhat.com>
8147 * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Handle also
8148 INTEGER_CST offset, add it together with bitpos / 8 and
8149 sign extend based on POINTER_SIZE.
8153 2017-04-20 Uros Bizjak <ubizjak@gmail.com>
8156 * config/i386/constraints.md (Yc): New register constraint.
8157 * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
8158 Use Yc constraint for alternative 2 of operand 0. Remove
8159 preferred_for_speed attribute.
8161 2018-03-14 Richard Biener <rguenther@suse.de>
8163 PR tree-optimization/84830
8164 * tree-ssa-pre.c (compute_antic_aux): Intersect the new ANTIC_IN
8165 with the old one to avoid oscillations.
8167 2018-03-13 Vladimir Makarov <vmakarov@redhat.com>
8170 * lra-assigns.c (find_all_spills_for): Ignore uninteresting
8172 (assign_by_spills): Return a flag of reload assignment failure.
8173 Do not process the reload assignment failures. Do not spill other
8174 reload pseudos if they has the same reg class. Update n if
8176 (lra_assign): Add a return arg. Set up from the result of
8177 assign_by_spills call.
8178 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8179 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8180 usage_insns if it is not NULL.
8181 (spill_hard_reg_in_range): New function.
8182 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8183 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8184 function prototypes.
8185 (lra_assign): Change prototype.
8186 * lra.c (lra): Add code to deal with fails by splitting hard reg
8189 2018-03-01 Palmer Dabbelt <palmer@sifive.com>
8191 * config/riscv/riscv.opt (mrelax): New option.
8192 * config/riscv/riscv.c (riscv_file_start): Emit ".option
8193 "norelax" when riscv_mrelax is disabled.
8194 * doc/invoke.texi (RISC-V): Document "-mrelax" and "-mno-relax".
8196 2018-03-13 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
8199 * config/rs6000/rs6000.c (rs6000_reassociation_width): Disable parallel
8200 reassociation for int modes.
8202 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8204 * tree-vect-loop-manip.c (vect_maybe_permute_loop_masks):
8205 Reverse the choice between VEC_UNPACK_LO_EXPR and VEC_UNPACK_HI_EXPR
8207 * config/aarch64/iterators.md (hi_lanes_optab): New int attribute.
8208 * config/aarch64/aarch64-sve.md
8209 (*aarch64_sve_<perm_insn><perm_hilo><mode>): Rename to...
8210 (aarch64_sve_<perm_insn><perm_hilo><mode>): ...this.
8211 (*extend<mode><Vwide>2): Rename to...
8212 (aarch64_sve_extend<mode><Vwide>2): ...this.
8213 (vec_unpack<su>_<perm_hilo>_<mode>): Turn into a define_expand,
8214 renaming the old pattern to...
8215 (aarch64_sve_punpk<perm_hilo>_<mode>): ...this. Only define
8217 (vec_unpack<su>_<perm_hilo>_<SVE_BHSI:mode>): Turn into a
8218 define_expand, renaming the old pattern to...
8219 (aarch64_sve_<su>unpk<perm_hilo>_<SVE_BHSI:mode>): ...this.
8220 (*vec_unpacku_<perm_hilo>_<mode>_no_convert): Delete.
8221 (vec_unpacks_<perm_hilo>_<mode>): Take BYTES_BIG_ENDIAN into
8222 account when deciding which SVE instruction the optab should use.
8223 (vec_unpack<su_optab>_float_<perm_hilo>_vnx4si): Likewise.
8225 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8227 * config/aarch64/aarch64.md (V4_REGNUM, V8_REGNUM, V12_REGNUM)
8228 (V20_REGNUM, V24_REGNUM, V28_REGNUM, P1_REGNUM, P2_REGNUM, P3_REGNUM)
8229 (P4_REGNUM, P5_REGNUM, P6_REGNUM, P8_REGNUM, P9_REGNUM, P10_REGNUM)
8230 (P11_REGNUM, P12_REGNUM, P13_REGNUM, P14_REGNUM): New define_constants.
8231 (tlsdesc_small_<mode>): Turn a define_expand and use
8232 tlsdesc_small_sve_<mode> for SVE. Rename original define_insn to...
8233 (tlsdesc_small_advsimd_<mode>): ...this.
8234 (tlsdesc_small_sve_<mode>): New pattern.
8236 2018-03-13 Richard Sandiford <richard.sandiford@linaro.org>
8238 * config/aarch64/iterators.md (UNSPEC_SMUL_HIGHPART)
8239 (UNSPEC_UMUL_HIGHPART): New constants.
8240 (MUL_HIGHPART): New int iteraor.
8241 (su): Handle UNSPEC_SMUL_HIGHPART and UNSPEC_UMUL_HIGHPART.
8242 * config/aarch64/aarch64-sve.md (<su>mul<mode>3_highpart): New
8244 (*<su>mul<mode>3_highpart): New define_insn.
8246 2018-03-13 Eric Botcazou <ebotcazou@adacore.com>
8249 * ipa-devirt.c (odr_subtypes_equivalent_p): Do not get the ODR type of
8252 2018-03-13 Martin Liska <mliska@suse.cz>
8255 * (sem_item_optimizer::sem_item_optimizer): Initialize new
8257 (sem_item_optimizer::~sem_item_optimizer): Release it.
8258 (sem_item_optimizer::merge_classes): Register variable aliases.
8259 (sem_item_optimizer::fixup_pt_set): New function.
8260 (sem_item_optimizer::fixup_points_to_sets): Likewise.
8261 * ipa-icf.h: Declare new variables and functions.
8263 2018-03-13 Jakub Jelinek <jakub@redhat.com>
8266 * match.pd ((A & C) != 0 ? D : 0): Use INTEGER_CST@2 instead of
8267 integer_pow2p@2 and test integer_pow2p in condition.
8268 (A < 0 ? C : 0): Similarly for @1.
8271 * stmt.c (parse_output_constraint): If the CONSTRAINT_LEN (*p, p)
8272 characters starting at p contain '\0' character, don't look beyond
8276 * config/i386/i386.md (round<mode>2): For 387 fancy math, disable
8277 pattern if -ftrapping-math -fno-fp-int-builtin-inexact.
8280 * reg-stack.c (change_stack): Change update_end var from int to
8281 rtx_insn *, if non-NULL don't update just BB_END (current_block), but
8282 also call set_block_for_insn on the newly added insns and rescan.
8285 * config/i386/sse.md (sse2_loadhpd): Use Yv constraint rather than v
8286 on the last operand.
8289 * tree.c (stabilize_reference_1): Return save_expr (e) for
8290 STATEMENT_LIST even if it doesn't have side-effects.
8292 2018-03-12 Jonathan Wakely <jwakely@redhat.com>
8294 * doc/invoke.texi (-mclflushopt): Fix spelling of option.
8296 2018-03-12 Renlin Li <renlin.li@arm.com>
8298 * config/aarch64/aarch64.md (movhf_aarch64): Fix mode argument to
8299 aarch64_output_scalar_simd_mov_immediate.
8301 2018-03-12 Martin Sebor <msebor@redhat.com>
8303 PR tree-optimization/83456
8304 * gimple-fold.c (gimple_fold_builtin_memory_op): Avoid warning
8305 for perfectly overlapping calls to memcpy.
8306 (gimple_fold_builtin_memory_chk): Same.
8307 (gimple_fold_builtin_strcpy): Handle no-warning.
8308 (gimple_fold_builtin_stxcpy_chk): Same.
8309 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Handle no-warning.
8311 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
8313 * config/rs6000/rs6000.c (abi_v4_pass_in_fpr): Add bool "named"
8314 parameter. Use it for SFmode.
8315 (rs6000_function_arg_advance_1): Adjust.
8316 (rs6000_function_arg): Adjust.
8317 (rs6000_gimplify_va_arg): Pass false for that new parameter.
8319 2018-03-12 Segher Boessenkool <segher@kernel.crashing.org>
8321 PR rtl-optimization/84169
8322 PR rtl-optimization/84780
8323 * combine.c (can_combine_p): Check for a 2-insn combination whether
8324 the destination register is used between the two insns, too.
8326 2018-03-12 Richard Biener <rguenther@suse.de>
8328 PR tree-optimization/84803
8329 * tree-if-conv.c (ifcvt_memrefs_wont_trap): Don't do anything
8330 for refs DR analysis didn't process.
8332 2018-03-12 Richard Biener <rguenther@suse.de>
8334 PR tree-optimization/84777
8335 * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): For
8336 force-vectorize loops ignore whether we are optimizing for size.
8338 2018-03-12 Chung-Ju Wu <jasonwucj@gmail.com>
8340 * config/nds32/nds32.c (nds32_md_asm_adjust): New function.
8341 (TARGET_MD_ASM_ADJUST): Define.
8343 2018-03-12 Monk Chiang <sh.chiang04@gmail.com>
8344 Kito Cheng <kito.cheng@gmail.com>
8345 Chung-Ju Wu <jasonwucj@gmail.com>
8347 * config/nds32/nds32.c (nds32_compute_stack_frame,
8348 nds32_emit_stack_push_multiple, nds32_emit_stack_pop_multiple,
8349 nds32_emit_stack_v3push, nds32_emit_stack_v3pop,
8350 nds32_emit_adjust_frame, nds32_expand_prologue, nds32_expand_epilogue,
8351 nds32_expand_prologue_v3push, nds32_expand_epilogue_v3pop): Refine.
8352 * config/nds32/nds32.h (NDS32_FIRST_CALLEE_SAVE_GPR_REGNUM,
8353 NDS32_LAST_CALLEE_SAVE_GPR_REGNUM, NDS32_V3PUSH_AVAILABLE_P): New.
8354 * config/nds32/nds32.md (prologue, epilogue): Use macro
8355 NDS32_V3PUSH_AVAILABLE_P to do checking.
8357 2018-03-11 Jakub Jelinek <jakub@redhat.com>
8360 * dwarf2out.c (gen_enumeration_type_die): Don't guard adding
8361 DW_AT_declaration for ENUM_IS_OPAQUE on -gdwarf-4 or -gno-strict-dwarf,
8362 but on TYPE_SIZE. Don't do anything for ENUM_IS_OPAQUE if not creating
8363 a new die. Don't set TREE_ASM_WRITTEN if ENUM_IS_OPAQUE. Guard
8364 addition of most attributes on !orig_type_die or the attribute not
8365 being present already. Assert TYPE_VALUES is NULL for ENUM_IS_OPAQUE.
8367 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8368 Chung-Ju Wu <jasonwucj@gmail.com>
8370 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): Modify to define
8372 * config/nds32/nds32.opt (mvh): New option.
8374 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8375 Chung-Ju Wu <jasonwucj@gmail.com>
8377 * config/nds32/nds32-protos.h (nds32_cpu_cpp_builtins): Declare
8379 * config/nds32/nds32.c (nds32_cpu_cpp_builtins): New function.
8380 * config/nds32/nds32.h (TARGET_CPU_CPP_BUILTINS): Modify its
8383 2018-03-11 Kito Cheng <kito.cheng@gmail.com>
8384 Chung-Ju Wu <jasonwucj@gmail.com>
8386 * config/nds32/nds32-memory-manipulation.c (nds32_expand_strlen): New
8388 * config/nds32/nds32-multiple.md (strlensi): New pattern.
8389 * config/nds32/nds32-protos.h (nds32_expand_strlen): Declare function.
8391 2018-03-11 Monk Chiang <sh.chiang04@gmail.com>
8392 Kito Cheng <kito.cheng@gmail.com>
8393 Chung-Ju Wu <jasonwucj@gmail.com>
8395 * config/nds32/constants.md (unspec_element): Add UNSPEC_FFB,
8396 UNSPEC_FFMISM and UNSPEC_FLMISM.
8397 * config/nds32/nds32-intrinsic.c (bdesc_2arg): Add builtin description
8398 for ffb, ffmism and flmism.
8399 * config/nds32/nds32-intrinsic.md (unspec_ffb): Define new pattern.
8400 (unspec_ffmism): Ditto.
8401 (unspec_flmism): Ditto.
8402 (nds32_expand_builtin_impl): Check if string extension is available.
8403 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_FFB,
8404 NDS32_BUILTIN_FFMISM and NDS32_BUILTIN_FLMISM.
8406 2018-03-10 Vladimir Makarov <vmakarov@redhat.com>
8409 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
8412 * lra-assigns.c (assign_by_spills): Return a flag of reload
8413 assignment failure. Do not process the reload assignment
8414 failures. Do not spill other reload pseudos if they has the same
8416 (lra_assign): Add a return arg. Set up from the result of
8417 assign_by_spills call.
8418 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8419 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8420 usage_insns if it is not NULL.
8421 (spill_hard_reg_in_range): New function.
8422 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8423 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8424 function prototypes.
8425 (lra_assign): Change prototype.
8426 * lra.c (lra): Add code to deal with fails by splitting hard reg
8429 2018-03-10 H.J. Lu <hongjiu.lu@intel.com>
8432 * config/i386/i386.opt: Replace Enforcment with Enforcement.
8434 2018-03-10 Alexandre Oliva <aoliva@redhat.com>
8437 * dwarf2out.h (dw_val_class): Add dw_val_class_symview.
8438 (dw_val_node): Add val_symbolic_view.
8439 * dwarf2out.c (dw_line_info_table): Add symviews_since_reset.
8440 (symview_upper_bound): New.
8441 (new_line_info_table): Initialize symviews_since_reset.
8442 (dwarf2out_source_line): Count symviews_since_reset and set
8443 symview_upper_bound.
8444 (dw_val_equal_p): Handle symview.
8445 (add_AT_symview): New.
8446 (print_dw_val): Handle symview.
8447 (attr_checksum, attr_checksum_ordered): Likewise.
8448 (same_dw_val_p, size_of_die): Likewise.
8449 (value_format, output_die): Likewise.
8450 (add_high_low_attributes): Use add_AT_symview for entry_view.
8451 (dwarf2out_finish): Reset symview_upper_bound, clear
8454 2018-03-09 Peter Bergner <bergner@vnet.ibm.com>
8457 * config/rs6000/rs6000.c (rs6000_offsettable_memref_p): New prototype.
8458 Add strict argument and use it.
8459 (rs6000_split_multireg_move): Update for new strict argument.
8460 (mem_operand_gpr): Disallow all non-offsettable addresses.
8461 * config/rs6000/rs6000.md (*movdi_internal64): Use YZ constraint.
8463 2018-03-09 Jakub Jelinek <jakub@redhat.com>
8466 * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Mark va_arg_tmp
8467 temporary TREE_ADDRESSABLE before gimplification of BUILT_IN_MEMCPY.
8468 * config/powerpcspe/powerpcspe.c (rs6000_gimplify_va_arg): Likewise.
8471 * tree-inline.c (copy_tree_body_r): For INDIRECT_REF of a remapped
8472 decl, use remap_type if we want to use the type.
8474 2018-03-09 Martin Sebor <msebor@redhat.com>
8476 PR tree-optimization/84526
8477 * gimple-ssa-warn-restrict.c (builtin_memref::set_base_and_offset):
8479 (builtin_access::generic_overlap): Be prepared to handle non-array
8482 2018-03-09 Alexandre Oliva <aoliva@redhat.com>
8484 PR rtl-optimization/84682
8485 * lra-constraints.c (process_address_1): Check is_address flag
8486 for address constraints.
8487 (process_alt_operands): Likewise.
8488 * lra.c (lra_set_insn_recog_data): Pass asm operand locs to
8489 preprocess_constraints.
8490 * recog.h (preprocess_constraints): Add oploc parameter.
8492 * recog.c (preprocess_constraints): Test address_operand for
8493 CT_ADDRESS constraints.
8495 2018-03-09 Vladimir Makarov <vmakarov@redhat.com>
8498 * lra-assigns.c (assign_by_spills): Return a flag of reload
8499 assignment failure. Do not process the reload assignment
8500 failures. Do not spill other reload pseudos if they has the same
8502 (lra_assign): Add a return arg. Set up from the result of
8503 assign_by_spills call.
8504 (find_reload_regno_insns, lra_split_hard_reg_for): New functions.
8505 * lra-constraints.c (split_reg): Add a new arg. Use it instead of
8506 usage_insns if it is not NULL.
8507 (spill_hard_reg_in_range): New function.
8508 (split_if_necessary, inherit_in_ebb): Pass a new arg to split_reg.
8509 * lra-int.h (spill_hard_reg_in_range, lra_split_hard_reg_for): New
8510 function prototypes.
8511 (lra_assign): Change prototype.
8512 * lra.c (lra): Add code to deal with fails by splitting hard reg
8515 2018-03-09 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8518 * common/config/arm/arm-common.c (arm_parse_arch_option_name):
8519 Accept complain bool parameter. Only emit errors if it is true.
8520 (arm_parse_cpu_option_name): Likewise.
8521 (arm_target_thumb_only): Adjust callers of the above.
8522 * config/arm/arm-protos.h (arm_parse_cpu_option_name): Adjust
8523 prototype to take a default true bool parameter.
8524 (arm_parse_arch_option_name): Likewise.
8526 2018-03-09 David Malcolm <dmalcolm@redhat.com>
8527 Francois-Xavier Coudert <fxcoudert@gcc.gnu.org>
8531 * Makefile.in (LD_VERSION_SCRIPT_OPTION, LD_SONAME_OPTION): New.
8532 * configure: Regenerate.
8533 * configure.ac ("linker --version-script option"): New.
8534 ("linker soname option"): New.
8536 2018-03-09 Richard Biener <rguenther@suse.de>
8538 PR tree-optimization/84775
8539 * tree-if-conv.c (add_bb_predicate_gimplified_stmts): Delink
8540 immediate uses of predicate stmts and mark them modified.
8543 PR tree-optimization/84178
8544 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
8546 (version_loop_for_if_conversion): Delay update_ssa call.
8547 (tree_if_conversion): Delay update_ssa until after predicate
8550 2018-03-09 Eric Botcazou <ebotcazou@adacore.com>
8553 * config/i386/winnt.c (i386_pe_seh_cold_init): Use small pre-allocation
8554 when the function accesses prior frames.
8556 2018-03-08 Jakub Jelinek <jakub@redhat.com>
8559 * dwarf2out.c (dw_loc_list): If list && loc_list->first->next, call
8560 gen_llsym, otherwise call maybe_gen_llsym.
8563 * recog.c (asm_operand_ok): Return 0 if multi-character constraint
8564 has ',' character inside of it.
8566 2018-03-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
8569 * config/aarch64/aarch64.md (*compare_cstore<mode>_insn): Mark pattern
8570 as clobbering CC_REGNUM.
8572 2018-03-08 Richard Biener <rguenther@suse.de>
8575 * tree-scalar-evolution.c: Include tree-into-ssa.h.
8576 (follow_copies_to_constant): Do not follow SSA names registered
8579 2018-03-08 Richard Biener <rguenther@suse.de>
8581 PR tree-optimization/84178
8582 * tree-if-conv.c (combine_blocks): Move insert_gimplified_predicates
8584 (version_loop_for_if_conversion): Delay update_ssa call.
8585 (tree_if_conversion): Delay update_ssa until after predicate
8588 2018-03-08 David Malcolm <dmalcolm@redhat.com>
8590 PR tree-optimization/84178
8591 * tree-if-conv.c (release_bb_predicate): Remove the
8592 the assertion that the stmts have NULL use_ops.
8593 Discard the statements, asserting that they haven't
8594 yet been added to a BB.
8596 2018-03-08 Richard Biener <rguenther@suse.de>
8598 PR tree-optimization/84746
8599 * tree-ssa-pre.c (find_leader_in_sets): Deal with SET1 being NULL.
8600 (phi_translate): Pass in destination ANTIC_OUT set.
8601 (phi_translate_1): Likewise. For a simplified result lookup
8602 a leader in ANTIC_OUT and AVAIL_OUT, not the ANTIC_IN sets.
8603 (phi_translate_set): Adjust.
8604 (do_pre_regular_insertion): Likewise.
8605 (do_pre_partial_partial_insertion): Likewise.
8607 2018-03-08 Martin Liska <mliska@suse.cz>
8609 PR gcov-profile/84735
8610 * doc/gcov.texi: Document usage of profile files.
8611 * gcov-io.h: Document changes in the format.
8613 2018-03-08 Alexandre Oliva <aoliva@redhat.com>
8617 * dwarf2out.c (struct dw_line_info_table): Update comments for
8619 (FORCE_RESET_NEXT_VIEW): New.
8620 (FORCE_RESETTING_VIEW_P): New.
8621 (RESETTING_VIEW_P): Check for -1 too.
8622 (ZERO_VIEW_P): Likewise.
8623 (new_line_info_table): Force-reset next view.
8624 (dwarf2out_begin_function): Likewise.
8625 (dwarf2out_source_line): Simplify zero_view_p initialization.
8626 Test FORCE_RESETTING_VIEW_P and RESETTING_VIEW_P instead of
8627 view directly. Omit view when omitting .loc at line 0.
8629 2018-03-08 Jakub Jelinek <jakub@redhat.com>
8631 PR tree-optimization/84740
8632 * tree-switch-conversion.c (process_switch): Call build_constructors
8633 only if info.phi_count is non-zero.
8635 PR tree-optimization/84739
8636 * tree-tailcall.c (find_tail_calls): Check call arguments against
8637 DECL_ARGUMENTS (current_function_decl) rather than
8638 DECL_ARGUMENTS (func) when checking for tail recursion.
8640 2018-03-07 Jakub Jelinek <jakub@redhat.com>
8642 * doc/contrib.texi: Add entries for Martin Liska, David Malcolm,
8643 Marek Polacek, extend Vladimir Makarov's, Jonathan Wakely's and
8644 Volker Reichelt's entry and add entries for people that perform
8645 GCC fuzzy testing and report numerous bugs.
8647 2018-03-07 Segher Boessenkool <segher@kernel.crashing.org>
8650 * config/rs6000/rs6000.c (rs6000_elf_in_small_data_p): Don't put
8651 readonly data in sdata, if that is disabled.
8652 * config/rs6000/sysv4.opt (mreadonly-in-sdata): New option.
8653 * doc/invoke.texi (RS/6000 and PowerPC Options): Document
8654 -mreadonly-in-sdata option.
8656 2018-03-07 Martin Sebor <msebor@redhat.com>
8658 PR tree-optimization/84468
8659 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Consider successor
8660 basic block when looking for nul assignment.
8662 2018-03-07 Eric Botcazou <ebotcazou@adacore.com>
8665 * except.h (output_function_exception_table): Adjust prototype.
8666 * except.c (output_function_exception_table): Remove FNNAME parameter
8667 and add SECTION parameter. Ouput one part of the table at a time.
8668 * final.c (final_scan_insn_1) <NOTE_INSN_SWITCH_TEXT_SECTIONS>: Output
8669 the first part of the exception table and emit unwind directives.
8670 * config/i386/i386-protos.h (i386_pe_end_cold_function): Declare.
8671 (i386_pe_seh_cold_init): Likewise.
8672 * config/i386/cygming.h (ASM_DECLARE_COLD_FUNCTION_NAME): New macro.
8673 (ASM_DECLARE_COLD_FUNCTION_SIZE): Likewise.
8674 * config/i386/i386.c (x86_expand_epilogue): Fix wording in comment.
8675 (ix86_output_call_insn): Emit a nop in one more case for SEH.
8676 * config/i386/winnt.c: Include except.h.
8677 (struct seh_frame_state): Add reg_offset, after_prologue and
8678 in_cold_section fields.
8679 (i386_pe_seh_end_prologue): Set seh->after_prologue.
8680 (i386_pe_seh_cold_init): New function.
8681 (i386_pe_seh_fini): Add COLD parameter and bail out if it is not equal
8682 to seh->in_cold_section.
8683 (seh_emit_push): Record the offset of the push.
8684 (seh_emit_save): Record the offet of the save.
8685 (i386_pe_seh_unwind_emit): Deal with NOTE_INSN_SWITCH_TEXT_SECTIONS.
8686 Test seh->after_prologue to disregard the epilogue.
8687 (i386_pe_end_function): Pass FALSE to i386_pe_seh_fini.
8688 (i386_pe_end_cold_function): New function.
8690 2018-03-07 Jakub Jelinek <jakub@redhat.com>
8693 * config/aarch64/predicates.md (aarch64_simd_reg_or_zero): Use
8694 aarch64_simd_or_scalar_imm_zero rather than aarch64_simd_imm_zero.
8697 * gimple-expr.c (create_tmp_var_raw): Set DECL_NAMELESS flag
8699 * tree-pretty-print.c (dump_decl_name): For TDF_COMPARE_DEBUG,
8700 don't print names of DECL_NAMELESS DECL_IGNORED_P decls.
8703 * multiple_target.c: Include tree-inline.h and intl.h.
8704 (expand_target_clones): Diagnose and fail if node->definition and
8705 !tree_versionable_function_p (node->decl).
8707 2018-03-06 John David Anglin <danglin@gcc.gnu.org>
8709 * config/pa/pa.h (ASM_GENERATE_INTERNAL_LABEL): Revise to use
8711 (ASM_OUTPUT_ADDR_VEC_ELT): Revise for above change.
8712 (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
8713 * config/pa/pa64-hpux.h (ASM_GENERATE_INTERNAL_LABEL): Revise as above.
8715 2018-03-06 Jakub Jelinek <jakub@redhat.com>
8718 * combine.c (try_combine): Use reg_or_subregno instead of handling
8719 just paradoxical SUBREGs and REGs.
8721 2018-03-06 Claudiu Zissulescu <claziss@synopsys.com>
8723 * config/arc/arc.c (arc_finalize_pic): Remove function.
8724 (arc_must_save_register): We use single base PIC register, remove
8725 checks to save/restore the PIC register.
8726 (arc_expand_prologue): Likewise.
8727 * config/arc/arc-protos.h (arc_set_default_type_attributes):
8729 (arc_verify_short): Likewise.
8730 (arc_attr_type): Likewise.
8731 * config/arc/arc.c (arc_set_default_type_attributes): Remove.
8732 (walk_stores): Likewise.
8733 (arc_address_cost): Make it static.
8734 (arc_verify_short): Likewise.
8735 (branch_dest): Likewise.
8736 (arc_attr_type): Likewise.
8737 * config/arc/arc.c (TARGET_ADJUST_INSN_LENGTH): Remove.
8738 (TARGET_INSN_LENGTH_PARAMETERS): Likewise.
8739 (arc_final_prescan_insn): Remove inserting the nops due to
8740 hardware hazards. It is done in reorg step.
8741 (insn_length_variant_t): Remove.
8742 (insn_length_parameters_t): Likewise.
8743 (arc_insn_length_parameters): Likewise.
8744 (arc_get_insn_variants): Likewise.
8745 * config/arc/arc.h (TARGET_UPSIZE_DBR): Remove.
8747 2018-03-06 Jakub Jelinek <jakub@redhat.com>
8750 * reg-stack.c (move_for_stack_reg): If any_malformed_asm, avoid
8753 PR tree-optimization/84687
8754 * omp-simd-clone.c (simd_clone_create): Clear DECL_BUILT_IN_CLASS
8756 * match.pd (pow(C,x)*expN(y) -> expN(logN(C)*x+y)): New optimization.
8758 2018-03-05 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
8760 * config/rs6000/rs6000-builtin.def (rs6000_speculation_barrier):
8761 Rename to ppc_speculation_barrier.
8762 * config/rs6000/rs6000.c (rs6000_init_builtins): Rename builtin to
8763 __builtin_ppc_speculation_barrier.
8765 2018-03-05 Jakub Jelinek <jakub@redhat.com>
8768 * combine.c (combine_simplify_rtx): Don't try to simplify if
8769 if_then_else_cond returned non-NULL, but either true_rtx or false_rtx
8772 2018-03-05 Segher Boessenkool <segher@kernel.crashing.org>
8774 * config/rs6000/rs6000.c (rs6000_loop_align): Don't align tiny loops
8775 to 32 bytes when compiling for POWER9.
8777 2018-03-05 Jakub Jelinek <jakub@redhat.com>
8780 * config/i386/i386.c (ix86_function_ok_for_sibcall): Check for
8781 regparm >= 3 with no arg reg available also for calls with
8782 flag_force_indirect_call. Pass decl to ix86_function_regparm.
8785 * config/i386/sse.md (*<code><mode>3): Replace <mask_prefix3> with
8787 (*<plusminus_insn><mode>3): Likewise. Remove <mask_operand3> uses.
8789 2018-03-05 Peter Bergner <bergner@vnet.ibm.com>
8792 * config/rs6000/vector.md (mov<mode>): Disallow altivec memory operands.
8794 2018-03-05 Richard Biener <rguenther@suse.de>
8796 PR tree-optimization/84486
8797 * tree-ssa-pre.c (create_expression_by_pieces): Remove dead code.
8798 When inserting a __builtin_assume_aligned call set the LHS
8799 SSA name alignment info accordingly.
8801 2018-03-05 Wilco Dijkstra <wdijkstr@arm.com>
8803 PR tree-optimization/84114
8804 * config/aarch64/aarch64.c (aarch64_reassociation_width)
8805 Avoid reassociation of FLOAT_MODE addition.
8807 2018-03-05 Olga Makhotina <olga.makhotina@intel.com>
8809 * common/config/i386/i386-common.c (OPTION_MASK_ISA_PCONFIG_SET,
8810 OPTION_MASK_ISA_PCONFIG_UNSET, OPTION_MASK_ISA_WBNOINVD_SET,
8811 OPTION_MASK_ISA_WBNOINVD_UNSET): New definitions.
8812 (ix86_handle_option): Handle -mpconfig and -mwbnoinvd.
8813 * config.gcc (pconfigintrin.h, wbnoinvdintrin.h) : Add headers.
8814 * config/i386/cpuid.h (bit_PCONFIG, bit_WBNOINVD): New.
8815 * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mpconfig
8817 * config/i386/i386-builtin.def (__builtin_ia32_wbnoinvd,
8818 __builtin_ia32_wbinvd): New builtins.
8819 (SPECIAL_ARGS2): New.
8820 * config/i386/i386-c.c (__WBNOINVD__, __PCONFIG__): New.
8821 (SPECIAL_ARGS2): New.
8822 * config/i386/i386.c (ix86_target_string): Add -mpconfig and -mwbnoinvd.
8823 (ix86_valid_target_attribute_inner_p): Ditto.
8824 (ix86_init_mmx_sse_builtins): Add special_args2.
8825 * config/i386/i386.h (TARGET_PCONFIG, TARGET_PCONFIG_P, TARGET_WBNOINVD,
8826 TARGET_WBNOINVD_P): New.
8827 * config/i386/i386.md (UNSPECV_WBINVD, UNSPECV_WBNOINVD): New.
8828 (define_insn "wbinvd", define_insn "wbnoinvd"): New.
8829 * config/i386/i386.opt: Add -mpconfig and -mwbnoinvd.
8830 * config/i386/immintrin.h (_wbinvd): New intrinsic.
8831 * config/i386/pconfigintrin.h: New file.
8832 * config/i386/wbnoinvdintrin.h: Ditto.
8833 * config/i386/x86intrin.h: Add headers pconfigintrin.h and
8835 * doc/invoke.texi (-mpconfig, -mwbnoinvd): New.
8837 2018-03-05 Richard Biener <rguenther@suse.de>
8839 PR tree-optimization/84670
8840 * tree-ssa-pre.c (struct bb_bitmap_sets): Add visited_with_visited_succs
8842 (BB_VISITED_WITH_VISITED_SUCCS): New define.
8843 (compute_antic): Initialize BB_VISITED_WITH_VISITED_SUCCS.
8844 (compute_antic_aux): Only assert the number of values in ANTIC_IN
8845 doesn't grow if all successors (recursively) were visited at least
8848 2018-03-05 Richard Biener <rguenther@suse.de>
8850 PR tree-optimization/84650
8851 * tree-ssa-loop-im.c (pass_lim::execute): Reset the SCEV cache
8852 if executed in the loop pipeline.
8854 2018-03-05 Sandra Loosemore <sandra@codesourcery.com>
8856 * doc/configfiles.texi (Configuration Files): Move info about
8857 conditionalizing $target-protos.h to...
8858 * doc/sourcebuild.texi (Back End): Here. Explain how $target.h
8859 differs from $target-protos.h.
8861 2018-03-05 Kito Cheng <kito.cheng@gmail.com>
8862 Chung-Ju Wu <jasonwucj@gmail.com>
8864 * config/nds32/nds32-protos.h (nds32_expand_setmem): Declare.
8865 * config/nds32/nds32-multiple.md (setmemsi): Define.
8866 * config/nds32/nds32-memory-manipulation.c
8867 (nds32_gen_dup_4_byte_to_word_value): New.
8868 (emit_setmem_word_loop): New.
8869 (emit_setmem_byte_loop): New.
8870 (nds32_expand_setmem_loop): New.
8871 (nds32_expand_setmem_loop_v3m): New.
8872 (nds32_expand_setmem_unroll): New.
8873 (nds32_expand_setmem): New.
8875 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
8876 Chung-Ju Wu <jasonwucj@gmail.com>
8878 * config/nds32/nds32-memory-manipulation.c
8879 (nds32_emit_load_store): New.
8880 (nds32_emit_post_inc_load_store): New.
8881 (nds32_emit_mem_move): New.
8882 (nds32_emit_mem_move_block): New.
8883 (nds32_expand_movmemsi_loop_unknown_size): New.
8884 (nds32_expand_movmemsi_loop_known_size): New.
8885 (nds32_expand_movmemsi_loop): New.
8886 (nds32_expand_movmemsi_unroll): New.
8887 (nds32_expand_movmemqi): Rename ...
8888 (nds32_expand_movmemsi): ... to this.
8889 * config/nds32/nds32-multiple.md (movmemqi): Rename ...
8890 (movmemsi): ... to this.
8891 * config/nds32/nds32-protos.h (nds32_expand_movmemqi): Rename ...
8892 (nds32_expand_movmemsi): ... to this.
8894 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
8895 Monk Chiang <sh.chiang04@gmail.com>
8896 Chung-Ju Wu <jasonwucj@gmail.com>
8898 * config/nds32/nds32-protos.h
8899 (nds32_expand_load_multiple): New arguments.
8900 (nds32_expand_store_multiple): Ditto.
8901 (nds32_valid_multiple_load_store): Rename ...
8902 (nds32_valid_multiple_load_store_p): ... to this.
8903 * config/nds32/nds32-memory-manipulation.c
8904 (nds32_expand_load_multiple): Refine implementation.
8905 (nds32_expand_store_multiple): Ditto.
8906 * config/nds32/nds32-multiple.md
8907 (load_multiple): Update nds32_expand_load_multiple interface.
8908 (store_multiple): Update nds32_expand_store_multiple interface.
8909 * config/nds32/nds32-predicates.c
8910 (nds32_valid_multiple_load_store): Rename ...
8911 (nds32_valid_multiple_load_store_p): ... to this and refine
8913 * config/nds32/predicates.md
8914 (nds32_load_multiple_and_update_address_operation): New predicate.
8915 (nds32_store_multiple_and_update_address_operation): New predicate.
8917 2018-03-04 Kito Cheng <kito.cheng@gmail.com>
8918 Chung-Ju Wu <jasonwucj@gmail.com>
8920 * config/nds32/nds32.md (type): Add load_multiple and store_multiple.
8921 (combo): New attribute.
8922 * config/nds32/nds32-multiple.md: Refine patterns with new attributes.
8924 2018-03-03 Chung-Ju Wu <jasonwucj@gmail.com>
8926 * config/nds32/nds32.opt: Change -mcmodel= default value.
8928 2018-03-03 Kito Cheng <kito.cheng@gmail.com>
8929 Monk Chiang <sh.chiang04@gmail.com>
8930 Chung-Ju Wu <jasonwucj@gmail.com>
8932 * config/nds32/constants.md (unspec_element): New enum.
8933 * config/nds32/constraints.md (Umw): New constraint.
8934 * config/nds32/nds32-intrinsic.c: Add more builtin functions.
8935 * config/nds32/nds32-intrinsic.md: Likewise.
8936 * config/nds32/nds32-md-auxiliary.c (nds32_regno_to_enable4): New.
8937 (nds32_valid_smw_lwm_base_p): New.
8938 (nds32_output_smw_single_word): New.
8939 (nds32_output_lmw_single_word): New.
8940 (nds32_expand_unaligned_load): New.
8941 (nds32_expand_unaligned_store): New.
8942 * config/nds32/nds32-protos.h (nds32_valid_smw_lwm_base_p): Declare.
8943 (nds32_output_smw_single_word): Declare.
8944 (nds32_output_lmw_single_word): Declare.
8945 (nds32_expand_unaligned_load): Declare.
8946 (nds32_expand_unaligned_store): Declare.
8947 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_UALOAD_HW,
8948 NDS32_BUILTIN_UALOAD_W, NDS32_BUILTIN_UALOAD_DW,
8949 NDS32_BUILTIN_UASTORE_HW, NDS32_BUILTIN_UASTORE_W,
8950 NDS32_BUILTIN_UASTORE_DW.
8951 * config/nds32/predicates.md (nds32_lmw_smw_base_operand): New
8954 2018-03-03 Monk Chiang <sh.chiang04@gmail.com>
8955 Kito Cheng <kito.cheng@gmail.com>
8956 Chung-Ju Wu <jasonwucj@gmail.com>
8958 * config/nds32/nds32-intrinsic.c
8959 (nds32_expand_builtin_null_ftype_reg): Delete.
8960 (nds32_expand_builtin_reg_ftype_imm): Ditto.
8961 (nds32_expand_builtin_null_ftype_reg_imm): Ditto.
8962 (nds32_read_argument): New.
8963 (nds32_legitimize_target): Ditto.
8964 (nds32_legitimize_argument): Ditto.
8965 (nds32_check_constant_argument): Ditto.
8966 (nds32_expand_unop_builtin): Ditto.
8967 (nds32_expand_unopimm_builtin): Ditto.
8968 (nds32_expand_binop_builtin): Ditto.
8969 (nds32_builtin_decl_impl): Ditto.
8970 (builtin_description): Ditto.
8971 (nds32_expand_builtin_impl): Rewrite with new infrastructure.
8972 (nds32_init_builtins_impl): Ditto.
8973 * config/nds32/nds32.c (TARGET_BUILTIN_DECL): Define.
8974 (nds32_builtin_decl): New.
8975 * config/nds32/nds32.h (nds32_builtins): Add NDS32_BUILTIN_COUNT.
8976 * config/nds32/nds32-protos.h (nds32_builtin_decl_impl): Declare.
8978 2018-03-02 Jeff Law <law@redhat.com>
8980 * reorg.c (stop_search_p): Handle DEBUG_INSN.
8981 (redundant_insn, fill_simple_delay_slots): Likewise.
8982 (fill_slots_from_thread): Likewise.
8983 * resource.c (mark_referenced_resources): Likewise.
8984 (mark_set_resources, find_dead_or_set_registers): Likewise.
8986 2018-03-02 Jakub Jelinek <jakub@redhat.com>
8988 * substring-locations.h (format_warning_va): Formatting fix for
8990 (format_warning_at_substring): Fix up ATTRIBUTE_GCC_DIAG second
8992 (format_warning_n_va, format_warning_at_substring_n): New prototypes.
8993 * substring-locations.c: Include intl.h.
8994 (format_warning_va): Turned into small wrapper around
8995 format_warning_n_va, renamed to ...
8996 (format_warning_n_va): ... this, add N and PLURAL_GMSGID arguments,
8997 rename GMSGID to SINGULAR_GMSGID, if SINGULAR_GMSGID != PLURAL_GMSGID,
8999 (format_warning_at_substring_n): New function.
9000 * gimple-ssa-sprintf.c: Remove GCC diagnostic ignored pragma.
9001 (fmtwarn): Add ATTRIBUTE_GCC_DIAG. Turn into a copy of
9002 format_warning_at_substring with just a shorter name instead of
9003 const function pointer.
9004 (fmtwarn_n): New function.
9005 (maybe_warn, format_directive, parse_directive): Use fmtwarn_n where
9006 appropriate, get rid of all the fmtstr temporaries, move conditionals
9007 with G_() wrapped string literals directly into fmtwarn arguments,
9008 cast dir.len to (int), formatting fixes.
9010 2018-03-02 Thomas Schwinge <thomas@codesourcery.com>
9012 * doc/invoke.texi: Remove "Cilk Plus" references.
9014 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9015 Richard Biener <rguenther@suse.de>
9018 * expr.c (expand_expr_real_1) <case CALL_EXPR>: Don't emit diagnostics
9019 for error or warning attributes if CALL_FROM_THUNK_P is set.
9022 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9025 * config/pa/pa.h (TARGET_CPU_CPP_BUILTINS): Predefine
9026 __SIZEOF_128__ macro if HPUX_LONG_DOUBLE_LIBRARY.
9029 * config/ia64/ia64.h (TARGET_CPU_CPP_BUILTINS): Predefine
9030 __SIZEOF_{FPREG,FLOAT{80,128}}__ macros.
9032 * predict.c (test_prediction_value_range): Use PROB_UNINITIALIZED
9033 instead of -1U in last predictors element's probability member.
9035 2018-03-02 Eric Botcazou <ebotcazou@adacore.com>
9038 * ipa-devirt.c (odr_subtypes_equivalent_p): Get the ODR type of both
9039 arguments if they are comparable.
9041 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
9043 PR tree-optimization/84634
9044 * tree-vect-stmts.c (vectorizable_store, vectorizable_load): Replace
9045 masks and masked_loop_p with a single loop_masks, making sure it's
9046 null for bb vectorization.
9048 2018-03-02 Richard Sandiford <richard.sandiford@linaro.org>
9050 * tree-vect-data-refs.c (vect_analyze_data_ref_dependence)
9051 (vect_analyze_data_ref_access): Use loop->safe_len rather than
9052 loop->force_vectorize to check whether there is no alias.
9054 2018-03-02 Jakub Jelinek <jakub@redhat.com>
9057 * rtl.h (prev_real_nondebug_insn, next_real_nondebug_insn): New
9059 * emit-rtl.c (next_real_insn, prev_real_insn): Fix up function
9061 (next_real_nondebug_insn, prev_real_nondebug_insn): New functions.
9062 * cfgcleanup.c (try_head_merge_bb): Use prev_real_nondebug_insn
9063 instead of a loop around prev_real_insn.
9064 * combine.c (move_deaths): Use prev_real_nondebug_insn instead of
9068 * config/i386/i386.c (ix86_print_operand): Use conditional
9069 output_operand_lossage instead of gcc_assert if CONST_VECTOR is not
9072 2018-03-02 Richard Biener <rguenther@suse.de>
9074 PR tree-optimization/84427
9075 * tree-ssa-pre.c (bitmap_remove_expr_from_set): Remove.
9076 (bitmap_set_subtract_values): Rewrite to handle multiple
9079 (prune_clobbered_mems): Likewise.
9080 (phi_translate): Take edge instead of pred/phiblock.
9081 (phi_translate_1): Likewise.
9082 (phi_translate_set): Likewise. Insert all translated
9083 exprs for a value into the set, keeping possibly multiple
9084 expressions per value.
9085 (compute_antic_aux): Adjust for phi_translate changes.
9086 When intersecting union the expressions and prune those
9087 not in the final value set, keeping possibly multiple
9088 expressions per value. Do not use value-insertion
9089 for unioning ANTIC_OUT U EXP_GEN - TMP_GEN but merge
9090 all expressions. Add verification that the value-sets
9091 only shrink during iteration.
9092 (compute_partial_antic_aux): Adjust for the phi_translate changes.
9093 (do_pre_regular_insertion): Likewise.
9094 (do_pre_partial_partial_insertion): Likewise.
9096 2018-03-02 Richard Biener <rguenther@suse.de>
9099 * config/darwin.c (saved_debug_info_level): New static global.
9100 (darwin_asm_lto_start): Disable debug info generation for LTO out.
9101 (darwin_asm_lto_end): Restore debug info generation settings.
9103 2018-03-01 Martin Liska <mliska@suse.cz>
9106 * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
9109 2018-03-01 Richard Biener <rguenther@suse.de>
9112 * dwarf2out.c (gen_variable_die): Properly handle late VLA
9113 type annotation with LTO when debug was disabled at compile-time.
9115 2018-03-01 Matthew Fortune <mfortune@gmail.com>
9117 * config/mips/mips.c (mips_final_prescan_insn): Fix incorrect
9119 (mips_final_postscan_insn): Likewise.
9121 2018-03-01 Richard Sandiford <richard.sandiford@linaro.org>
9123 PR rtl-optimization/84528
9124 * alias.c (init_alias_target): Add commentary.
9125 (init_alias_analysis): Only give HARD_FRAME_POINTER_REGNUM
9126 a unique base value if the frame pointer is not eliminated
9127 to the stack pointer.
9129 2018-03-01 Tom de Vries <tom@codesourcery.com>
9131 PR rtl-optimization/83327
9132 * lra-int.h (hard_regs_spilled_into): Declare.
9133 * lra.c (hard_regs_spilled_into): Define.
9134 (init_reg_info): Init hard_regs_spilled_into.
9135 * lra-spills.c (assign_spill_hard_regs): Update hard_regs_spilled_into.
9136 * lra-lives.c (make_hard_regno_born, make_hard_regno_dead)
9137 (process_bb_lives): Handle hard_regs_spilled_into.
9138 (lra_create_live_ranges_1): Before doing liveness propagation, clear
9139 regs in all_hard_regs_bitmap if set in hard_regs_spilled_into.
9141 2018-02-28 David Edelsohn <dje.gcc@gmail.com>
9143 * config.gcc (powerpc-ibm-aix7.1.*): New stanza.
9144 (powerpc-ibm-aix[789]*): Default to AIX 7.2.
9145 * config/rs6000/aix71.h (TARGET_DEFAULT): Revert to Power4 ISA.
9146 * config/rs6000/aix72.h: New file.
9148 2018-02-28 Jakub Jelinek <jakub@redhat.com>
9150 * gimple-ssa-warn-restrict.c (maybe_diag_overlap): Use warning_n
9151 instead of warning_at with conditional singular and plural messages
9155 * stor-layout.c (update_alignment_for_field): For
9156 targetm.ms_bitfield_layout_p (rli->t), if !is_bitfield
9157 && !DECL_PACKED (field), do the alignment update, just use
9158 only desired_align instead of MAX (type_align, desired_align)
9160 (place_field): Don't do known_align < desired_align handling
9161 early if targetm.ms_bitfield_layout_p (rli->t) and rli->prev_field
9162 is non-NULL, instead do it after rli->prev_field handling and
9163 only if not within a bitfield word. For DECL_PACKED (field)
9164 use type_align of BITS_PER_UNIT.
9166 2018-02-28 Eric Botcazou <ebotcazou@adacore.com>
9168 * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Remove
9169 superfluous parentheses and trailing spaces.
9171 2018-02-28 Richard Biener <rguenther@suse.de>
9173 PR tree-optimization/84584
9174 * graphite-scop-detection.c (scop_detection::add_scop): Discard
9175 SCoPs with fake exit edge.
9177 2018-02-28 Martin Liska <mliska@suse.cz>
9180 * timevar.c (timer::print): Fix format to properly print 100%
9183 2018-02-28 Richard Biener <rguenther@suse.de>
9186 * genmatch.c (capture_info::walk_match): Do not mark
9187 captured expressions without operands as expr_p given
9188 they act more like predicates and should be subject to
9189 "lost tail" side-effect preserving.
9191 2018-02-28 Alexandre Oliva <aoliva@redhat.com>
9193 PR rtl-optimization/81611
9194 * auto-inc-dec.c (attempt_change): Move dead note from
9195 mem_insn if it's the next use of regno
9196 (find_address): Take address use of reg holding
9197 non-incremented value. Add parm to limit search to the named
9199 (merge_in_block): Attempt to use a mem insn that is the next
9200 use of the original regno.
9202 2018-02-27 Martin Sebor <msebor@redhat.com>
9205 * gcc/doc/invoke.texi (-Wmissing-attributes): New option.
9206 * gcc/print-tree.c (print_node): Handle DECL_UNINLINABLE.
9208 2018-02-27 Martin Sebor <msebor@redhat.com>
9210 PR translation/84207
9211 * diagnostic-core.h (warning_n, error_n, inform_n): Change
9212 n argument to unsigned HOST_WIDE_INT.
9213 * diagnostic.c (warning_n, error_n, inform_n): Ditto.
9214 (diagnostic_n_impl): Ditto. Handle arguments in excess of LONG_MAX.
9215 * gimple-ssa-sprintf.c (format_directive): Simplify inform_n call.
9216 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Use warning_n.
9218 2018-02-27 Richard Biener <rguenther@suse.de>
9220 PR tree-optimization/84512
9221 * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost):
9222 Do not use the estimate returned from record_stmt_cost for
9223 the scalar iteration cost but sum properly using add_stmt_cost.
9225 2018-02-27 Richard Biener <rguenther@suse.de>
9227 PR tree-optimization/84466
9228 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
9229 Adjust last change to less strictly validate use operands.
9231 2018-02-27 Martin Liska <mliska@suse.cz>
9233 PR gcov-profile/84548
9234 * gcov.c (process_file): Allow partial overlap and consider it
9235 also as group functions.
9236 (output_lines): Properly calculate range of lines for a group.
9238 2018-02-27 Martin Liska <mliska@suse.cz>
9240 * timevar.c (timer::print_row): Remove 'usr', 'sys', 'wall' and
9241 'ggc' suffixes. Change first column width.
9242 (timer::print): Fix formatting of the column.
9244 2018-02-27 Alexandre Oliva <aoliva@redhat.com>
9246 * tree-ssa-live.c (remove_unused_scope_block_p): Do not
9247 preserve inline entry blocks for the sake of debug inline
9248 entry point markers alone.
9249 (remove_unused_locals): Suggest in comments a better place to
9250 force the preservation of inline entry blocks that are
9251 otherwise unused, but do not preserve them.
9253 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9255 * config/i386/i386.c (ix86_output_indirect_jmp): Update comments.
9257 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9260 * config/i386/constraints.md (Bs): Replace
9261 ix86_indirect_branch_register with
9262 TARGET_INDIRECT_BRANCH_REGISTER.
9264 * config/i386/i386.md (indirect_jump): Likewise.
9265 (tablejump): Likewise.
9266 (*sibcall_memory): Likewise.
9267 (*sibcall_value_memory): Likewise.
9268 Peepholes of indirect call and jump via memory: Likewise.
9269 (*sibcall_GOT_32): Disallowed for TARGET_INDIRECT_BRANCH_REGISTER.
9270 (*sibcall_value_GOT_32): Likewise.
9271 * config/i386/predicates.md (indirect_branch_operand): Likewise.
9272 (GOT_memory_operand): Likewise.
9273 (call_insn_operand): Likewise.
9274 (sibcall_insn_operand): Likewise.
9275 (GOT32_symbol_operand): Likewise.
9276 * config/i386/i386.h (TARGET_INDIRECT_BRANCH_REGISTER): New.
9278 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
9280 PR rtl-optimization/83496
9281 * reorg.c (steal_delay_list_from_target): Change REDUNDANT array from
9282 booleans to RTXes. Call fix_reg_dead_note on every non-null element.
9283 (steal_delay_list_from_fallthrough): Call fix_reg_dead_note on a
9284 redundant insn, if any.
9285 (relax_delay_slots): Likewise.
9286 (update_reg_unused_notes): Rename REDUNDANT_INSN to OTHER_INSN.
9288 2018-02-26 Richard Sandiford <richard.sandiford@linaro.org>
9290 PR tree-optimization/83965
9291 * tree-vect-patterns.c (vect_reassociating_reduction_p): Assume
9292 that grouped statements are part of a reduction chain. Return
9293 true if the statement is not marked as a reduction itself but
9295 (vect_recog_dot_prod_pattern): Don't check whether the statement
9296 is part of a group here.
9297 (vect_recog_sad_pattern): Likewise.
9298 (vect_recog_widen_sum_pattern): Likewise.
9300 2018-02-26 Eric Botcazou <ebotcazou@adacore.com>
9303 * final.c (rest_of_clean_state): Also look for calls inside sequences.
9305 2018-02-26 H.J. Lu <hongjiu.lu@intel.com>
9308 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Remove
9310 (ix86_output_indirect_function_return): New prototype.
9311 (ix86_split_simple_return_pop_internal): Likewise.
9312 * config/i386/i386.c (indirect_return_via_cx): New.
9313 (indirect_return_via_cx_bnd): Likewise.
9314 (indirect_thunk_name): Handle return va CX_REG.
9315 (output_indirect_thunk_function): Create alias for
9316 __x86_return_thunk_[re]cx and __x86_return_thunk_[re]cx_bnd.
9317 (ix86_output_indirect_jmp): Remove the bool argument.
9318 (ix86_output_indirect_function_return): New function.
9319 (ix86_split_simple_return_pop_internal): Likewise.
9320 * config/i386/i386.md (*indirect_jump): Don't pass false
9321 to ix86_output_indirect_jmp.
9322 (*tablejump_1): Likewise.
9323 (simple_return_pop_internal): Change it to define_insn_and_split.
9324 Call ix86_split_simple_return_pop_internal to split it for
9326 (simple_return_indirect_internal): Call
9327 ix86_output_indirect_function_return instead of
9328 ix86_output_indirect_jmp.
9330 2018-02-26 Jakub Jelinek <jakub@redhat.com>
9333 * vec.h (vec_default_construct): For BROKEN_VALUE_INITIALIZATION use
9334 memset and value initialization afterwards.
9336 2018-02-26 Christophe Lyon <christophe.lyon@linaro.org>
9338 * Makefile.in (lto-wrapper): Use ALL_LINKERFLAGS.
9340 2018-02-26 Ramana Radhakrishnan <ramana.radhakrishnan@arm.com>
9343 * common/config/aarch64/aarch64-common.c
9344 (aarch_option_optimization_table[]): Switch
9345 off fomit-frame-pointer
9347 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
9348 Chung-Ju Wu <jasonwucj@gmail.com>
9350 * config/nds32/nds32-multiple.md (load_multiple): Disallow
9352 (store_multiple): Ditto.
9354 2018-02-26 Kito Cheng <kito.cheng@gmail.com>
9356 * config.gcc: Add --with-cpu support for nds32 target.
9357 * config/nds32/nds32-opts.h (nds32_cpu_type): New.
9358 * config/nds32/nds32.opt: Add -mcpu= option.
9360 2018-02-25 Segher Boessenkool <segher@kernel.crashing.org>
9362 * config/rs6000/rs6000.opt (mvrsave=no, mvrsave=yes, isel=no,
9363 isel=yes): Warn for these deprecated options.
9365 2018-02-23 David Edelsohn <dje.gcc@gmail.com>
9367 * config/rs6000/aix71.h (TARGET_DEFAULT): Change to
9368 ISA_2_5_MASKS_EMBEDDED.
9370 2018-02-23 Jakub Jelinek <jakub@redhat.com>
9372 * ipa-prop.c (ipa_vr_ggc_hash_traits::hash): Hash p->min and
9373 p->max as pointers rather than using iterative_hash_expr.
9375 2018-02-23 Carl Love <cel@us.ibm.com>
9377 * config/rs6000/rs6000-builtin.def: Change VSIGNED2 and VUNSIGNED2
9378 macro expansions from BU_VSX_2 to BU_P8V_VSX_2 and BU_VSX_OVERLOAD_2 to
9380 * config/rs6000/rs6000-c.c: Change VSX_BUILTIN_VEC_VSIGNED2 to
9381 P8V_BUILTIN_VEC_VSIGNED2. Change VSX_BUILTIN_VEC_VUNSIGNED2 to
9382 P8V_BUILTIN_VEC_VUNSIGNED2.
9384 2018-02-22 Vladimir Makarov <vmakarov@redhat.com>
9387 * lra-int.h (LRA_UNKNOWN_ALT, LRA_NON_CLOBBERED_ALT): New macros.
9388 * lra.c (lra_set_insn_recog_data, lra_update_insn_recog_data): Use
9390 * lra-constraints.c (curr_insn_transform): Set up
9391 LRA_NON_CLOBBERED_ALT for moves processed on the fast path. Use
9393 (remove_inheritance_pseudos): Use LRA_UNKNOWN_ALT.
9394 * lra-eliminations.c (spill_pseudos): Ditto.
9395 (process_insn_for_elimination): Ditto.
9396 * lra-lives.c (reg_early_clobber_p): Use the new macros.
9397 * lra-spills.c (spill_pseudos): Use LRA_UNKNOWN_ALT and
9398 LRA_NON_CLOBBERED_ALT.
9400 2018-02-22 Martin Sebor <msebor@redhat.com>
9402 PR tree-optimization/84480
9403 * gimple-fold.c (gimple_fold_builtin_strcpy): Move warnings
9404 to maybe_diag_stxncpy_trunc. Call it.
9405 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Integrate warnings
9406 from gimple_fold_builtin_strcpy. Print inlining stack.
9407 (handle_builtin_stxncpy): Print inlining stack.
9408 * tree-ssa-strlen.h (maybe_diag_stxncpy_trunc): Declare.
9410 2018-02-22 H.J. Lu <hongjiu.lu@intel.com>
9413 * config/i386/i386.c (ix86_set_indirect_branch_type): Issue an
9414 error when -mindirect-branch=thunk-extern, -fcf-protection=branch
9415 and -fcheck-pointer-bounds are used together.
9416 (indirect_thunk_prefix): New enum.
9417 (indirect_thunk_need_prefix): New function.
9418 (indirect_thunk_name): Replace need_bnd_p with need_prefix. Use
9419 "_nt" instead of "_bnd" for NOTRACK prefix.
9420 (output_indirect_thunk): Replace need_bnd_p with need_prefix.
9421 (output_indirect_thunk_function): Likewise.
9423 (ix86_code_end): Update output_indirect_thunk_function calls.
9424 (ix86_output_indirect_branch_via_reg): Replace
9425 ix86_bnd_prefixed_insn_p with indirect_thunk_need_prefix.
9426 (ix86_output_indirect_branch_via_push): Likewise.
9427 (ix86_output_function_return): Likewise.
9428 * doc/invoke.texi: Document -mindirect-branch=thunk-extern is
9429 incompatible with -fcf-protection=branch and
9430 -fcheck-pointer-bounds.
9432 2018-02-22 Steve Ellcey <sellcey@cavium.com>
9435 * config/aarch64/aarch64.c (aarch64_print_address_internal):
9436 Change gcc_assert call to output_operand_lossage.
9438 2018-02-22 Steve Ellcey <sellcey@cavium.com>
9440 * doc/extend.texi (__builtin_extend_pointer): Document builtin.
9442 2018-02-22 DJ Delorie <dj@redhat.com>
9443 Sebastian Perta <sebastian.perta@renesas.com>
9444 Oleg Endo <olegendo@gcc.gnu.org>
9446 * config/rx/rx.c (rx_rtx_costs): New function.
9447 (TARGET_RTX_COSTS): Override to use rx_rtx_costs.
9449 2018-02-22 Thomas Preud'homme <thomas.preudhomme@arm.com>
9451 * config/arm/t-multilib: Map Armv8-R to Armv7 multilibs.
9453 2018-02-22 Martin Liska <mliska@suse.cz>
9456 * common/config/arm/arm-common.c (arm_print_hint_for_cpu_option):
9457 Add "native" as a possible value.
9459 2018-02-22 Martin Liska <mliska@suse.cz>
9462 * config/i386/i386.c (ix86_option_override_internal):
9463 Add "native" as a possible value for -march and -mtune.
9465 2018-02-22 Jakub Jelinek <jakub@redhat.com>
9468 * stor-layout.c (finalize_type_size): Propagate TYPE_EMPTY_P flag
9469 to all type variants.
9471 PR tree-optimization/84503
9472 * gimple-ssa-store-merging.c (merged_store_group::merge_into): Compute
9473 width as info->bitpos + info->bitsize - start.
9474 (merged_store_group::merge_overlapping): Simplify width computation.
9475 (check_no_overlap): New function.
9476 (imm_store_chain_info::try_coalesce_bswap): Compute expected
9477 start + width and last_order of the group, fail if check_no_overlap
9479 (imm_store_chain_info::coalesce_immediate_stores): Don't merge info
9480 to group if check_no_overlap fails.
9482 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
9484 * config/rs6000/altivec.md: Delete contraint arguments to
9485 define_expand, define_split, and define_peephole2, and in
9486 define_insn_and_split if always unused.
9487 * config/rs6000/darwin.md: Ditto.
9488 * config/rs6000/dfp.md: Ditto.
9489 * config/rs6000/rs6000.md: Ditto.
9490 * config/rs6000/sync.md: Ditto.
9491 * config/rs6000/vector.md: Ditto.
9492 * config/rs6000/vsx.md: Ditto.
9494 2018-02-21 Segher Boessenkool <segher@kernel.crashing.org>
9496 * config/rs6000/altivec.md: Write output control strings as braced
9497 blocks instead of double-quoted strings.
9498 * config/rs6000/darwin.md: Ditto.
9499 * config/rs6000/rs6000.md: Ditto.
9500 * config/rs6000/vector.md: Ditto.
9501 * config/rs6000/vsx.md: Ditto.
9503 2018-02-21 Jason Merrill <jason@redhat.com>
9505 PR c++/84314 - ICE with templates and fastcall attribute.
9506 * attribs.c (build_type_attribute_qual_variant): Remove assert.
9508 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
9510 * ipa-cp.c (determine_versionability): Fix comment typos.
9512 2018-02-21 Jan Hubicka <hubicka@ucw.cz>
9515 * ipa-cp.c (determine_versionability): Do not version functions caling
9518 2018-02-21 Martin Liska <mliska@suse.cz>
9521 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch):
9522 Add "native" as a possible value.
9523 * config/aarch64/aarch64.h (HAVE_LOCAL_CPU_DETECT): Define
9524 the macro when native cpu detection is available.
9526 2018-02-21 Martin Liska <mliska@suse.cz>
9529 * common/config/arm/arm-common.c (arm_print_hint_for_arch_option):
9530 Add "native" as a possible value.
9531 * config/arm/arm.h (HAVE_LOCAL_CPU_DETECT): Define the macro
9532 when native cpu detection is available.
9534 2018-02-21 Jakub Jelinek <jakub@redhat.com>
9535 Martin Sebor <msebor@redhat.com>
9537 PR tree-optimization/84478
9538 * gimple-fold.h (get_range_strlen): Add a bool argument defaulted to
9540 * gimple-fold.c (get_range_strlen): Make minlen const and assume it
9541 can't be NULL. Change FUZZY from bool to int, for 1 add PHI/COND_EXPR
9542 support which is conservatively correct, for 2 only stay conservative
9543 for maxlen. Formatting and comment capitalization fixes. Add STRICT
9544 argument to the 2 argument get_range_strlen, adjust 6 arg
9545 get_range_strlen caller and clear minmaxlen[0] and [1] if it returned
9547 (get_maxval_strlen): Adjust 6 arg get_range_strlen caller.
9548 (gimple_fold_builtin_strlen): Pass true as last argument to
9551 2018-02-20 Martin Sebor <msebor@redhat.com>
9554 * gimple-ssa-warn-restrict.c (builtin_memref::extend_offset_range): New.
9555 (builtin_memref::set_base_and_offset): Same. Handle inner references.
9556 (builtin_memref::builtin_memref): Factor out parts into
9557 set_base_and_offset and call it.
9559 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
9562 * optabs-query.c (find_widening_optab_handler_and_mode): If from_mode
9563 is a scalar_int_mode, assert that to_mode is a scalar_int_mode with
9564 greater precision. If to_mode is a MODE_PARTIAL_INT, stop the
9565 search at the associated MODE_INT.
9567 2018-02-20 Jeff Law <law@redhat.com>
9570 PR tree-optimization/81592
9572 * gimple-ssa-sprintf.c (format_integer): Query EVRP range analyzer
9573 for range data rather than using global data.
9574 * gimple-ssa-sprintf.c (get_int_range): Query EVRP range analyzer for
9575 range data rather than using global data.
9576 * gimple-ssa-sprintf.c (get_int_range): Accept vr_values parameter
9577 pass it to children as needed.
9578 (struct directive::fmtresult): Similarly.
9579 (struct directive::set_width): Similarly.
9580 (struct directive::set_precision): Similarly.
9581 (format_integer, format_directive, parse_directive): Similarly.
9582 (format_none): Accept unnamed vr_values parameter.
9583 (format_percent, format_floating, format_character): Similarly.
9584 (format_string, format_plain): Similarly.
9585 * gimple-ssa-sprintf.c (sprintf_dom_walker::handle_gimple_call): Query
9586 the EVRP range analyzer for range data rather than using global data.
9587 * gimple-ssa-sprintf.c: Include alloc-pool.h, vr-values.h and
9588 gimple-ssa-evrp-analyze.h
9589 (class sprintf_dom_walker): Add after_dom_children member function.
9590 Add evrp_range_analyzer member.
9591 (sprintf_dom_walker::before_dom_children): Call into the EVRP
9592 range analyzer as needed.
9593 (sprintf_dom_walker::after_dom_children): New member function.
9594 * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::enter): Do nothing
9596 (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
9597 (evrp_range_analyzer::pop_to_marker): Likewise.
9599 2018-02-20 Richard Sandiford <richard.sandiford@linaro.org>
9601 PR tree-optimization/84419
9602 * internal-fn.c (expand_call_mem_ref): Create a TARGET_MEM_REF
9603 with the required type if its current type is compatible but
9606 2018-02-20 Jakub Jelinek <jakub@redhat.com>
9609 * match.pd (pow(C,x) -> exp(log(C)*x)): Delay all folding until
9610 after vectorization.
9612 2018-02-20 Martin Liska <mliska@suse.cz>
9615 * config/aarch64/aarch64.c (aarch64_print_hint_for_core_or_arch): Print
9616 possible values if we don't have a hint.
9618 2018-02-20 Martin Liska <mliska@suse.cz>
9622 * final.c (shorten_branches): Build align_tab array with one
9624 * opts.c (finish_options): Add alignment option limit check.
9625 (MAX_CODE_ALIGN): Likewise.
9626 (MAX_CODE_ALIGN_VALUE): Likewise.
9627 * doc/invoke.texi: Document maximum allowed option value for
9628 all -falign-* options.
9630 2018-02-19 Jakub Jelinek <jakub@redhat.com>
9633 * reg-notes.def (REG_CALL_ARG_LOCATION): New reg note.
9634 * insn-notes.def (NOTE_INSN_CALL_ARG_LOCATION): Remove.
9635 * var-tracking.c (emit_note_insn_var_location): Remove all references
9636 to NOTE_INSN_CALL_ARG_LOCATION.
9637 (emit_notes_in_bb): Emit arguments as REG_CALL_ARG_LOCATION note on
9638 the CALL_INSN rather than separate NOTE_INSN_CALL_ARG_LOCATION note.
9639 Use copy_rtx_if_shared.
9640 * dwarf2out.c (gen_subprogram_die): Use XEXP with 0 instead of
9641 NOTE_VAR_LOCATION on ca_loc->call_arg_loc_note.
9642 (dwarf2out_var_location): Remove handling of
9643 NOTE_INSN_CALL_ARG_LOCATION, instead handle REG_CALL_ARG_LOCATION note
9645 * final.c (final_scan_insn): Remove all references to
9646 NOTE_INSN_CALL_ARG_LOCATION.
9647 (rest_of_clean_state): Likewise. Remove REG_CALL_ARG_LOCATION notes
9648 before dumping final insns.
9649 * except.c (emit_note_eh_region_end): Remove all references to
9650 NOTE_INSN_CALL_ARG_LOCATION.
9651 * config/alpha/alpha.c (alpha_pad_function_end): Likewise.
9652 * config/c6x/c6x.c (c6x_gen_bundles): Likewise.
9653 * config/arc/arc.c (hwloop_optimize): Likewise.
9654 * config/arm/arm.c (create_fix_barrier): Likewise.
9655 * config/s390/s390.c (s390_chunkify_start): Likewise.
9656 * config/sh/sh.c (find_barrier): Likewise.
9657 * config/i386/i386.c (rest_of_insert_endbranch,
9658 ix86_seh_fixup_eh_fallthru): Likewise.
9659 * config/xtensa/xtensa.c (hwloop_optimize): Likewise.
9660 * config/iq2000/iq2000.c (final_prescan_insn): Likewise.
9661 * config/frv/frv.c (frv_function_prologue): Likewise.
9662 * emit-rtl.c (try_split): Likewise. Copy over REG_CALL_ARG_LOCATION
9664 (note_outside_basic_block_p): Remove all references to
9665 NOTE_INSN_CALL_ARG_LOCATION.
9666 * gengtype.c (adjust_field_rtx_def): Likewise.
9667 * print-rtl.c (rtx_writer::print_rtx_operand_code_0, print_insn):
9669 * jump.c (cleanup_barriers, delete_related_insns): Likewise.
9670 * cfgrtl.c (force_nonfallthru_and_redirect): Likewise.
9673 * builtins.c (builtin_mathfn_code): Don't check if CALL_EXPR_FN (t)
9676 PR tree-optimization/84452
9677 * tree-vect-patterns.c (vect_recog_pow_pattern): Don't call
9678 expand_simd_clones if targetm.simd_clone.compute_vecsize_and_simdlen
9681 2018-02-19 Martin Liska <mliska@suse.cz>
9684 * passes.def: Put pass_sancov_O0 before pass_lower_switch with -O0.
9686 2018-02-19 Martin Liska <mliska@suse.cz>
9687 Richard Sandiford <richard.sandiford@linaro.org>
9689 PR tree-optimization/82491
9690 * gimple-fold.c (get_base_constructor): Make earlier bail out
9693 2018-02-19 Carl Love <cel@us.ibm.com>
9695 * config/rs6000/rs6000-builtin.def: Change NEG macro expansions from
9696 BU_ALTIVEC_A to BU_P8V_AV_1 and BU_ALTIVEC_OVERLOAD_1 to
9698 * config/rs6000/rs6000-c.c: Change ALTIVEC_BUILTIN_VEC_NEG to
9699 P8V_BUILTIN_VEC_NEG.
9701 2018-02-19 Sebastian Perta <sebastian.perta@renesas.com>
9703 * config/rl78/rl78.md (movdf): New define expand.
9705 2018-02-19 Martin Liska <mliska@suse.cz>
9708 * doc/invoke.texi: Fix typo.
9709 * params.def (PARAM_MAX_LOOP_HEADER_INSNS): Likewise.
9711 2018-02-18 Segher Boessenkool <segher@kernel.crashing.org>
9713 * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
9714 handle rs6000_single_float and rs6000_double_float specially for
9717 2018-02-16 Jeff Law <law@redhat.com>
9719 * config/rx/rx.c (add_pop_cfi_notes): New function.;
9722 2018-02-16 Jakub Jelinek <jakub@redhat.com>
9725 * ipa-inline.c (inline_small_functions): Fix a typo.
9727 2018-02-16 Nathan Sidwell <nathan@acm.org>
9729 * doc/extend.texi (Backwards Compatibility): Americanize 'behaviour'.
9731 2018-02-16 Carl Love <cel@us.ibm.com>
9733 * config/rs6000/rs6000-builtin.def: Add BU_P8V_VSX_2 macro definition.
9734 Change FLOAT2 expansions from BU_VSX_2 to BU_P8V_VSX_2 and
9735 from BU_VSX_OVERLOAD_2 to BU_P8V_OVERLOAD_2.
9736 * config/rs6000/rs6000-c.c: Changed macro VSX_BUILTIN_VEC_FLOAT2
9737 expansion to P8V_BUILTIN_VEC_FLOAT2.
9739 2018-02-16 Vladimir Makarov <vmakarov@redhat.com>
9741 PR rtl-optimization/70023
9742 * lra-constraints.c (inherit_in_ebb): Take hard reg mode of
9743 src_regno into account.
9745 2018-02-16 Carl Love <cel@us.ibm.com>
9747 * config/rs6000/altivec.h: Remove vec_vextract4b and vec_vinsert4b.
9748 * config/rs6000/rs6000-builtin.def: Remove macro expansion for
9749 VEXTRACT4B, VINSERT4B, VINSERT4B_DI and VEXTRACT4B.
9750 * config/rs6000/rs6000.c: Remove case statements for
9751 P9V_BUILTIN_VEXTRACT4B, P9V_BUILTIN_VEC_VEXTRACT4B,
9752 P9V_BUILTIN_VINSERT4B, P9V_BUILTIN_VINSERT4B_DI,
9753 and P9V_BUILTIN_VEC_VINSERT4B.
9754 * config/rs6000/rs6000-c.c (altivec_expand_builtin): Remove entries for
9755 P9V_BUILTIN_VEC_VEXTRACT4B and P9V_BUILTIN_VEC_VINSERT4B.
9756 * config/rs6000/vsx.md:
9757 * doc/extend.texi: Remove vec_vextract4b, non ABI definitions for
9760 2018-02-16 Carl Love <cel@us.ibm.com>
9762 * config/rs6000/altivec.h: Add builtin names vec_extract4b
9764 * config/rs6000/rs6000-builtin.def: Add INSERT4B and EXTRACT4B
9766 * config/rs6000/rs6000-c.c: Add the definitions for
9767 P9V_BUILTIN_VEC_EXTRACT4B and P9V_BUILTIN_VEC_INSERT4B.
9768 * config/rs6000/rs6000.c (altivec_expand_builtin): Add
9769 P9V_BUILTIN_EXTRACT4B and P9V_BUILTIN_INSERT4B case statements.
9770 * config/rs6000/vsx.md: Add define_insn extract4b. Add define_expand
9771 definition for insert4b and define insn *insert3b_internal.
9772 * doc/extend.texi: Add documentation for vec_extract4b.
9774 2018-02-16 Nathan Sidwell <nathan@acm.org>
9776 * doc/extend.texi (Backwards Compatibility): Mention friend
9777 injection. Note for-scope is deprecated.
9778 * doc/invoke.texi (-ffriend-injection): Deprecate.
9780 2018-02-16 Segher Boessenkool <segher@kernel.crashing.org>
9782 * combine.c (try_combine): When adjusting LOG_LINKS for the destination
9783 that moved to I2, also allow destinations that are a paradoxical
9784 subreg (instead of a normal reg).
9786 2018-02-16 Oleg Endo <olegendo@gcc.gnu.org>
9789 * config/rx/rx.c (rx_fuse_in_memory_bitop): Convert shift operand
9792 2018-02-16 Richard Biener <rguenther@suse.de>
9794 PR tree-optimization/84037
9795 PR tree-optimization/84016
9797 * config/i386/i386.c (ix86_builtin_vectorization_cost):
9798 Adjust vec_construct for the fact we need additional higher latency
9799 128bit inserts for AVX256 and AVX512 vector builds.
9800 (ix86_add_stmt_cost): Scale vector construction cost for
9803 2018-02-16 Richard Biener <rguenther@suse.de>
9805 PR tree-optimization/84417
9806 * tree-ssa.c (non_rewritable_mem_ref_base): Properly constrain
9807 the MEM_REF offset when conversion to BIT_FIELD_REF is desired.
9808 (non_rewritable_lvalue_p): Likewise, use poly-ints.
9810 2018-02-16 Martin Liska <mliska@suse.cz>
9813 * internal-fn.def (ASAN_CHECK): Set proper flags.
9814 (ASAN_MARK): Likewise.
9816 2018-02-16 Julia Koval <julia.koval@intel.com>
9818 * config/i386/i386.c (ix86_option_override_internal): Remove PTA_CLWB
9819 from PTA_CANNONLAKE.
9821 2018-02-16 Jakub Jelinek <jakub@redhat.com>
9824 * config/aarch64/cortex-a57-fma-steering.c (fma_forest::merge_forest):
9825 Use ++iter rather than iter++ for std::list iterators.
9826 (func_fma_steering::dfs): Likewise. Don't delete nodes right away,
9827 defer deleting them until all nodes in the forest are processed. Do
9828 free even leaf nodes. Change to_process into auto_vec.
9831 * system.h (BROKEN_VALUE_INITIALIZATION): Define for GCC < 4.3.
9832 * vec.h (vec_default_construct): Use memset instead of placement new
9833 if BROKEN_VALUE_INITIALIZATION is defined.
9834 * hash-table.h (hash_table<Descriptor, Allocator>::empty_slow): Use
9835 memset instead of value initialization if BROKEN_VALUE_INITIALIZATION
9838 PR rtl-optimization/83723
9839 * lra-int.h (lra_substitute_pseudo): Add DEBUG_P argument.
9840 * lra.c (lra_substitute_pseudo): Likewise. If true, use
9841 gen_rtx_raw_SUBREG instead of gen_rtx_SUBREG. Pass DEBUG_P to
9843 (lra_substitute_pseudo_within_insn): Adjust lra_substitute_pseudo
9845 * lra-constraints.c (inherit_reload_reg, split_reg): Likewise.
9847 2018-02-16 Eric Botcazou <ebotcazou@adacore.com>
9849 PR rtl-optimization/81443
9850 * rtlanal.c (num_sign_bit_copies1) <SUBREG>: Do not propagate results
9851 from inner REGs to paradoxical SUBREGs.
9853 2018-02-16 Richard Biener <rguenther@suse.de>
9855 PR tree-optimization/84399
9856 * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
9857 For operands we can analyze at their definition make sure we can
9858 analyze them at each use as well.
9860 2018-02-16 Richard Biener <rguenther@suse.de>
9862 PR tree-optimization/84190
9863 * tree-ssa.c (non_rewritable_mem_ref_base): Do not touch
9864 volatile accesses if the decl isn't volatile.
9866 2018-02-15 Jason Merrill <jason@redhat.com>
9868 PR c++/84314 - ICE with templates and fastcall attribute.
9869 * attribs.c (build_type_attribute_qual_variant): Don't clobber
9870 TYPE_CANONICAL on an existing type.
9872 2018-02-15 Jakub Jelinek <jakub@redhat.com>
9874 PR tree-optimization/84383
9875 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Don't look at
9876 dstoff nor call operand_equal_p if dstbase is NULL.
9878 PR tree-optimization/84334
9879 * match.pd ((A +- CST1) +- CST2 -> A + CST3): If A is
9880 also a CONSTANT_CLASS_P, punt.
9882 2018-02-14 Jim Wilson <jimw@sifive.com>
9884 * config/riscv/riscv.c (riscv_first_stack_step): Move locals after
9885 first SMALL_OPERAND check. New local min_second_step. Move assert
9886 to where locals are set. Add TARGET_RVC support.
9887 * config/riscv/riscv.h (C_SxSP_BITS, SWSP_REACH, SDSP_REACH): New.
9889 2018-02-14 Indu Bhagat <indu.bhagat@oracle.com>
9891 * doc/invoke.texi: Correct -Wformat-overflow code sample.
9893 2018-02-14 Martin Sebor <msebor@redhat.com>
9895 PR tree-optimization/83698
9896 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For
9897 arrays constrain the offset range to their bounds.
9898 (builtin_access::strcat_overlap): Adjust the bounds of overlap offset.
9899 (builtin_access::overlap): Avoid setting the size of overlap if it's
9901 (maybe_diag_overlap): Also consider arrays when deciding what values
9902 of offsets to include in diagnostics.
9904 2018-02-14 Martin Sebor <msebor@redhat.com>
9907 * attribs.c (diag_attr_exclusions): Consider the exclusion(s)
9908 that correspond to the kind of a declaration.
9910 2018-02-14 John David Anglin <danglin@gcc.gnu.org>
9913 * config/pa/pa.md: Load address of PIC label using the linkage table
9914 if the label is nonlocal.
9916 2018-02-14 Kelvin Nilsen <kelvin@gcc.gnu.org>
9918 * config/rs6000/rs6000.c (rs6000_option_override_internal): Issue
9919 warning message if user requests -maltivec=be.
9920 * doc/invoke.texi: Document deprecation of -maltivec=be.
9922 2018-02-14 Will Schmidt <will_schmidt@vnet.ibm.com>
9925 * config/rs6000/rs6000-c.c: Update definitions for
9926 ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SLDW,
9927 VEC_XXSLDWI and ALTIVEC_BUILTIN_VEC_XXPERMDI builtins.
9929 2018-02-14 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
9932 * config/i386/cetintrin.h: Remove _rdssp[d|q] and
9933 add _get_ssp intrinsics. Remove argument from
9934 __builtin_ia32_rdssp[d|q].
9935 * config/i386/i386-builtin-types.def: Add UINT_FTYPE_VOID.
9936 * config/i386/i386-builtin.def: Remove argument from
9937 __builtin_ia32_rdssp[d|q].
9938 * config/i386/i386.c: Use UINT_FTYPE_VOID. Use
9939 ix86_expand_special_args_builtin for _rdssp[d|q].
9940 * config/i386/i386.md: Remove argument from rdssp[si|di] insn.
9941 Clear register before usage.
9942 * doc/extend.texi: Remove argument from __builtin_ia32_rdssp[d|q].
9943 Add documentation for new _get_ssp and _inc_ssp intrinsics.
9945 2018-02-14 Richard Sandiford <richard.sandiford@linaro.org>
9947 PR tree-optimization/84357
9948 * tree-data-ref.c (object_address_invariant_in_loop_p): Check
9949 operand 1 of an ARRAY_REF too.
9951 2018-02-14 Oleg Endo <olegendo@gcc.gnu.org>
9954 * config/rx/rx-protos.h (rx_reg_dead_or_unused_after_insn,
9955 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
9957 (set_of_reg): New struct.
9958 (rx_find_set_of_reg, rx_find_use_of_reg): New functions.
9959 * config/rx/rx.c (rx_reg_dead_or_unused_after_insn,
9960 rx_copy_reg_dead_or_unused_notes, rx_fuse_in_memory_bitop): New
9962 * config/rx/rx.md (andsi3, iorsi3, xorsi3): Convert to insn_and_split.
9963 Split into bitclr, bitset, bitinvert patterns if appropriate.
9964 (*bitset, *bitinvert, *bitclr): Convert to named insn_and_split and
9965 use rx_fuse_in_memory_bitop.
9966 (*bitset_in_memory, *bitinvert_in_memory, *bitclr_in_memory): Convert
9967 to named insn, correct maximum insn length.
9969 2018-02-14 Jozef Lawrynowicz <jozefl.gcc@gmail.com>
9972 * machmode.def: Define a complex mode for PARTIAL_INT.
9973 * genmodes.c (complex_class): Return MODE_COMPLEX_INT for
9975 * doc/rtl.texi: Document CSPImode.
9976 * config/msp430/msp430.c (msp430_hard_regno_nregs): Add CPSImode
9978 (msp430_hard_regno_nregs_with_padding): Likewise.
9980 2018-02-13 Peter Bergner <bergner@vnet.ibm.com>
9983 * config/rs6000/rs6000.c (mem_operand_gpr): Disallow altivec addresses.
9985 2018-02-13 Segher Boessenkool <segher@kernel.crashing.org>
9987 PR rtl-optimization/84169
9988 * combine.c (try_combine): New variable split_i2i3. Set it to true if
9989 we generated a parallel as new i3 and we split that to new i2 and i3
9990 instructions. Handle split_i2i3 similar to swap_i2i3: scan the
9991 LOG_LINKs of i3 to see which of those need to link to i2 now. Link
9992 those to i2, not i1. Partially rewrite this scan code.
9994 2018-02-13 Jakub Jelinek <jakub@redhat.com>
9997 * stor-layout.c (place_field): For variable length fields, adjust
9998 offset_align afterwards not just based on the field's alignment,
9999 but also on the size.
10001 PR middle-end/84309
10002 * match.pd (pow(C,x) -> exp(log(C)*x)): Use exp2s and log2s instead
10003 of exps and logs in the use_exp2 case.
10005 2018-02-13 Jeff Law <law@redhat.com>
10007 * config/rl/rl78.c (rl78_attribute_table): Fix terminator and
10008 entry for "vector".
10010 * config/rl78/rl78.c (rl78_handle_func_attribute): Mark
10013 2018-02-13 Alexandre Oliva <aoliva@redhat.com>
10017 * common.opt (gas-loc-support, gas-locview-support): New.
10018 (ginline-points, ginternal-reset-location-views): New.
10019 * doc/invoke.texi: Document them. Use @itemx where intended.
10020 (gvariable-location-views): Adjust.
10021 * target.def (reset_location_view): New.
10022 * doc/tm.texi.in (DWARF2_ASM_VIEW_DEBUG_INFO): New.
10023 (TARGET_RESET_LOCATION_VIEW): New.
10024 * doc/tm.texi: Rebuilt.
10025 * dwarf2out.c (dwarf2out_default_as_loc_support): New.
10026 (dwarf2out_default_as_locview_support): New.
10027 (output_asm_line_debug_info): Use option variables.
10028 (dwarf2out_maybe_output_loclist_view_pair): Likewise.
10029 (output_loc_list): Likewise.
10030 (add_high_low_attributes): Check option variables.
10031 Don't output entry view attribute in strict mode.
10032 (gen_inlined_subroutine_die): Check option variables.
10033 (dwarf2out_inline_entry): Likewise.
10034 (init_sections_and_labels): Likewise.
10035 (dwarf2out_early_finish): Likewise.
10036 (maybe_reset_location_view): New, from...
10037 (dwarf2out_var_location): ... here. Call it.
10038 * debug.h (dwarf2out_default_as_loc_support): Declare.
10039 (dwarf2out_default_as_locview_support): Declare.
10040 * hooks.c (hook_int_rtx_insn_0): New.
10041 * hooks.h (hook_int_rtx_insn_0): Declare.
10042 * toplev.c (process_options): Take -gas-loc-support and
10043 -gas-locview-support from dwarf2out. Enable
10044 -gvariable-location-views by default only with locview
10045 assembler support. Enable -ginternal-reset-location-views by
10046 default only if the target defines the corresponding hook.
10047 Enable -ginline-points by default if location views are
10048 enabled; force it disabled if statement frontiers are
10050 * tree-inline.c (expand_call_inline): Check option variables.
10051 * tree-ssa-live.c (remove_unused_scope_block_p): Likewise.
10053 2018-02-13 Richard Sandiford <richard.sandiford@linaro.org>
10055 PR tree-optimization/84321
10056 * tree-vrp.c (intersect_range_with_nonzero_bits): Fix VR_ANTI_RANGE
10057 handling. Also check whether the anti-range contains any values
10058 that satisfy the mask; switch to a VR_RANGE if not.
10060 2018-02-13 Paolo Bonzini <bonzini@gnu.org>
10063 * internal-fn.def (ASAN_CHECK, ASAN_MARK): Revert changes to fnspec.
10065 2018-02-13 Martin Jambor <mjambor@suse.cz>
10068 * ipa-param-manipulation.c (ipa_modify_call_arguments): Use location
10069 of call statements, also set location of a load to a temporary.
10071 2018-02-13 Sebastian Perta <sebastian.perta@renesas.com>
10073 * config/rl78/rl78.c (add_vector_labels): New function.
10074 * config/rl78/rl78.c (rl78_handle_vector_attribute): New function.
10075 * config/rl78/rl78.c (rl78_start_function): Call add_vector_labels.
10076 * config/rl78/rl78.c (rl78_handle_func_attribute): Removed the assert
10077 which checks that no arguments are passed.
10078 * config/rl78/rl78.c (rl78_attribute_table): Add "vector" attribute.
10079 * doc/extend.texi: Documentation for the new attribute.
10081 2018-02-13 Andreas Schwab <schwab@suse.de>
10083 * config/riscv/linux.h (CPP_SPEC): Define.
10085 2018-02-13 Jakub Jelinek <jakub@redhat.com>
10088 * config/i386/i386.c (ix86_init_mmx_sse_builtins): Pass
10089 OPTION_MASK_ISA_AES | OPTION_MASK_ISA_SSE2 instead of
10090 OPTION_MASK_ISA_AES as first argument to def_builtin_const
10091 for AES builtins. Pass OPTION_MASK_ISA_PCLMUL | OPTION_MASK_ISA_SSE2
10092 instead of OPTION_MASK_ISA_PCLMUL as first argument to
10093 def_builtin_const for __builtin_ia32_pclmulqdq128 builtin.
10094 * config/i386/wmmintrin.h: If __SSE2__ is not defined, enable it
10095 temporarily for AES and PCLMUL builtins.
10097 PR tree-optimization/84339
10098 * gimple-fold.c (get_range_strlen): Set *FLEXP to true when handling
10099 ARRAY_REF where first operand is array_at_struct_end_p COMPONENT_REF.
10102 PR middle-end/84309
10103 * match.pd (pow(C,x) -> exp(log(C)*x)): Optimize instead into
10104 exp2(log2(C)*x) if C is a power of 2 and c99 runtime is available.
10105 * generic-match-head.c (canonicalize_math_after_vectorization_p): New
10107 * gimple-match-head.c (canonicalize_math_after_vectorization_p): New
10109 * omp-simd-clone.h: New file.
10110 * omp-simd-clone.c: Include omp-simd-clone.h.
10111 (expand_simd_clones): No longer static.
10112 * tree-vect-patterns.c: Include fold-const-call.h, attribs.h,
10113 cgraph.h and omp-simd-clone.h.
10114 (vect_recog_pow_pattern): Optimize pow(C,x) to exp(log(C)*x).
10115 (vect_recog_widen_shift_pattern): Formatting fix.
10116 (vect_pattern_recog_1): Don't check optab for calls.
10119 * config/i386/sse.md (<avx512>_vpermi2var<mode>3_mask): Force
10120 operands[2] into a REG before using gen_lowpart on it.
10122 2018-02-12 Jeff Law <law@redhat.com>
10125 * config/sh/sh.c (find_barrier): Consider a sibling call
10128 * cse.c (try_back_substitute_reg): Move any REG_ARGS_SIZE note when
10129 successfully back substituting a reg.
10131 2018-02-12 Richard Biener <rguenther@suse.de>
10133 PR tree-optimization/84037
10134 * tree-vect-slp.c (vect_analyze_slp_cost): Add visited
10135 parameter, move visited init to caller.
10136 (vect_slp_analyze_operations): Separate cost from validity
10137 check, initialize visited once for all instances.
10138 (vect_schedule_slp): Analyze map to CSE vectorized nodes once
10140 * tree-vect-stmts.c (vect_model_simple_cost): Make early
10142 (vect_model_promotion_demotion_cost): Likewise.
10143 (vectorizable_bswap): Guard cost modeling with !slp_node
10144 instead of !PURE_SLP_STMT to avoid double-counting on hybrid
10146 (vectorizable_call): Likewise.
10147 (vectorizable_conversion): Likewise.
10148 (vectorizable_assignment): Likewise.
10149 (vectorizable_shift): Likewise.
10150 (vectorizable_operation): Likewise.
10151 (vectorizable_store): Likewise.
10152 (vectorizable_load): Likewise.
10153 (vectorizable_condition): Likewise.
10154 (vectorizable_comparison): Likewise.
10156 2018-02-12 Paolo Bonzini <bonzini@gnu.org>
10159 * internal-fn.def (ASAN_CHECK): Fix fnspec to account for return value.
10160 (ASAN_MARK): Fix fnspec to account for return value, change pointer
10161 argument from 'R' to 'W' so that the pointed-to datum is clobbered.
10163 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
10165 PR middle-end/83665
10166 * params.def (inline-min-speedup): Increase from 8 to 15.
10167 (max-inline-insns-auto): Decrease from 40 to 30.
10168 * ipa-split.c (consider_split): Add some buffer for function to
10169 be considered inlining candidate.
10170 * invoke.texi (max-inline-insns-auto, inline-min-speedup): UPdate
10173 2018-02-12 Richard Biener <rguenther@suse.de>
10175 PR tree-optimization/84037
10176 * tree-vect-slp.c (vect_build_slp_tree_2): Try swapping the
10177 matched stmts if we cannot swap the non-matched ones.
10179 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10181 * config/i386/avx512fintrin.h (_mm_mask_scalef_round_sd,
10182 _mm_maskz_scalef_round_sd, _mm_mask_scalef_round_ss,
10183 _mm_maskz_scalef_round_ss): New intrinsics.
10184 (__builtin_ia32_scalefsd_round, __builtin_ia32_scalefss_round): Fix.
10185 * config/i386/i386-builtin.def (__builtin_ia32_scalefsd_round,
10186 __builtin_ia32_scalefss_round): Remove.
10187 (__builtin_ia32_scalefsd_mask_round,
10188 __builtin_ia32_scalefss_mask_round): New intrinsics.
10189 * config/i386/sse.md (vmscalef<mode><round_name>): Renamed to ...
10190 (vmscalef<mode><mask_scalar_name><round_scalar_name>): ... this.
10191 ((match_operand:VF_128 2 "<round_nimm_predicate>"
10192 "<round_constraint>")): Changed to ...
10193 ((match_operand:VF_128 2 "<round_scalar_nimm_predicate>"
10194 "<round_scalar_constraint>")): ... this.
10195 ("vscalef<ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|
10196 %0, %1, %2<round_op3>}"): Changed to ...
10197 ("vscalef<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%2, %1,
10198 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %1,
10199 %2<round_scalar_mask_op3>}"): ... this.
10200 * config/i386/subst.md (round_scalar_nimm_predicate): New.
10202 2018-02-12 Olga Makhotina <olga.makhotina@intel.com>
10204 * config/i386/avx512fintrin.h (_mm_mask_sqrt_round_sd)
10205 (_mm_maskz_sqrt_round_sd, _mm_mask_sqrt_round_ss)
10206 (_mm_maskz_sqrt_round_ss): New intrinsics.
10207 (__builtin_ia32_sqrtsd_round, __builtin_ia32_sqrtss_round): Remove.
10208 (__builtin_ia32_sqrtsd_mask_round)
10209 (__builtin_ia32_sqrtss_mask_round): New builtins.
10210 * config/i386/i386-builtin.def (__builtin_ia32_sqrtsd_round)
10211 (__builtin_ia32_sqrtss_round): Remove.
10212 (__builtin_ia32_sqrtsd_mask_round)
10213 (__builtin_ia32_sqrtss_mask_round): New builtins.
10214 * config/i386/sse.md (vmsqrt<mode>2<round_name>): Renamed to ...
10215 (vmsqrt<mode>2<mask_scalar_name><round_scalar_name>): ... this.
10216 ((match_operand:VF_128 1 "vector_operand"
10217 "xBm,<round_constraint>")): Changed to ...
10218 ((match_operand:VF_128 1 "vector_operand"
10219 "xBm,<round_scalar_constraint>")): ... this.
10220 (vsqrt<ssescalarmodesuffix>\t{<round_op3>%1, %2, %0|
10221 %0, %2, %<iptr>1<round_op3>}): Changed to ...
10222 (vsqrt<ssescalarmodesuffix>\t{<round_scalar_mask_op3>%1, %2,
10223 %0<mask_scalar_operand3>|%0<mask_scalar_operand3>, %2,
10224 %<iptr>1<round_scalar_mask_op3>}): ... this.
10225 ((set_attr "prefix" "<round_prefix>")): Changed to ...
10226 ((set_attr "prefix" "<round_scalar_prefix>")): ... this.
10228 2018-02-11 Steven Munroe <munroesj@gcc.gnu.org>
10231 * config/rs6000/mmintrin.h (_mm_cmpeq_pi32 [_ARCH_PWR9]):
10232 Cast vec_cmpeq result to correct type.
10233 * config/rs6000/mmintrin.h (_mm_cmpgt_pi32 [_ARCH_PWR9]):
10234 Cast vec_cmpgt result to correct type.
10236 2018-02-11 Alexandre Oliva <aoliva@redhat.com>
10238 * final.c (final_scan_insn_1): Renamed from...
10239 (final_scan_insn): ... this. New wrapper, to recover
10240 seen from the outermost call in recursive ones.
10241 * config/sparc/sparc.c (output_return): Drop seen from call.
10242 (output_sibcall): Likewise.
10243 * config/visium/visium.c (output_branch): Likewise.
10245 2018-02-10 John David Anglin <danglin@gcc.gnu.org>
10247 * config/pa/pa.c (hppa_profile_hook): Mark SYMBOL_REF for _mcount as
10250 2018-02-10 Alan Modra <amodra@gmail.com>
10253 * config/rs6000/rs6000.md (split_stack_return): Remove (use ..).
10254 Specify LR as an input.
10256 2018-02-10 Jakub Jelinek <jakub@redhat.com>
10259 * omp-low.c (maybe_remove_omp_member_access_dummy_vars,
10260 remove_member_access_dummy_vars): New functions.
10261 (lower_omp_for, lower_omp_taskreg, lower_omp_target,
10262 lower_omp_1, execute_lower_omp): Use them.
10264 PR rtl-optimization/84308
10265 * shrink-wrap.c (spread_components): Release todo vector.
10267 2018-02-09 Vladimir Makarov <vmakarov@redhat.com>
10269 PR rtl-optimization/57193
10270 * ira-color.c (struct allocno_color_data): Add member
10271 conflict_allocno_hard_prefs.
10272 (update_conflict_allocno_hard_prefs): New.
10273 (bucket_allocno_compare_func): Add a preference based on
10274 conflict_allocno_hard_prefs.
10275 (push_allocno_to_stack): Update conflict_allocno_hard_prefs.
10276 (color_allocnos): Remove a dead code. Initiate
10277 conflict_allocno_hard_prefs. Call update_costs_from_prefs.
10279 2018-02-09 Jakub Jelinek <jakub@redhat.com>
10282 * config/rs6000/vsx.md (p9_xxbrq_v16qi): Change input operand
10283 constraint from =wa to wa. Avoid a subreg on the output operand,
10284 instead use a pseudo and subreg it in a move.
10285 (p9_xxbrd_<mode>): Changed to ...
10286 (p9_xxbrd_v2di): ... this insn, without VSX_D iterator.
10287 (p9_xxbrd_v2df): New expander.
10288 (p9_xxbrw_<mode>): Changed to ...
10289 (p9_xxbrw_v4si): ... this insn, without VSX_W iterator.
10290 (p9_xxbrw_v4sf): New expander.
10292 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
10294 * config/rx/rx.md (movsicc): Update expander to be matched by GCC.
10296 2018-02-09 Peter Bergner <bergner@vnet.ibm.com>
10299 * config/rs6000/vsx.md (vsx_mul_v2di): Handle generating a 64-bit
10300 multiply in 32-bit mode.
10301 (vsx_div_v2di): Handle generating a 64-bit signed divide in 32-bit mode.
10302 (vsx_udiv_v2di): Handle generating a 64-bit unsigned divide in 32-bit
10305 2018-02-09 Sebastian Perta <sebastian.perta@renesas.com>
10307 * config/rx/constraints.md (CALL_OP_SYMBOL_REF): Added new constraint
10308 to allow or block "symbol_ref" depending on the value of TARGET_JSR.
10309 * config/rx/rx.md (call_internal): Use CALL_OP_SYMBOL_REF.
10310 * config/rx/rx.md (call_value_internal): Use CALL_OP_SYMBOL_REF.
10312 2018-02-09 Pierre-Marie de Rodat <derodat@adacore.com>
10315 * dwarf2out.c (is_trivial_indirect_ref): New function.
10316 (dwarf2out_late_global_decl): Do not generate a location
10317 attribute for variables that have a non-trivial DECL_VALUE_EXPR
10318 and that are not defined in the current unit.
10320 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
10322 * optabs.c (prepare_cmp_insn): Try harder to emit a direct comparison
10323 instead of a libcall for UNORDERED.
10325 2018-02-09 Tamar Christina <tamar.christina@arm.com>
10328 * config/arm/arm-c.c (arm_cpu_builtins): Un-define __ARM_FEATURE_LDREX,
10329 __ARM_ARCH_PROFILE, __ARM_ARCH_ISA_THUMB, __ARM_FP and __ARM_NEON_FP.
10331 2018-02-09 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10334 * config/s390/s390.c (s390_set_current_function): Invoke
10335 s390_indirect_branch_settings also if fndecl didn't change.
10337 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
10339 * config/rs6000/rs6000.md (blockage): Set length to zero.
10341 2018-02-09 Eric Botcazou <ebotcazou@adacore.com>
10343 * expr.c (optimize_bitfield_assignment_op): Remove obsolete assertion.
10345 2018-02-09 Jakub Jelinek <jakub@redhat.com>
10348 * gcc.c (STATIC_LIBASAN_LIBS, STATIC_LIBTSAN_LIBS,
10349 STATIC_LIBLSAN_LIBS, STATIC_LIBUBSAN_LIBS): Handle -static like
10353 * var-tracking.c (vt_add_function_parameter): Punt for non-onepart
10354 PARALLEL incoming that failed vt_get_decl_and_offset check.
10356 PR middle-end/84237
10357 * output.h (bss_initializer_p): Add NAMED argument, defaulted to false.
10358 * varasm.c (bss_initializer_p): Add NAMED argument, if true, ignore
10360 (get_variable_section): For decls in named .bss* sections pass true as
10361 second argument to bss_initializer_p.
10363 2018-02-09 Marek Polacek <polacek@redhat.com>
10364 Jakub Jelinek <jakub@redhat.com>
10367 * fold-const.c (fold_indirect_ref_1): Use VECTOR_TYPE_P macro.
10368 Formatting fixes. Verify first that tree_fits_poly_int64_p (op01).
10369 Sync some changes from cxx_fold_indirect_ref.
10371 2018-02-09 Alexandre Oliva <aoliva@redhat.com>
10373 * cfgexpand.c (expand_gimple_basic_block): Handle inline entry
10375 * dwarf2out.c (dwarf2_debug_hooks): Enable inline_entry hook.
10376 (BLOCK_INLINE_ENTRY_LABEL): New.
10377 (dwarf2out_var_location): Disregard inline entry markers.
10378 (inline_entry_data): New struct.
10379 (inline_entry_data_hasher): New hashtable type.
10380 (inline_entry_data_hasher::hash): New.
10381 (inline_entry_data_hasher::equal): New.
10382 (inline_entry_data_table): New variable.
10383 (add_high_low_attributes): Add DW_AT_entry_pc and
10384 DW_AT_GNU_entry_view attributes if a pending entry is found
10385 in inline_entry_data_table. Add old entry_pc attribute only
10386 if debug nonbinding markers are disabled.
10387 (gen_inlined_subroutine_die): Set BLOCK_DIE if nonbinding
10388 markers are enabled.
10389 (block_within_block_p, dwarf2out_inline_entry): New.
10390 (dwarf2out_finish): Check that no entries remained in
10391 inline_entry_data_table.
10392 * final.c (reemit_insn_block_notes): Handle inline entry notes.
10393 (final_scan_insn, notice_source_line): Likewise.
10394 (rest_of_clean_state): Skip inline entry markers.
10395 * gimple-pretty-print.c (dump_gimple_debug): Handle inline entry
10397 * gimple.c (gimple_build_debug_inline_entry): New.
10398 * gimple.h (enum gimple_debug_subcode): Add
10399 GIMPLE_DEBUG_INLINE_ENTRY.
10400 (gimple_build_debug_inline_entry): Declare.
10401 (gimple_debug_inline_entry_p): New.
10402 (gimple_debug_nonbind_marker_p): Adjust.
10403 * insn-notes.def (INLINE_ENTRY): New.
10404 * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
10405 inline entry marker notes.
10406 (print_insn): Likewise.
10407 * rtl.h (NOTE_MARKER_P): Add INLINE_ENTRY support.
10408 (INSN_DEBUG_MARKER_KIND): Likewise.
10409 (GEN_RTX_DEBUG_MARKER_INLINE_ENTRY_PAT): New.
10410 * tree-inline.c (expand_call_inline): Build and insert
10411 debug_inline_entry stmt.
10412 * tree-ssa-live.c (remove_unused_scope_block_p): Preserve
10413 inline entry blocks early, if nonbind markers are enabled.
10414 (dump_scope_block): Dump fragment info.
10415 * var-tracking.c (reemit_marker_as_note): Handle inline entry note.
10416 * doc/gimple.texi (gimple_debug_inline_entry_p): New.
10417 (gimple_build_debug_inline_entry): New.
10418 * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers):
10419 Enable/disable inline entry points too.
10420 * doc/rtl.texi (NOTE_INSN_INLINE_ENTRY): New.
10421 (DEBUG_INSN): Describe inline entry markers.
10423 * common.opt (gvariable-location-views): New.
10424 (gvariable-location-views=incompat5): New.
10425 * config.in: Rebuilt.
10426 * configure: Rebuilt.
10427 * configure.ac: Test assembler for view support.
10428 * dwarf2asm.c (dw2_asm_output_symname_uleb128): New.
10429 * dwarf2asm.h (dw2_asm_output_symname_uleb128): Declare.
10430 * dwarf2out.c (var_loc_view): New typedef.
10431 (struct dw_loc_list_struct): Add vl_symbol, vbegin, vend.
10432 (dwarf2out_locviews_in_attribute): New.
10433 (dwarf2out_locviews_in_loclist): New.
10434 (dw_val_equal_p): Compare val_view_list of dw_val_class_view_lists.
10435 (enum dw_line_info_opcode): Add LI_adv_address.
10436 (struct dw_line_info_table): Add view.
10437 (RESET_NEXT_VIEW, RESETTING_VIEW_P): New macros.
10438 (DWARF2_ASM_VIEW_DEBUG_INFO): Define default.
10439 (zero_view_p): New variable.
10440 (ZERO_VIEW_P): New macro.
10441 (output_asm_line_debug_info): New.
10442 (struct var_loc_node): Add view.
10443 (add_AT_view_list, AT_loc_list): New.
10444 (add_var_loc_to_decl): Add view param. Test it against last.
10445 (new_loc_list): Add view params. Record them.
10446 (AT_loc_list_ptr): Handle loc and view lists.
10447 (view_list_to_loc_list_val_node): New.
10448 (print_dw_val): Handle dw_val_class_view_list.
10449 (size_of_die): Likewise.
10450 (value_format): Likewise.
10451 (loc_list_has_views): New.
10452 (gen_llsym): Set vl_symbol too.
10453 (maybe_gen_llsym, skip_loc_list_entry): New.
10454 (dwarf2out_maybe_output_loclist_view_pair): New.
10455 (output_loc_list): Output view list or entries too.
10456 (output_view_list_offset): New.
10457 (output_die): Handle dw_val_class_view_list.
10458 (output_dwarf_version): New.
10459 (output_compilation_unit_header): Use it.
10460 (output_skeleton_debug_sections): Likewise.
10461 (output_rnglists, output_line_info): Likewise.
10462 (output_pubnames, output_aranges): Update version comments.
10463 (output_one_line_info_table): Output view numbers in asm comments.
10464 (dw_loc_list): Determine current endview, pass it to new_loc_list.
10465 Call maybe_gen_llsym.
10466 (loc_list_from_tree_1): Adjust.
10467 (add_AT_location_description): Create view list attribute if
10468 needed, check it's absent otherwise.
10469 (convert_cfa_to_fb_loc_list): Adjust.
10470 (maybe_emit_file): Call output_asm_line_debug_info for test.
10471 (dwarf2out_var_location): Reset views as needed. Precompute
10472 add_var_loc_to_decl args. Call get_attr_min_length only if we have the
10473 attribute. Set view.
10474 (new_line_info_table): Reset next view.
10475 (set_cur_line_info_table): Call output_asm_line_debug_info for test.
10476 (dwarf2out_source_line): Likewise. Output view resets and labels to
10477 the assembler, or select appropriate line info opcodes.
10478 (prune_unused_types_walk_attribs): Handle dw_val_class_view_list.
10479 (optimize_string_length): Catch it. Adjust.
10480 (resolve_addr): Copy vl_symbol along with ll_symbol. Handle
10481 dw_val_class_view_list, and remove it if no longer needed.
10482 (hash_loc_list): Hash view numbers.
10483 (loc_list_hasher::equal): Compare them.
10484 (optimize_location_lists): Check whether a view list symbol is
10485 needed, and whether the locview attribute is present, and
10486 whether they match. Remove the locview attribute if no longer
10488 (index_location_lists): Call skip_loc_list_entry for test.
10489 (dwarf2out_finish): Call output_asm_line_debug_info for test.
10490 Use output_dwarf_version.
10491 * dwarf2out.h (enum dw_val_class): Add dw_val_class_view_list.
10492 (struct dw_val_node): Add val_view_list.
10493 * final.c (SEEN_NEXT_VIEW): New.
10494 (set_next_view_needed): New.
10495 (clear_next_view_needed): New.
10496 (maybe_output_next_view): New.
10497 (final_start_function): Rename to...
10498 (final_start_function_1): ... this. Take pointer to FIRST,
10499 add SEEN parameter. Emit param bindings in the initial view.
10500 (final_start_function): Reintroduce SEEN-less interface.
10501 (final): Rename to...
10502 (final_1): ... this. Take SEEN parameter. Output final pending
10503 next view at the end.
10504 (final): Reintroduce seen-less interface.
10505 (final_scan_insn): Output pending next view before switching
10506 sections or ending a block. Mark the next view as needed when
10507 outputting variable locations. Notify debug backend of section
10508 changes, and of location view changes.
10509 (rest_of_handle_final): Adjust.
10510 * toplev.c (process_options): Autodetect value for debug variable
10511 location views option. Warn on incompat5 without -gdwarf-5.
10512 * doc/invoke.texi (gvariable-location-views): New.
10513 (gvariable-location-views=incompat5): New.
10514 (gno-variable-location-views): New.
10516 2018-02-08 David Malcolm <dmalcolm@redhat.com>
10518 PR tree-optimization/84136
10519 * tree-cfg.c (find_taken_edge_computed_goto): Remove assertion
10520 that the result of find_edge is non-NULL.
10522 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
10525 * config/i386/x86-tune-costs.h (skylake_cost): Fix cost of
10526 storing integer register in SImode. Fix cost of 256 and 512
10527 byte aligned SSE register store.
10529 2018-02-08 Sergey Shalnov <sergey.shalnov@intel.com>
10531 * config/i386/i386.c (ix86_multiplication_cost): Fix
10532 multiplication cost for TARGET_AVX512DQ.
10534 2018-02-08 Marek Polacek <polacek@redhat.com>
10536 PR tree-optimization/84238
10537 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Verify the result of
10540 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10542 PR tree-optimization/84265
10543 * tree-vect-stmts.c (vectorizable_store): Don't treat
10544 VMAT_CONTIGUOUS accesses as grouped.
10545 (vectorizable_load): Likewise.
10547 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10549 PR tree-optimization/81635
10550 * wide-int.h (wi::round_down_for_mask, wi::round_up_for_mask): Declare.
10551 * wide-int.cc (wi::round_down_for_mask, wi::round_up_for_mask)
10552 (test_round_for_mask): New functions.
10553 (wide_int_cc_tests): Call test_round_for_mask.
10554 * tree-vrp.h (intersect_range_with_nonzero_bits): Declare.
10555 * tree-vrp.c (intersect_range_with_nonzero_bits): New function.
10556 * tree-data-ref.c (split_constant_offset_1): Use it to refine the
10557 range returned by get_range_info.
10559 2018-02-08 Jan Hubicka <hubicka@ucw.cz>
10562 * cgraph.h (symtab_node::output_to_lto_symbol_table_p): Declare
10563 * symtab.c: Include builtins.h
10564 (symtab_node::output_to_lto_symbol_table_p): Move here
10565 from lto-streamer-out.c:output_symbol_p.
10566 * lto-streamer-out.c (write_symbol): Turn early exit to assert.
10567 (output_symbol_p): Move all logic to symtab.c
10568 (produce_symtab): Update.
10570 2018-02-08 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
10572 * config/s390/s390-opts.h (enum indirect_branch): Define.
10573 * config/s390/s390-protos.h (s390_return_addr_from_memory)
10574 (s390_indirect_branch_via_thunk)
10575 (s390_indirect_branch_via_inline_thunk): Add function prototypes.
10576 (enum s390_indirect_branch_type): Define.
10577 * config/s390/s390.c (struct s390_frame_layout, struct
10578 machine_function): Remove.
10579 (indirect_branch_prez10thunk_mask, indirect_branch_z10thunk_mask)
10580 (indirect_branch_table_label_no, indirect_branch_table_name):
10582 (INDIRECT_BRANCH_NUM_OPTIONS): Define macro.
10583 (enum s390_indirect_branch_option): Define.
10584 (s390_return_addr_from_memory): New function.
10585 (s390_handle_string_attribute): New function.
10586 (s390_attribute_table): Add new attribute handler.
10587 (s390_execute_label): Handle UNSPEC_EXECUTE_JUMP patterns.
10588 (s390_indirect_branch_via_thunk): New function.
10589 (s390_indirect_branch_via_inline_thunk): New function.
10590 (s390_function_ok_for_sibcall): When jumping via thunk disallow
10591 sibling call optimization for non z10 compiles.
10592 (s390_emit_call): Force indirect branch target to be a single
10593 register. Add r1 clobber for non-z10 compiles.
10594 (s390_emit_epilogue): Emit return jump via return_use expander.
10595 (s390_reorg): Handle JUMP_INSNs as execute targets.
10596 (s390_option_override_internal): Perform validity checks for the
10597 new command line options.
10598 (s390_indirect_branch_attrvalue): New function.
10599 (s390_indirect_branch_settings): New function.
10600 (s390_set_current_function): Invoke s390_indirect_branch_settings.
10601 (s390_output_indirect_thunk_function): New function.
10602 (s390_code_end): Implement target hook.
10603 (s390_case_values_threshold): Implement target hook.
10604 (TARGET_ASM_CODE_END, TARGET_CASE_VALUES_THRESHOLD): Define target
10606 * config/s390/s390.h (struct s390_frame_layout)
10607 (struct machine_function): Move here from s390.c.
10608 (TARGET_INDIRECT_BRANCH_NOBP_RET)
10609 (TARGET_INDIRECT_BRANCH_NOBP_JUMP)
10610 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_THUNK)
10611 (TARGET_INDIRECT_BRANCH_NOBP_JUMP_INLINE_THUNK)
10612 (TARGET_INDIRECT_BRANCH_NOBP_CALL)
10613 (TARGET_DEFAULT_INDIRECT_BRANCH_TABLE)
10614 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EXRL)
10615 (TARGET_INDIRECT_BRANCH_THUNK_NAME_EX)
10616 (TARGET_INDIRECT_BRANCH_TABLE): Define macros.
10617 * config/s390/s390.md (UNSPEC_EXECUTE_JUMP)
10618 (INDIRECT_BRANCH_THUNK_REGNUM): Define constants.
10619 (mnemonic attribute): Add values which aren't recognized
10621 ("*cjump_long", "*icjump_long", "*basr", "*basr_r"): Disable
10622 pattern for branch conversion. Fix mnemonic attribute.
10623 ("*c<code>", "*sibcall_br", "*sibcall_value_br", "*return"): Emit
10624 indirect branch via thunk if requested.
10625 ("indirect_jump", "<code>"): Expand patterns for branch conversion.
10626 ("*indirect_jump"): Disable for branch conversion using out of
10628 ("indirect_jump_via_thunk<mode>_z10")
10629 ("indirect_jump_via_thunk<mode>")
10630 ("indirect_jump_via_inlinethunk<mode>_z10")
10631 ("indirect_jump_via_inlinethunk<mode>", "*casesi_jump")
10632 ("casesi_jump_via_thunk<mode>_z10", "casesi_jump_via_thunk<mode>")
10633 ("casesi_jump_via_inlinethunk<mode>_z10")
10634 ("casesi_jump_via_inlinethunk<mode>", "*basr_via_thunk<mode>_z10")
10635 ("*basr_via_thunk<mode>", "*basr_r_via_thunk_z10")
10636 ("*basr_r_via_thunk", "return<mode>_prez10"): New pattern.
10637 ("*indirect2_jump"): Disable for branch conversion.
10638 ("casesi_jump"): Turn into expander and expand patterns for branch
10640 ("return_use"): New expander.
10641 ("*return"): Emit return via thunk and rename it to ...
10642 ("*return<mode>"): ... this one.
10643 * config/s390/s390.opt: Add new options and and enum for the
10646 2018-02-08 Richard Sandiford <richard.sandiford@linaro.org>
10648 * lra-constraints.c (match_reload): Unconditionally use
10649 gen_lowpart_SUBREG, rather than selecting between that
10650 and equivalent gen_rtx_SUBREG code.
10652 2018-02-08 Richard Biener <rguenther@suse.de>
10654 PR tree-optimization/84233
10655 * tree-ssa-phiprop.c (propagate_with_phi): Use separate
10656 changed flag instead of boguously re-using phi_inserted.
10658 2018-02-08 Martin Jambor <mjambor@suse.cz>
10660 * hsa-gen.c (get_symbol_for_decl): Set program allocation for
10661 static local variables.
10663 2018-02-08 Richard Biener <rguenther@suse.de>
10665 PR tree-optimization/84278
10666 * tree-vect-stmts.c (vectorizable_store): When looking for
10667 smaller vector types to perform grouped strided loads/stores
10668 make sure the mode is supported by the target.
10669 (vectorizable_load): Likewise.
10671 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
10673 * config/aarch64/aarch64.c (aarch64_components_for_bb):
10674 Increase LDP/STP opportunities by adding adjacent callee-saves.
10676 2018-02-08 Wilco Dijkstra <wdijkstr@arm.com>
10678 PR rtl-optimization/84068
10679 PR rtl-optimization/83459
10680 * haifa-sched.c (rank_for_schedule): Fix SCHED_PRESSURE_MODEL sorting.
10682 2018-02-08 Aldy Hernandez <aldyh@redhat.com>
10684 PR tree-optimization/84224
10685 * gimple-ssa-warn-alloca.c (pass_walloca::execute): Remove assert.
10686 * calls.c (gimple_alloca_call_p): Only return TRUE when we have
10687 non-zero arguments.
10689 2018-02-07 Iain Sandoe <iain@codesourcery.com>
10692 * config/rs6000/altivec.md (*restore_world): Remove LR use.
10693 * config/rs6000/predicates.md (restore_world_operation): Adjust op
10694 count, remove one USE.
10696 2018-02-07 Michael Meissner <meissner@linux.vnet.ibm.com>
10698 * doc/install.texi (Configuration): Document the
10699 --with-long-double-format={ibm,ieee} PowerPC configuration
10703 * config/rs6000/rs6000.md (fix_trunc<SFDF:mode><QHI:mode>2):
10704 Convert from define_expand to be define_insn_and_split. Rework
10705 float/double/_Float128 conversions to QI/HI/SImode to work with
10706 both ISA 2.07 (power8) or ISA 3.0 (power9). Fix regression where
10707 conversions to QI/HImode types did a store and then a load to
10708 truncate the value. For conversions to VSX registers, don't split
10709 the insn, instead emit the code directly. Use the code iterator
10710 any_fix to combine signed and unsigned conversions.
10711 (fix<uns>_trunc<SFDF:mode>si2_p8): Likewise.
10712 (fixuns_trunc<SFDF:mode><QHI:mode>2): Likewise.
10713 (fix_trunc<IEEE128:mode><QHI:mode>2): Likewise.
10714 (fix<uns>_trunc<SFDF:mode><QHI:mode>2): Likewise.
10715 (fix_<mode>di2_hw): Likewise.
10716 (fixuns_<mode>di2_hw): Likewise.
10717 (fix_<mode>si2_hw): Likewise.
10718 (fixuns_<mode>si2_hw): Likewise.
10719 (fix<uns>_<IEEE128:mode><SDI:mode>2_hw): Likewise.
10720 (fix<uns>_trunc<IEEE128:mode><QHI:mode>2): Likewise.
10721 (fctiw<u>z_<mode>_smallint): Rename fctiw<u>z_<mode>_smallint to
10722 fix<uns>_trunc<SFDF:mode>si2_p8.
10723 (fix_trunc<SFDF:mode><QHI:mode>2_internal): Delete, no longer used.
10724 (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
10725 (fix<uns>_<mode>_mem): Likewise.
10726 (fctiw<u>z_<mode>_mem): Likewise.
10727 (fix<uns>_<mode>_mem): Likewise.
10728 (fix<uns>_trunc<SFDF:mode><QHSI:mode>2_mem): On ISA 3.0, prevent
10729 the register allocator from doing a direct move to the GPRs to do
10730 a store, and instead use the ISA 3.0 store byte/half-word from
10731 vector register instruction. For IEEE 128-bit floating point,
10732 also optimize stores of 32-bit ints.
10733 (fix<uns>_trunc<IEEE128:mode><QHSI:mode>2_mem): Likewise.
10735 2018-02-07 Alan Hayward <alan.hayward@arm.com>
10737 * genextract.c (push_pathstr_operand): New function to support
10739 (walk_rtx): Call push_pathstr_operand.
10740 (print_path): Support [a-zA-Z].
10742 2018-02-07 Richard Biener <rguenther@suse.de>
10744 PR tree-optimization/84037
10745 * tree-vectorizer.h (struct _loop_vec_info): Add ivexpr_map member.
10746 (cse_and_gimplify_to_preheader): Declare.
10747 (vect_get_place_in_interleaving_chain): Likewise.
10748 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
10750 (_loop_vec_info::~_loop_vec_info): Delete it.
10751 (cse_and_gimplify_to_preheader): New function.
10752 * tree-vect-slp.c (vect_get_place_in_interleaving_chain): Export.
10753 * tree-vect-stmts.c (vectorizable_store): CSE base and steps.
10754 (vectorizable_load): Likewise. For grouped stores always base
10755 the IV on the first element.
10756 * tree-vect-loop-manip.c (vect_loop_versioning): Unshare versioning
10757 condition before gimplifying.
10759 2018-02-07 Jakub Jelinek <jakub@redhat.com>
10761 * tree-eh.c (operation_could_trap_helper_p): Ignore honor_trapv for
10762 *DIV_EXPR and *MOD_EXPR.
10764 2018-02-07 H.J. Lu <hongjiu.lu@intel.com>
10767 * config/i386/i386.c (ix86_option_override_internal): Mask out
10768 the CF_SET bit when checking -fcf-protection.
10770 2018-02-07 Tom de Vries <tom@codesourcery.com>
10773 * omp-expand.c (expand_oacc_collapse_init): Ensure diff_type is large
10776 2018-02-07 Richard Biener <rguenther@suse.de>
10778 PR tree-optimization/84204
10779 * tree-chrec.c (chrec_fold_plus_1): Remove size limiting in
10782 PR tree-optimization/84205
10783 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Also
10784 special-case isl_ast_op_zdiv_r.
10786 PR tree-optimization/84223
10787 * graphite-scop-detection.c (gather_bbs::before_dom_children):
10788 Only add conditions from within the region.
10789 (gather_bbs::after_dom_children): Adjust.
10791 2018-02-07 Georg-Johann Lay <avr@gjlay.de>
10794 * config/avr/avr.h (GENERAL_REGNO_P, GENERAL_REG_P): New macros.
10795 * config/avr/avr.md: Only post-reload split REG-REG moves if
10796 either register is GENERAL_REG_P.
10798 2018-02-07 Jakub Jelinek <jakub@redhat.com>
10800 PR tree-optimization/84235
10801 * tree-ssa-scopedtables.c
10802 (avail_exprs_stack::simplify_binary_operation): Fir MINUS_EXPR, punt
10803 if the subtraction is performed in floating point type where NaNs are
10804 honored. For *DIV_EXPR, punt for ALL_FRACT_MODE_Ps where we can't
10805 build 1. Formatting fix.
10807 2018-02-06 Jakub Jelinek <jakub@redhat.com>
10810 * config/i386/i386.c (rest_of_insert_endbranch): Only skip
10811 NOTE_INSN_CALL_ARG_LOCATION after a call, not anything else,
10812 and skip it regardless of bb boundaries. Use CALL_P macro,
10813 don't test INSN_P (insn) together with CALL_P or JUMP_P check
10814 unnecessarily, formatting fix.
10816 2018-02-06 Michael Collison <michael.collison@arm.com>
10818 * config/arm/thumb2.md:
10819 (*thumb2_mov_negscc): Split only if TARGET_THUMB2 && !arm_restrict_it.
10820 (*thumb_mov_notscc): Ditto.
10822 2018-02-06 Michael Meissner <meissner@linux.vnet.ibm.com>
10825 * config/rs6000/rs6000.md (su code attribute): Use "u" for
10826 unsigned_fix, not "s".
10828 2018-02-06 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
10830 * configure.ac (gcc_fn_eh_frame_ro): New function.
10831 (gcc_cv_as_cfi_directive): Check both 32 and 64-bit assembler for
10832 correct .eh_frame permissions.
10833 * configure: Regenerate.
10835 2018-02-06 Andrew Jenner <andrew@codeourcery.com>
10837 * doc/invoke.texi: Add section for the PowerPC SPE backend. Remove
10838 irrelevant options.
10840 2018-02-06 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
10842 * config/rs6000/rs6000.c (rs6000_option_override_internal):
10843 Display warning message for -mno-speculate-indirect-jumps.
10845 2018-02-06 Andrew Jenner <andrew@codesourcery.com>
10847 * config/powerpcspe/powerpcspe.opt: (msimple-fpu, mfpu) Add
10849 * config/powerpcspe/sysv4.opt (mbit-align): Likewise.
10851 2018-02-06 Aldy Hernandez <aldyh@redhat.com>
10853 PR tree-optimization/84225
10854 * tree-eh.c (find_trapping_overflow): Only call
10855 operation_no_trapping_overflow when ANY_INTEGRAL_TYPE_P.
10857 2018-02-06 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10860 * config/i386/i386.c: Reimplement the check of possible options
10861 -mibt/-mshstk conbination. Change error messages.
10862 * doc/invoke.texi: Fix a typo: remove extra '='.
10864 2018-02-06 Marek Polacek <polacek@redhat.com>
10866 PR tree-optimization/84228
10867 * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Skip debug statements.
10869 2018-02-06 Tamar Christina <tamar.christina@arm.com>
10872 * config/arm/arm.c (arm_print_asm_arch_directives): Record already
10873 emitted arch directives.
10874 * config/arm/arm-c.c (arm_cpu_builtins): Undefine __ARM_ARCH and
10875 __ARM_FEATURE_COPROC before changing architectures.
10877 2018-02-06 Richard Biener <rguenther@suse.de>
10879 * config/i386/i386.c (print_reg): Fix typo.
10880 (ix86_loop_unroll_adjust): Do not unroll beyond the original nunroll.
10882 2018-02-06 Eric Botcazou <ebotcazou@adacore.com>
10884 * configure: Regenerate.
10886 2018-02-05 Martin Sebor <msebor@redhat.com>
10888 PR tree-optimization/83369
10889 * tree-ssa-ccp.c (pass_post_ipa_warn::execute): Use %G to print
10892 2018-02-05 Martin Liska <mliska@suse.cz>
10894 * doc/invoke.texi: Cherry-pick upstream r323995.
10896 2018-02-05 Richard Sandiford <richard.sandiford@linaro.org>
10898 * ira.c (ira_init_register_move_cost): Adjust comment.
10900 2018-02-05 Martin Liska <mliska@suse.cz>
10902 PR gcov-profile/84137
10903 * doc/gcov.texi: Fix typo in documentation.
10905 2018-02-05 Martin Liska <mliska@suse.cz>
10907 PR gcov-profile/83879
10908 * doc/gcov.texi: Document necessity of --dynamic-list-data when
10909 using dlopen functionality.
10911 2018-02-05 Olga Makhotina <olga.makhotina@intel.com>
10913 * config/i386/avx512dqintrin.h (_mm_mask_range_sd, _mm_maskz_range_sd,
10914 _mm_mask_range_round_sd, _mm_maskz_range_round_sd, _mm_mask_range_ss,
10915 _mm_maskz_range_ss, _mm_mask_range_round_ss,
10916 _mm_maskz_range_round_ss): New intrinsics.
10917 (__builtin_ia32_rangesd128_round)
10918 (__builtin_ia32_rangess128_round): Remove.
10919 (__builtin_ia32_rangesd128_mask_round,
10920 __builtin_ia32_rangess128_mask_round): New builtins.
10921 * config/i386/i386-builtin.def (__builtin_ia32_rangesd128_round,
10922 __builtin_ia32_rangess128_round): Remove.
10923 (__builtin_ia32_rangesd128_mask_round,
10924 __builtin_ia32_rangess128_mask_round): New builtins.
10925 * config/i386/sse.md (ranges<mode><round_saeonly_name>): Renamed to ...
10926 (ranges<mode><mask_scalar_name><round_saeonly_scalar_name>): ... this.
10927 ((match_operand:VF_128 2 "<round_saeonly_nimm_predicate>"
10928 "<round_saeonly_constraint>")): Changed to ...
10929 ((match_operand:VF_128 2 "<round_saeonly_scalar_nimm_predicate>"
10930 "<round_saeonly_scalar_constraint>")): ... this.
10931 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
10932 %0, %1, %2<round_saeonly_op4>, %3}"): Changed to ...
10933 ("vrange<ssescalarmodesuffix>\t{%3, <round_saeonly_scalar_mask_op4>%2,
10934 %1, %0<mask_scalar_operand4>|%0<mask_scalar_operand4>, %1,
10935 %2<round_saeonly_scalar_mask_op4>, %3}"): ... this.
10937 2018-02-02 Andrew Jenner <andrew@codesourcery.com>
10939 * config/powerpcspe/powerpcspe.opt: Add Undocumented to irrelevant
10941 * config/powerpcspe/powerpcspe-tables.opt (rs6000_cpu_opt_value):
10942 Remove all values except native, 8540 and 8548.
10944 2018-02-02 H.J. Lu <hongjiu.lu@intel.com>
10946 * config/i386/i386.c (ix86_output_function_return): Pass
10947 INVALID_REGNUM, instead of -1, as invalid register number to
10948 indirect_thunk_name and output_indirect_thunk.
10950 2018-02-02 Julia Koval <julia.koval@intel.com>
10952 * config.gcc: Add -march=icelake.
10953 * config/i386/driver-i386.c (host_detect_local_cpu): Detect icelake.
10954 * config/i386/i386-c.c (ix86_target_macros_internal): Handle icelake.
10955 * config/i386/i386.c (processor_costs): Add m_ICELAKE.
10956 (PTA_ICELAKE, PTA_AVX512VNNI, PTA_GFNI, PTA_VAES, PTA_AVX512VBMI2,
10957 PTA_VPCLMULQDQ, PTA_RDPID, PTA_AVX512BITALG): New.
10958 (processor_target_table): Add icelake.
10959 (ix86_option_override_internal): Handle new PTAs.
10960 (get_builtin_code_for_version): Handle icelake.
10961 (M_INTEL_COREI7_ICELAKE): New.
10962 (fold_builtin_cpu): Handle icelake.
10963 * config/i386/i386.h (TARGET_ICELAKE, PROCESSOR_ICELAKE): New.
10964 * doc/invoke.texi: Add -march=icelake.
10966 2018-02-02 Julia Koval <julia.koval@intel.com>
10968 * config/i386/i386.c (ix86_option_override_internal): Change flags type
10969 to wide_int_bitmask.
10970 * wide-int-bitmask.h: New.
10972 2018-02-02 Igor Tsimbalist <igor.v.tsimbalist@intel.com>
10975 * config/i386/i386.md: Replace Pmode with word_mode in
10976 builtin_setjmp_setup and builtin_longjmp to support x32.
10978 2018-02-01 Peter Bergner <bergner@vnet.ibm.com>
10982 * config/rs6000/driver-rs6000.c: #include "diagnostic.h".
10984 (rs6000_supported_cpu_names): New static variable.
10985 (linux_cpu_translation_table): Likewise.
10986 (elf_platform) <cpu>: Define new static variable and use it.
10987 Translate kernel AT_PLATFORM name to canonical name if needed.
10988 Error if platform name is unknown.
10990 2018-02-01 Aldy Hernandez <aldyh@redhat.com>
10993 * config/pa/predicates.md (base14_operand): Handle E_VOIDmode.
10995 2018-02-01 Jeff Law <law@redhat.com>
10998 * config/i386/i386.c (release_scratch_register_on_entry): Add new
10999 OFFSET and RELEASE_VIA_POP arguments. Use SP+OFFSET to restore
11000 the scratch if RELEASE_VIA_POP is false.
11001 (ix86_adjust_stack_and_probe_stack_clash): Un-constify SIZE.
11002 If we have to save a temporary register, decrement SIZE appropriately.
11003 Pass new arguments to release_scratch_register_on_entry.
11004 (ix86_adjust_stack_and_probe): Likewise.
11005 (ix86_emit_probe_stack_range): Pass new arguments to
11006 release_scratch_register_on_entry.
11008 2018-02-01 Uros Bizjak <ubizjak@gmail.com>
11010 PR rtl-optimization/84157
11011 * combine.c (change_zero_ext): Use REG_P predicate in
11012 front of HARD_REGISTER_P predicate.
11014 2018-02-01 Georg-Johann Lay <avr@gjlay.de>
11016 * config/avr/avr.c (avr_option_override): Move disabling of
11017 -fdelete-null-pointer-checks to...
11018 * common/config/avr/avr-common.c (avr_option_optimization_table):
11021 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11023 PR tree-optimization/81635
11024 * tree-data-ref.c (split_constant_offset_1): For types that
11025 wrap on overflow, try to use range info to prove that wrapping
11028 2018-02-01 Renlin Li <renlin.li@arm.com>
11031 * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
11032 TAILCALL_ADDR_REGS.
11033 (aarch64_register_move_cost): Likewise.
11034 * config/aarch64/aarch64.h (reg_class): Rename CALLER_SAVE_REGS to
11035 TAILCALL_ADDR_REGS.
11036 (REG_CLASS_NAMES): Likewise.
11037 (REG_CLASS_CONTENTS): Rename CALLER_SAVE_REGS to
11038 TAILCALL_ADDR_REGS. Remove IP registers.
11039 * config/aarch64/aarch64.md (Ucs): Update register constraint.
11041 2018-02-01 Richard Biener <rguenther@suse.de>
11043 * domwalk.h (dom_walker::dom_walker): Add additional constructor
11044 for specifying RPO order and allow NULL for that.
11045 * domwalk.c (dom_walker::dom_walker): Likewise.
11046 (dom_walker::walk): Handle NULL RPO order.
11047 * tree-into-ssa.c (rewrite_dom_walker): Do not walk dom children
11049 (rewrite_update_dom_walker): Likewise.
11050 (mark_def_dom_walker): Likewise.
11052 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11054 * config/aarch64/aarch64-protos.h (aarch64_split_sve_subreg_move)
11055 (aarch64_maybe_expand_sve_subreg_move): Declare.
11056 * config/aarch64/aarch64.md (UNSPEC_REV_SUBREG): New unspec.
11057 * config/aarch64/predicates.md (aarch64_any_register_operand): New
11059 * config/aarch64/aarch64-sve.md (mov<mode>): Optimize subreg moves
11060 that are semantically a reverse operation.
11061 (*aarch64_sve_mov<mode>_subreg_be): New pattern.
11062 * config/aarch64/aarch64.c (aarch64_maybe_expand_sve_subreg_move):
11063 (aarch64_replace_reg_mode, aarch64_split_sve_subreg_move): New
11065 (aarch64_can_change_mode_class): For big-endian, forbid changes
11066 between two SVE modes if they have different element sizes.
11068 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11070 * config/aarch64/aarch64.c (aarch64_expand_sve_const_vector): Prefer
11071 the TImode handling for big-endian targets.
11073 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11075 * config/aarch64/aarch64-sve.md (sve_ld1rq): Replace with...
11076 (*sve_ld1rq<Vesize>): ... this new pattern. Handle all element sizes,
11078 * config/aarch64/aarch64.c (aarch64_expand_sve_widened_duplicate):
11079 Remove BSWAP handing for big-endian targets and use the form of
11080 LD1RQ appropariate for the mode.
11082 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11084 * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Handle
11085 all CONST_VECTOR_DUPLICATE_P vectors, not just those with a single
11086 duplicated element.
11088 2018-02-01 Richard Sandiford <richard.sandiford@linaro.org>
11091 * config/aarch64/aarch64.c (aarch64_secondary_reload): Tighten
11092 check for operands that need to go through aarch64_sve_reload_be.
11094 2018-02-01 Jakub Jelinek <jakub@redhat.com>
11096 PR tree-optimization/81661
11097 PR tree-optimization/84117
11098 * tree-eh.h (rewrite_to_non_trapping_overflow): Declare.
11099 * tree-eh.c: Include gimplify.h.
11100 (find_trapping_overflow, replace_trapping_overflow,
11101 rewrite_to_non_trapping_overflow): New functions.
11102 * tree-vect-loop.c: Include tree-eh.h.
11103 (vect_get_loop_niters): Use rewrite_to_non_trapping_overflow.
11104 * tree-data-ref.c: Include tree-eh.h.
11105 (get_segment_min_max): Use rewrite_to_non_trapping_overflow.
11107 2018-01-31 Uros Bizjak <ubizjak@gmail.com>
11109 PR rtl-optimization/84123
11110 * combine.c (change_zero_ext): Check if hard register satisfies
11111 can_change_dest_mode before calling gen_lowpart_SUBREG.
11113 2018-01-31 Vladimir Makarov <vmakarov@redhat.com>
11116 * ira.c (ira_init_register_move_cost): Remove assert.
11118 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11120 PR rtl-optimization/84071
11121 * doc/tm.texi.in (WORD_REGISTER_OPERATIONS): Add explicit case.
11122 * doc/tm.texi: Regenerate.
11124 2018-01-31 Richard Biener <rguenther@suse.de>
11126 PR tree-optimization/84132
11127 * tree-data-ref.c (analyze_miv_subscript): Properly
11128 check whether evolution_function_is_affine_multivariate_p
11129 before calling gcd_of_steps_may_divide_p.
11131 2018-01-31 Julia Koval <julia.koval@intel.com>
11134 * config/i386/i386.c (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
11135 * config/i386/i386.md (rdpid_rex64) New.
11136 (rdpid): Make 32bit only.
11138 2018-01-29 Aldy Hernandez <aldyh@redhat.com>
11141 * tree-pretty-print.c (dump_generic_node): Handle a TYPE_NAME with
11142 an IDENTIFIER_NODE for FUNCTION_TYPE's.
11144 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11147 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
11149 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
11151 2018-01-31 Eric Botcazou <ebotcazou@adacore.com>
11153 PR rtl-optimization/84071
11154 * combine.c (record_dead_and_set_regs_1): Record the source unmodified
11155 for a paradoxical SUBREG on a WORD_REGISTER_OPERATIONS target.
11157 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11159 * config/arc/arc.c (arc_handle_aux_attribute): New function.
11160 (arc_attribute_table): Add 'aux' attribute.
11161 (arc_in_small_data_p): Consider aux like variables.
11162 (arc_is_aux_reg_p): New function.
11163 (arc_asm_output_aligned_decl_local): Ignore 'aux' like variables.
11164 (arc_get_aux_arg): New function.
11165 (prepare_move_operands): Handle aux-register access.
11166 (arc_handle_aux_attribute): New function.
11167 * doc/extend.texi (ARC Variable attributes): Add subsection.
11169 2018-01-31 Claudiu Zissulescu <claziss@synopsys.com>
11171 * config/arc/arc-protos.h (arc_is_uncached_mem_p): Function proto.
11172 * config/arc/arc.c (arc_handle_uncached_attribute): New function.
11173 (arc_attribute_table): Add 'uncached' attribute.
11174 (arc_print_operand): Print '.di' flag for uncached memory
11176 (arc_in_small_data_p): Do not consider for small data the uncached
11178 (arc_is_uncached_mem_p): New function.
11179 * config/arc/predicates.md (compact_store_memory_operand): Check
11180 for uncached memory accesses.
11181 (nonvol_nonimm_operand): Likewise.
11182 * gcc/doc/extend.texi (ARC Type Attribute): New subsection.
11184 2018-01-31 Jakub Jelinek <jakub@redhat.com>
11187 * common.opt (falign-functions=, falign-jumps=, falign-labels=,
11188 falign-loops=): Add Optimization flag.
11190 2018-01-30 Jeff Law <law@redhat.com>
11193 * i386.c (ix86_adjust_stack_and_probe_stack_clash): New argument
11194 INT_REGISTERS_SAVED. Check it prior to calling
11195 get_scratch_register_on_entry.
11196 (ix86_adjust_stack_and_probe): Similarly.
11197 (ix86_emit_probe_stack_range): Similarly.
11198 (ix86_expand_prologue): Corresponding changes.
11200 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11203 * config/sol2.h (STARTFILE_ARCH_SPEC): Use -std=c*,
11204 -std=iso9899:199409 instead of -pedantic to select values-Xc.o.
11206 2018-01-30 Vladimir Makarov <vmakarov@redhat.com>
11209 * lra-constraints.c (curr_insn_transform): Process AND in the
11212 2018-01-30 Jakub Jelinek <jakub@redhat.com>
11214 PR rtl-optimization/83986
11215 * sched-deps.c (sched_analyze_insn): For frame related insns, add anti
11216 dependence against last_pending_memory_flush in addition to
11217 pending_jump_insns.
11219 2018-01-30 Alexandre Oliva <aoliva@redhat.com>
11221 PR tree-optimization/81611
11222 * tree-ssa-dom.c (simple_iv_increment_p): Skip intervening
11225 2018-01-30 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
11228 * config/rs6000/rs6000.c (rs6000_internal_arg_pointer): Only return
11231 2018-01-30 Richard Biener <rguenther@suse.de>
11232 Jakub Jelinek <jakub@redhat.com>
11234 PR tree-optimization/84111
11235 * tree-ssa-loop-ivcanon.c (tree_unroll_loops_completely_1): Skip
11236 inner loops added during recursion, as they don't have up-to-date
11239 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
11242 * ipa-inline.c (can_inline_edge_p): Break out late tests to...
11243 (can_inline_edge_by_limits_p): ... here.
11244 (can_early_inline_edge_p, check_callers,
11245 update_caller_keys, update_callee_keys, recursive_inlining,
11246 add_new_edges_to_heap, speculation_useful_p,
11247 inline_small_functions,
11248 inline_small_functions, flatten_function,
11249 inline_to_all_callers_1): Update.
11251 2018-01-30 Jan Hubicka <hubicka@ucw.cz>
11253 * profile-count.c (profile_count::combine_with_ipa_count): Handle
11256 2018-01-30 Richard Biener <rguenther@suse.de>
11258 PR tree-optimization/83008
11259 * tree-vect-slp.c (vect_analyze_slp_cost_1): Properly cost
11260 invariant and constant vector uses in stmts when they need
11261 more than one stmt.
11263 2018-01-30 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11266 * configure.ac (gcc_cv_as_shf_merge): Disable on Solaris 10/x86.
11267 * configure: Regenerate.
11269 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11271 * config/aarch64/aarch64-sve.md (*vec_extract<mode><Vel>_0): New
11273 (*vec_extract<mode><Vel>_v128): Require a nonzero lane number.
11274 Use gen_rtx_REG rather than gen_lowpart.
11276 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11278 * lra-constraints.c (match_reload): Use subreg_lowpart_offset
11279 rather than 0 when creating partial subregs.
11281 2018-01-30 Richard Sandiford <richard.sandiford@linaro.org>
11283 * vec-perm-indices.c (vec_perm_indices::series_p): Give examples
11286 2018-01-29 Michael Meissner <meissner@linux.vnet.ibm.com>
11289 * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): If DFmode
11290 and SFmode can go in Altivec registers (-mcpu=power7 for DFmode,
11291 -mcpu=power8 for SFmode) don't set the PRE_INCDEC or PRE_MODIFY
11292 flags. This restores the settings used before the 2017-07-24.
11293 Turning off pre increment/decrement/modify allows IVOPTS to
11294 optimize DF/SF loops where the index is an int.
11296 2018-01-29 Richard Biener <rguenther@suse.de>
11297 Kelvin Nilsen <kelvin@gcc.gnu.org>
11300 * tree-vect-stmts.c (vectorizable_call): Don't call
11301 targetm.vectorize_builtin_md_vectorized_function if callee is
11304 2018-01-22 Carl Love <cel@us.ibm.com>
11306 * doc/extend.tex: Fix typo in second arg in
11307 __builtin_bcdadd_{lt|eq|gt|ov} and __builtin_bcdsub_{lt|eq|gt|ov}.
11309 2018-01-29 Richard Biener <rguenther@suse.de>
11311 PR tree-optimization/84086
11312 * tree-ssanames.c: Include cfgloop.h and tree-scalar-evolution.h.
11313 (flush_ssaname_freelist): When SSA names were released reset
11314 the SCEV hash table.
11316 2018-01-29 Richard Biener <rguenther@suse.de>
11318 PR tree-optimization/84057
11319 * tree-ssa-loop-ivcanon.c (unloop_loops): Deal with already
11320 removed paths when removing edges.
11322 2018-01-27 H.J. Lu <hongjiu.lu@intel.com>
11324 * doc/invoke.texi: Replace -mfunction-return==@var{choice} with
11325 -mfunction-return=@var{choice}.
11327 2018-01-27 Bernd Edlinger <bernd.edlinger@hotmail.de>
11329 PR diagnostic/84034
11330 * diagnostic-show-locus.c (get_line_width_without_trailing_whitespace):
11331 Handle CR like TAB.
11332 (layout::print_source_line): Likewise.
11333 (test_get_line_width_without_trailing_whitespace): Add test cases.
11335 2018-01-27 Jakub Jelinek <jakub@redhat.com>
11337 PR middle-end/84040
11338 * sched-deps.c (sched_macro_fuse_insns): Return immediately for
11341 2018-01-26 Jim Wilson <jimw@sifive.com>
11343 * config/riscv/riscv.h (MAX_FIXED_MODE_SIZE): New.
11345 * config/riscv/elf.h (LIB_SPEC): Don't include -lgloss when nosys.specs
11348 2018-01-26 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11350 * config/aarch64/aarch64.md: Add peepholes for CMP + SUB -> SUBS
11351 and CMP + SUB-immediate -> SUBS.
11353 2018-01-26 Martin Sebor <msebor@redhat.com>
11355 PR tree-optimization/83896
11356 * tree-ssa-strlen.c (get_string_len): Rename...
11357 (get_string_cst_length): ...to this. Return HOST_WIDE_INT.
11358 Avoid assuming length is constant.
11359 (handle_char_store): Use HOST_WIDE_INT for string length.
11361 2018-01-26 Uros Bizjak <ubizjak@gmail.com>
11364 * config/i386/i386.md (*andndi3_doubleword): Add earlyclobber
11365 to (=&r,r,rm) alternative. Add (=r,0,rm) and (=r,r,0) alternatives.
11367 2018-01-26 Richard Biener <rguenther@suse.de>
11369 PR rtl-optimization/84003
11370 * dse.c (record_store): Only record redundant stores when
11371 the earlier store aliases at least all accesses the later one does.
11373 2018-01-26 Jakub Jelinek <jakub@redhat.com>
11375 PR rtl-optimization/83985
11376 * dce.c (deletable_insn_p): Return false for separate shrink wrapping
11377 REG_CFA_RESTORE insns.
11378 (delete_unmarked_insns): Don't ignore separate shrink wrapping
11379 REG_CFA_RESTORE insns here.
11382 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Don't
11383 use SSA_NAME_VAR as base for SSA_NAMEs with non-NULL SSA_NAME_VAR.
11385 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11387 * config/arc/arc-arch.h (arc_tune_attr): Add ARC_TUNE_CORE_3.
11388 * config/arc/arc.c (arc_sched_issue_rate): Use ARC_TUNE_... .
11389 (arc_init): Likewise.
11390 (arc_override_options): Likewise.
11391 (arc_file_start): Choose Tag_ARC_CPU_variation based on arc_tune
11393 (hwloop_fail): Use TARGET_DBNZ when we want to check for dbnz insn
11395 * config/arc/arc.h (TARGET_DBNZ): Define.
11396 * config/arc/arc.md (attr tune): Add core_3, use ARC_TUNE_... to
11397 properly set the tune attribute.
11398 (dbnz): Use TARGET_DBNZ guard.
11399 * config/arc/arc.opt (mtune): Add core3 option.
11401 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11403 * config/arc/arc.c (arc_delegitimize_address_0): Refactored to
11404 recognize new pic like addresses.
11405 (arc_delegitimize_address): Clean up.
11407 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11409 * config/arc/arc-arches.def: Option mrf16 valid for all
11411 * config/arc/arc-c.def (__ARC_RF16__): New predefined macro.
11412 * config/arc/arc-cpus.def (em_mini): New cpu with rf16 on.
11413 * config/arc/arc-options.def (FL_RF16): Add mrf16 option.
11414 * config/arc/arc-tables.opt: Regenerate.
11415 * config/arc/arc.c (arc_conditional_register_usage): Handle
11416 reduced register file case.
11417 (arc_file_start): Set must have build attributes.
11418 * config/arc/arc.h (MAX_ARC_PARM_REGS): Conditional define using
11419 mrf16 option value.
11420 * config/arc/arc.opt (mrf16): Add new option.
11421 * config/arc/elf.h (ATTRIBUTE_PCS): Define.
11422 * config/arc/genmultilib.awk: Handle new mrf16 option.
11423 * config/arc/linux.h (ATTRIBUTE_PCS): Define.
11424 * config/arc/t-multilib: Regenerate.
11425 * doc/invoke.texi (ARC Options): Document mrf16 option.
11427 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11429 * config/arc/arc-protos.h: Add arc_is_secure_call_p proto.
11430 * config/arc/arc.c (arc_handle_secure_attribute): New function.
11431 (arc_attribute_table): Add 'secure_call' attribute.
11432 (arc_print_operand): Print secure call operand.
11433 (arc_function_ok_for_sibcall): Don't optimize tail calls when
11435 (arc_is_secure_call_p): New function. * config/arc/arc.md
11436 (call_i): Add support for sjli instruction.
11437 (call_value_i): Likewise.
11438 * config/arc/constraints.md (Csc): New constraint.
11440 2018-01-26 Claudiu Zissulescu <claziss@synopsys.com>
11441 John Eric Martin <John.Martin@emmicro-us.com>
11443 * config/arc/arc-protos.h: Add arc_is_jli_call_p proto.
11444 * config/arc/arc.c (_arc_jli_section): New struct.
11445 (arc_jli_section): New type.
11446 (rc_jli_sections): New static variable.
11447 (arc_handle_jli_attribute): New function.
11448 (arc_attribute_table): Add jli_always and jli_fixed attribute.
11449 (arc_file_end): New function.
11450 (TARGET_ASM_FILE_END): Define.
11451 (arc_print_operand): Reuse 'S' letter for JLI output instruction.
11452 (arc_add_jli_section): New function.
11453 (jli_call_scan): Likewise.
11454 (arc_reorg): Call jli_call_scan.
11455 (arc_output_addsi): Remove 'S' from printing asm operand.
11456 (arc_is_jli_call_p): New function.
11457 * config/arc/arc.md (movqi_insn): Remove 'S' from printing asm
11459 (movhi_insn): Likewise.
11460 (movsi_insn): Likewise.
11461 (movsi_set_cc_insn): Likewise.
11462 (loadqi_update): Likewise.
11463 (load_zeroextendqisi_update): Likewise.
11464 (load_signextendqisi_update): Likewise.
11465 (loadhi_update): Likewise.
11466 (load_zeroextendhisi_update): Likewise.
11467 (load_signextendhisi_update): Likewise.
11468 (loadsi_update): Likewise.
11469 (loadsf_update): Likewise.
11470 (movsicc_insn): Likewise.
11471 (bset_insn): Likewise.
11472 (bxor_insn): Likewise.
11473 (bclr_insn): Likewise.
11474 (bmsk_insn): Likewise.
11475 (bicsi3_insn): Likewise.
11476 (cmpsi_cc_c_insn): Likewise.
11477 (movsi_ne): Likewise.
11478 (movsi_cond_exec): Likewise.
11479 (clrsbsi2): Likewise.
11480 (norm_f): Likewise.
11490 (call_i): Remove 'S' asm letter, add jli instruction.
11491 (call_value_i): Likewise.
11492 * config/arc/arc.op (mjli-always): New option.
11493 * config/arc/constraints.md (Cji): New constraint.
11494 * config/arc/fpx.md (addsf3_fpx): Remove 'S' from printing asm
11496 (subsf3_fpx): Likewise.
11497 (mulsf3_fpx): Likewise.
11498 * config/arc/simdext.md (vendrec_insn): Remove 'S' from printing
11500 * doc/extend.texi (ARC): Document 'jli-always' and 'jli-fixed'
11501 function attrbutes.
11502 * doc/invoke.texi (ARC): Document mjli-always option.
11504 2018-01-26 Sebastian Perta <sebastian.perta@renesas.com>
11506 * config/rl78/rl78.c (rl78_addsi3_internal): If operand 2 is const
11507 avoid addition with 0 and use incw and decw where possible.
11509 2018-01-26 Richard Biener <rguenther@suse.de>
11511 PR tree-optimization/81082
11512 * fold-const.c (fold_plusminus_mult_expr): Do not perform the
11513 association if it requires casting to unsigned.
11514 * match.pd ((A * C) +- (B * C) -> (A+-B)): New patterns derived
11515 from fold_plusminus_mult_expr to catch important cases late when
11516 range info is available.
11518 2018-01-26 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
11520 * config/i386/sol2.h (USE_HIDDEN_LINKONCE): Remove.
11521 * configure.ac (hidden_linkonce): New test.
11522 * configure: Regenerate.
11523 * config.in: Regenerate.
11525 2018-01-26 Julia Koval <julia.koval@intel.com>
11527 * config/i386/avx512bitalgintrin.h (_mm512_bitshuffle_epi64_mask,
11528 _mm512_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
11529 _mm256_mask_bitshuffle_epi64_mask, _mm_bitshuffle_epi64_mask,
11530 _mm_mask_bitshuffle_epi64_mask): Fix type.
11531 * config/i386/i386-builtin-types.def (UHI_FTYPE_V2DI_V2DI_UHI,
11532 USI_FTYPE_V4DI_V4DI_USI): Remove.
11533 * config/i386/i386-builtin.def (__builtin_ia32_vpshufbitqmb512_mask,
11534 __builtin_ia32_vpshufbitqmb256_mask,
11535 __builtin_ia32_vpshufbitqmb128_mask): Fix types.
11536 * config/i386/i386.c (ix86_expand_args_builtin): Remove old types.
11537 * config/i386/sse.md (VI1_AVX512VLBW): Change types.
11539 2018-01-26 Alan Modra <amodra@gmail.com>
11542 * config/rs6000/rs6000-p8swap.c (rtx_is_swappable_p): Exclude
11543 UNSPEC_VBPERMQ. Sort other unspecs.
11545 2018-01-25 David Edelsohn <dje.gcc@gmail.com>
11547 * doc/invoke.texi (PowerPC Options): Document 'native' cpu type.
11549 2018-01-25 Jan Hubicka <hubicka@ucw.cz>
11551 PR middle-end/83055
11552 * predict.c (drop_profile): Do not push/pop cfun; update also
11554 (handle_missing_profiles): Fix logic looking for zero profiles.
11556 2018-01-25 Jakub Jelinek <jakub@redhat.com>
11558 PR middle-end/83977
11559 * ipa-fnsummary.c (compute_fn_summary): Clear can_change_signature
11560 on functions with #pragma omp declare simd or functions with simd
11562 * omp-simd-clone.c (expand_simd_clones): Revert 2018-01-24 change.
11563 * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
11564 Remove trailing \n from warning_at calls.
11566 2018-01-25 Tom de Vries <tom@codesourcery.com>
11569 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
11570 for neutered workers.
11572 2018-01-24 Joseph Myers <joseph@codesourcery.com>
11575 * config/m68k/m68k.c (m68k_promote_function_mode): New function.
11576 (TARGET_PROMOTE_FUNCTION_MODE): New macro.
11578 2018-01-24 Jeff Law <law@redhat.com>
11581 * i386.c (get_probe_interval): Move to earlier point.
11582 (ix86_compute_frame_layout): If -fstack-clash-protection and
11583 the frame is larger than the probe interval, then use pushes
11584 to save registers rather than reg->mem moves.
11585 (ix86_expand_prologue): Remove conditional for int_registers_saved
11588 2018-01-24 Vladimir Makarov <vmakarov@redhat.com>
11591 * ira-build.c (setup_min_max_allocno_live_range_point): Set up
11592 min/max for never referenced object.
11594 2018-01-24 Jakub Jelinek <jakub@redhat.com>
11596 PR middle-end/83977
11597 * tree.c (free_lang_data_in_decl): Don't clear DECL_ABSTRACT_ORIGIN
11599 * omp-low.c (create_omp_child_function): Remove "omp declare simd"
11600 attributes from DECL_ATTRIBUTES (decl) without affecting
11601 DECL_ATTRIBUTES (current_function_decl).
11602 * omp-simd-clone.c (expand_simd_clones): Ignore DECL_ARTIFICIAL
11603 functions with non-NULL DECL_ABSTRACT_ORIGIN.
11605 2018-01-24 Richard Sandiford <richard.sandiford@linaro.org>
11607 PR tree-optimization/83979
11608 * fold-const.c (fold_comparison): Use constant_boolean_node
11609 instead of boolean_{true,false}_node.
11611 2018-01-24 Jan Hubicka <hubicka@ucw.cz>
11613 * ipa-profile.c (ipa_propagate_frequency_1): Fix logic skipping calls
11616 2018-01-24 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11618 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11619 Simplify the clause that sets the length attribute.
11620 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11621 (*sibcall_nonlocal_sysv<mode>): Clean up code block; simplify the
11622 clause that sets the length attribute.
11623 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11625 2018-01-24 Tom de Vries <tom@codesourcery.com>
11628 * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG_2): Define to 1.
11629 (nvptx_pc_set, nvptx_condjump_label): New function. Copy from jump.c.
11630 Add strict parameter.
11631 (prevent_branch_around_nothing): Insert dummy insn between branch to
11632 label and label with no ptx insn inbetween.
11633 * config/nvptx/nvptx.md (define_insn "fake_nop"): New insn.
11635 2018-01-24 Tom de Vries <tom@codesourcery.com>
11638 * config/nvptx/nvptx.c (nvptx_single): Add exit insn after noreturn call
11639 for neutered threads in warp.
11640 * config/nvptx/nvptx.md (define_insn "exit"): New insn.
11642 2018-01-24 Richard Biener <rguenther@suse.de>
11644 PR tree-optimization/83176
11645 * tree-chrec.c (chrec_fold_plus_1): Handle (signed T){(T) .. }
11648 2018-01-24 Richard Biener <rguenther@suse.de>
11650 PR tree-optimization/82819
11651 * graphite-isl-ast-to-gimple.c (binary_op_to_tree): Avoid
11652 code generating pluses that are no-ops in the target precision.
11654 2018-01-24 Richard Biener <rguenther@suse.de>
11656 PR middle-end/84000
11657 * tree-cfg.c (replace_loop_annotate): Handle annot_expr_parallel_kind.
11659 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11661 * cfgcleanup.c (try_crossjump_to_edge): Use combine_with_count
11662 to merge probabilities.
11663 * predict.c (probably_never_executed): Also mark as cold functions
11664 with global 0 profile and guessed local profile.
11665 * profile-count.c (profile_probability::combine_with_count): New
11667 * profile-count.h (profile_probability::operator*,
11668 profile_probability::operator*=, profile_probability::operator/,
11669 profile_probability::operator/=): Reduce precision to adjusted
11670 and set value to guessed on contradictory divisions.
11671 (profile_probability::combine_with_freq): Remove.
11672 (profile_probability::combine_wiht_count): Declare.
11673 (profile_count::force_nonzero):: Set to adjusted.
11674 (profile_count::probability_in):: Set quality to adjusted.
11675 * tree-ssa-tail-merge.c (replace_block_by): Use
11676 combine_with_count.
11678 2018-01-23 Andrew Waterman <andrew@sifive.com>
11679 Jim Wilson <jimw@sifive.com>
11681 * config/riscv/riscv.c (riscv_stack_boundary): New.
11682 (riscv_option_override): Set riscv_stack_boundary. Handle
11683 riscv_preferred_stack_boundary_arg.
11684 * config/riscv/riscv.h (MIN_STACK_BOUNDARY, ABI_STACK_BOUNDARY): New.
11685 (BIGGEST_ALIGNMENT): Set to STACK_BOUNDARY.
11686 (STACK_BOUNDARY): Set to riscv_stack_boundary.
11687 (RISCV_STACK_ALIGN): Use STACK_BOUNDARY.
11688 * config/riscv/riscv.opt (mpreferred-stack-boundary): New.
11689 * doc/invoke.tex (RISC-V Options): Add -mpreferred-stack-boundary.
11691 2018-01-23 H.J. Lu <hongjiu.lu@intel.com>
11694 * config/i386/i386.c (ix86_expand_prologue): Use cost reference
11695 of struct ix86_frame.
11696 (ix86_expand_epilogue): Likewise. Add a local variable for
11697 the reg_save_offset field in struct ix86_frame.
11699 2018-01-23 Bin Cheng <bin.cheng@arm.com>
11701 PR tree-optimization/82604
11702 * tree-loop-distribution.c (enum partition_kind): New enum item
11703 PKIND_PARTIAL_MEMSET.
11704 (partition_builtin_p): Support above new enum item.
11705 (generate_code_for_partition): Ditto.
11706 (compute_access_range): Differentiate cases that equality can be
11707 proven at all loops, the innermost loops or no loops.
11708 (classify_builtin_st, classify_builtin_ldst): Adjust call to above
11709 function. Set PKIND_PARTIAL_MEMSET for partition appropriately.
11710 (finalize_partitions, distribute_loop): Don't fuse partition of
11711 PKIND_PARTIAL_MEMSET kind when distributing 3-level loop nest.
11712 (prepare_perfect_loop_nest): Distribute 3-level loop nest only if
11713 parloop is enabled.
11715 2018-01-23 Martin Liska <mliska@suse.cz>
11717 * predict.def (PRED_INDIR_CALL): Set probability to PROB_EVEN in
11718 order to ignore the predictor.
11719 (PRED_POLYMORPHIC_CALL): Likewise.
11720 (PRED_RECURSIVE_CALL): Likewise.
11722 2018-01-23 Martin Liska <mliska@suse.cz>
11724 * tree-profile.c (tree_profiling): Print function header to
11725 aware reader which function we are working on.
11726 * value-prof.c (gimple_find_values_to_profile): Do not print
11727 not interesting value histograms.
11729 2018-01-23 Martin Liska <mliska@suse.cz>
11731 * profile-count.h (enum profile_quality): Add
11732 profile_uninitialized as the first value. Do not number values
11733 as they are zero based.
11734 (profile_count::verify): Update sanity check.
11735 (profile_probability::verify): Likewise.
11737 2018-01-23 Nathan Sidwell <nathan@acm.org>
11739 * doc/invoke.texi (ffor-scope): Deprecate.
11741 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11743 PR tree-optimization/83510
11744 * domwalk.c (set_all_edges_as_executable): New function.
11745 (dom_walker::dom_walker): Convert bool param
11746 "skip_unreachable_blocks" to enum reachability. Move setup of
11747 edge flags to set_all_edges_as_executable and only do it when
11748 reachability is REACHABLE_BLOCKS.
11749 * domwalk.h (enum dom_walker::reachability): New enum.
11750 (dom_walker::dom_walker): Convert bool param
11751 "skip_unreachable_blocks" to enum reachability.
11752 (set_all_edges_as_executable): New decl.
11753 * graphite-scop-detection.c (gather_bbs::gather_bbs): Convert
11754 from false for "skip_unreachable_blocks" to ALL_BLOCKS for
11756 * tree-ssa-dom.c (dom_opt_dom_walker::dom_opt_dom_walker): Likewise,
11757 but converting true to REACHABLE_BLOCKS.
11758 * tree-ssa-sccvn.c (sccvn_dom_walker::sccvn_dom_walker): Likewise.
11760 (check_array_bounds_dom_walker::check_array_bounds_dom_walker):
11761 Likewise, but converting it to REACHABLE_BLOCKS_PRESERVING_FLAGS.
11762 (vrp_dom_walker::vrp_dom_walker): Likewise, but converting it to
11764 (vrp_prop::vrp_finalize): Call set_all_edges_as_executable
11765 if check_all_array_refs will be called.
11767 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11769 * tree.c (selftest::test_location_wrappers): Add more test
11772 2018-01-23 David Malcolm <dmalcolm@redhat.com>
11774 * sbitmap.c (selftest::test_set_range): Fix memory leaks.
11775 (selftest::test_bit_in_range): Likewise.
11777 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
11780 * doc/sourcebuild.texi (vect_float): Say that the selector
11781 only describes the situation when -funsafe-math-optimizations is on.
11782 (vect_float_strict): Document.
11784 2018-01-23 Richard Sandiford <richard.sandiford@linaro.org>
11786 PR tree-optimization/83965
11787 * tree-vect-patterns.c (vect_reassociating_reduction_p): New function.
11788 (vect_recog_dot_prod_pattern, vect_recog_sad_pattern): Use it
11789 instead of checking only for a reduction.
11790 (vect_recog_widen_sum_pattern): Likewise.
11792 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11794 * predict.c (probably_never_executed): Only use precise profile info.
11795 (compute_function_frequency): Skip after inlining hack since we now
11796 have quality checking.
11798 2018-01-23 Jan Hubicka <hubicka@ucw.cz>
11800 * profile-count.h (profile_probability::very_unlikely,
11801 profile_probability::unlikely, profile_probability::even): Set
11802 precision to guessed.
11804 2018-01-23 Richard Biener <rguenther@suse.de>
11806 PR tree-optimization/83963
11807 * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
11808 Properly terminate dominator walk when crossing the exit edge not
11809 when visiting its source block.
11811 2018-01-23 Jakub Jelinek <jakub@redhat.com>
11814 * tree.c (maybe_wrap_with_location): Use NON_LVALUE_EXPR rather than
11815 VIEW_CONVERT_EXPR to wrap CONST_DECLs.
11817 2018-01-22 Jakub Jelinek <jakub@redhat.com>
11819 PR tree-optimization/83957
11820 * omp-expand.c (expand_omp_for_generic): Ignore virtual PHIs. Remove
11821 semicolon after for body surrounded by braces.
11823 PR tree-optimization/83081
11824 * profile-count.h (profile_probability::split): New method.
11825 * dojump.c (do_jump_1) <case TRUTH_ANDIF_EXPR, case TRUTH_ORIF_EXPR>:
11826 Use profile_probability::split.
11827 (do_compare_rtx_and_jump): Fix adjustment of probabilities
11828 when splitting a single conditional jump into 2.
11830 2018-01-22 David Malcolm <dmalcolm@redhat.com>
11832 PR tree-optimization/69452
11833 * tree-ssa-loop-im.c (class move_computations_dom_walker): Remove
11836 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11838 * config/rl78/rl78-expand.md (bswaphi2): New define_expand.
11839 * config/rl78/rl78-virt.md (*bswaphi2_virt): New define_insn.
11840 * config/rl78/rl78-real.md (*bswaphi2_real): New define_insn.
11842 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11844 * config/rl78/rl78-protos.h (rl78_split_movdi): New function
11846 * config/rl78/rl78.md (movdi): New define_expand.
11847 * config/rl78/rl78.c (rl78_split_movdi): New function.
11849 2018-01-22 Michael Meissner <meissner@linux.vnet.ibm.com>
11852 * config/rs6000/rs6000-protos.h (rs6000_split_signbit): Delete,
11854 * config/rs6000/rs6000.c (rs6000_split_signbit): Likewise.
11855 * config/rs6000/rs6000.md (signbit<mode>2): Change code for IEEE
11856 128-bit to produce an UNSPEC move to get the double word with the
11857 signbit and then a shift directly to do signbit.
11858 (signbit<mode>2_dm): Replace old IEEE 128-bit signbit
11859 implementation with a new version that just does either a direct
11860 move or a regular move. Move memory interface to separate insns.
11861 Move insns so they are next to the expander.
11862 (signbit<mode>2_dm_mem_be): New combiner insns to combine load
11863 with signbit move. Split big and little endian case.
11864 (signbit<mode>2_dm_mem_le): Likewise.
11865 (signbit<mode>2_dm_<su>ext): Delete, no longer used.
11866 (signbit<mode>2_dm2): Likewise.
11868 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11870 * config/rl78/rl78.md (anddi3): New define_expand.
11872 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11874 * config/rl78/rl78.md (umindi3): New define_expand.
11876 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11878 * config/rl78/rl78.md (smindi3): New define_expand.
11880 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11882 * config/rl78/rl78.md (smaxdi3): New define_expand.
11884 2018-01-22 Carl Love <cel@us.ibm.com>
11886 * config/rs6000/rs6000-builtin.def (ST_ELEMREV_V1TI, LD_ELEMREV_V1TI,
11887 LVX_V1TI): Add macro expansion.
11888 * config/rs6000/rs6000-c.c (altivec_builtin_types): Add argument
11889 definitions for VSX_BUILTIN_VEC_XST_BE, VSX_BUILTIN_VEC_ST,
11890 VSX_BUILTIN_VEC_XL, LD_ELEMREV_V1TI builtins.
11891 * config/rs6000/rs6000-p8swap.c (insn_is_swappable_p);
11892 Change check to determine if the instruction is a byte reversing
11893 entry. Fix typo in comment.
11894 * config/rs6000/rs6000.c (altivec_expand_builtin): Add case entry
11895 for VSX_BUILTIN_ST_ELEMREV_V1TI and VSX_BUILTIN_LD_ELEMREV_V1TI.
11896 Add def_builtin calls for new builtins.
11897 * config/rs6000/vsx.md (vsx_st_elemrev_v1ti, vsx_ld_elemrev_v1ti):
11898 Add define_insn expansion.
11900 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11902 * config/rl78/rl78.md (umaxdi3): New define_expand.
11904 2018-01-22 Sebastian Perta <sebastian.perta@renesas.com>
11906 * config/rl78/rl78.c (rl78_note_reg_set): Fixed dead reg check
11907 for non-QImode registers.
11909 2018-01-22 Richard Biener <rguenther@suse.de>
11911 PR tree-optimization/83963
11912 * graphite-scop-detection.c (scop_detection::get_sese): Delay
11913 including the loop exit block.
11914 (scop_detection::merge_sese): Likewise.
11915 (scop_detection::add_scop): Do it here instead.
11917 2018-01-22 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
11919 * doc/sourcebuild.texi (arm_softfloat): Document.
11921 2018-01-21 John David Anglin <danglin@gcc.gnu.org>
11924 * config/pa/pa.c (pa_function_ok_for_sibcall): Use
11925 targetm.binds_local_p instead of TREE_PUBLIC to check local binding.
11926 Move TARGET_PORTABLE_RUNTIME check after TARGET_64BIT check.
11928 2018-01-21 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
11929 David Edelsohn <dje.gcc@gmail.com>
11932 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
11933 Change "crset eq" to "crset 2".
11934 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
11935 (*call_indirect_aix<mode>_nospec): Likewise.
11936 (*call_value_indirect_aix<mode>_nospec): Likewise.
11937 (*call_indirect_elfv2<mode>_nospec): Likewise.
11938 (*call_value_indirect_elfv2<mode>_nospec): Likewise.
11939 (*sibcall_nonlocal_sysv<mode>): Change "crset eq" to "crset 2";
11940 change assembly output from . to $.
11941 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
11942 (indirect_jump<mode>_nospec): Change assembly output from . to $.
11943 (*tablejump<mode>_internal1_nospec): Likewise.
11945 2018-01-21 Oleg Endo <olegendo@gcc.gnu.org>
11948 * config/sh/sh_optimize_sett_clrt.cc:
11949 Use INCLUDE_ALGORITHM and INCLUDE_VECTOR instead of direct includes.
11951 2018-01-20 Richard Sandiford <richard.sandiford@linaro.org>
11953 PR tree-optimization/83940
11954 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): Set
11955 offset_dt to vect_constant_def rather than vect_unknown_def_type.
11956 (vect_check_load_store_mask): Add a mask_dt_out parameter and
11957 use it to pass back the definition type.
11958 (vect_check_store_rhs): Likewise rhs_dt_out.
11959 (vect_build_gather_load_calls): Add a mask_dt argument and use
11960 it instead of a call to vect_is_simple_use.
11961 (vectorizable_store): Update calls to vect_check_load_store_mask
11962 and vect_check_store_rhs. Use the dt returned by the latter instead
11963 of scatter_src_dt. Use the cached mask_dt and gs_info.offset_dt
11964 instead of calls to vect_is_simple_use. Pass the scalar rather
11965 than the vector operand to vect_is_simple_use when handling
11966 second and subsequent copies of an rhs value.
11967 (vectorizable_load): Update calls to vect_check_load_store_mask
11968 and vect_build_gather_load_calls. Use the cached mask_dt and
11969 gs_info.offset_dt instead of calls to vect_is_simple_use.
11971 2018-01-20 Jakub Jelinek <jakub@redhat.com>
11973 PR middle-end/83945
11974 * tree-emutls.c: Include gimplify.h.
11975 (lower_emutls_2): New function.
11976 (lower_emutls_1): If ADDR_EXPR is a gimple invariant and walk_tree
11977 with lower_emutls_2 callback finds some TLS decl in it, unshare_expr
11978 it before further processing.
11981 * simplify-rtx.c (simplify_binary_operation_1) <case UMOD>: Use
11982 UINTVAL (trueop1) instead of INTVAL (op1).
11984 2018-01-19 Jakub Jelinek <jakub@redhat.com>
11988 * dwarf2cfi.c (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define to
11989 INCOMING_FRAME_SP_OFFSET if not defined.
11990 (scan_trace): Add ENTRY argument. If true and
11991 DEFAULT_INCOMING_FRAME_SP_OFFSET != INCOMING_FRAME_SP_OFFSET,
11992 emit a note to adjust the CFA offset.
11993 (create_cfi_notes): Adjust scan_trace callers.
11994 (create_cie_data): Use DEFAULT_INCOMING_FRAME_SP_OFFSET rather than
11995 INCOMING_FRAME_SP_OFFSET in the CIE.
11996 * config/i386/i386.h (DEFAULT_INCOMING_FRAME_SP_OFFSET): Define.
11997 * config/stormy16/stormy16.h (DEFAULT_INCOMING_FRAME_SP_OFFSET):
11999 * doc/tm.texi.in (DEFAULT_INCOMING_FRAME_SP_OFFSET): Document.
12000 * doc/tm.texi: Regenerated.
12002 2018-01-19 Andreas Krebbel <krebbel@linux.vnet.ibm.com>
12004 PR rtl-optimization/83147
12005 * lra-constraints.c (remove_inheritance_pseudos): Use
12006 lra_substitute_pseudo_within_insn.
12008 2018-01-19 Tom de Vries <tom@codesourcery.com>
12009 Cesar Philippidis <cesar@codesourcery.com>
12012 * config/nvptx/nvptx.c (nvptx_single): Fix jit workaround.
12014 2018-01-19 Cesar Philippidis <cesar@codesourcery.com>
12017 * config/nvptx/nvptx.c (output_init_frag): Don't use generic address
12018 spaces for function labels.
12020 2018-01-19 Martin Liska <mliska@suse.cz>
12022 * predict.def (PRED_LOOP_EXIT): Change from 85 to 89.
12023 (PRED_LOOP_EXIT_WITH_RECURSION): Change from 72 to 78.
12024 (PRED_LOOP_EXTRA_EXIT): Change from 83 to 67.
12025 (PRED_OPCODE_POSITIVE): Change from 64 to 59.
12026 (PRED_TREE_OPCODE_POSITIVE): Change from 64 to 59.
12027 (PRED_CONST_RETURN): Change from 69 to 65.
12028 (PRED_NULL_RETURN): Change from 91 to 71.
12029 (PRED_LOOP_IV_COMPARE_GUESS): Change from 98 to 64.
12030 (PRED_LOOP_GUARD): Change from 66 to 73.
12032 2018-01-19 Martin Liska <mliska@suse.cz>
12034 * predict.c (predict_insn_def): Add new assert.
12035 (struct branch_predictor): Change type to signed integer.
12036 (test_prediction_value_range): Amend test to cover
12037 PROB_UNINITIALIZED.
12038 * predict.def (PRED_LOOP_ITERATIONS): Use the new constant.
12039 (PRED_LOOP_ITERATIONS_GUESSED): Likewise.
12040 (PRED_LOOP_ITERATIONS_MAX): Likewise.
12041 (PRED_LOOP_IV_COMPARE): Likewise.
12042 * predict.h (PROB_UNINITIALIZED): Define new constant.
12044 2018-01-19 Martin Liska <mliska@suse.cz>
12046 * predict.c (dump_prediction): Add new format for
12047 analyze_brprob.py script which is enabled with -details
12049 * profile-count.h (precise_p): New function.
12051 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
12053 PR tree-optimization/83922
12054 * tree-vect-loop.c (vect_verify_full_masking): Return false if
12055 there are no statements that need masking.
12056 (vect_active_double_reduction_p): New function.
12057 (vect_analyze_loop_operations): Use it when handling phis that
12058 are not in the loop header.
12060 2018-01-19 Richard Sandiford <richard.sandiford@linaro.org>
12062 PR tree-optimization/83914
12063 * tree-vect-loop.c (vectorizable_induction): Don't convert
12064 init_expr or apply the peeling adjustment for inductions
12065 that are nested within the vectorized loop.
12067 2018-01-19 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12069 * config/arm/thumb2.md (*thumb2_negsi2_short): Use RSB mnemonic
12072 2018-01-18 Jakub Jelinek <jakub@redhat.com>
12076 * function.h (gimplify_parameters): Add gimple_seq * argument.
12077 * function.c: Include gimple.h and options.h.
12078 (gimplify_parameters): Add cleanup argument, add CLOBBER stmts
12079 for the added local temporaries if needed.
12080 * gimplify.c (gimplify_body): Adjust gimplify_parameters caller,
12081 if there are any parameter cleanups, wrap whole body into a
12082 try/finally with the cleanups.
12084 2018-01-18 Wilco Dijkstra <wdijkstr@arm.com>
12087 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
12088 Use GET_MODE_CLASS for scalar floating point.
12090 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
12094 * cgraphclones.c (cgraph_node::create_version_clone_with_body):
12095 Fix call of call_cgraph_insertion_hooks.
12097 2018-01-18 Martin Sebor <msebor@redhat.com>
12099 * doc/invoke.texi (-Wclass-memaccess): Tweak text.
12101 2018-01-18 Jan Hubicka <hubicka@ucw.cz>
12104 * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Update edge
12107 2018-01-18 Boris Kolpackov <boris@codesynthesis.com>
12110 * common.opt: (-ffile-prefix-map): New option.
12111 * opts.c (common_handle_option): Defer it.
12112 * opts-global.c (handle_common_deferred_options): Handle it.
12113 * debug.h (remap_debug_filename, add_debug_prefix_map): Move to...
12114 * file-prefix-map.h: New file.
12115 (remap_debug_filename, add_debug_prefix_map): ...here.
12116 (add_macro_prefix_map, add_file_prefix_map, remap_macro_filename): New.
12117 * final.c (debug_prefix_map, add_debug_prefix_map
12118 remap_debug_filename): Move to...
12119 * file-prefix-map.c: New file.
12120 (file_prefix_map, add_prefix_map, remap_filename) ...here and rename,
12121 generalize, get rid of alloca(), use strrchr() instead of strchr().
12122 (add_macro_prefix_map, add_debug_prefix_map, add_file_prefix_map):
12123 Implement in terms of add_prefix_map().
12124 (remap_macro_filename, remap_debug_filename): Implement in term of
12126 * Makefile.in (OBJS, PLUGIN_HEADERS): Add new files.
12127 * builtins.c (fold_builtin_FILE): Call remap_macro_filename().
12128 * dbxout.c: Include file-prefix-map.h.
12129 * varasm.c: Likewise.
12130 * vmsdbgout.c: Likewise.
12131 * xcoffout.c: Likewise.
12132 * dwarf2out.c: Likewise plus omit new options from DW_AT_producer.
12133 * doc/cppopts.texi (-fmacro-prefix-map): Document.
12134 * doc/invoke.texi (-ffile-prefix-map): Document.
12135 (-fdebug-prefix-map): Update description.
12137 2018-01-18 Martin Liska <mliska@suse.cz>
12139 * config/i386/i386.c (indirect_thunk_name): Document that also
12141 (output_indirect_thunk): Document why both instructions
12142 (pause and lfence) are generated.
12144 2018-01-18 Richard Biener <rguenther@suse.de>
12146 PR tree-optimization/83887
12147 * graphite-scop-detection.c
12148 (scop_detection::get_nearest_dom_with_single_entry): Remove.
12149 (scop_detection::get_nearest_pdom_with_single_exit): Likewise.
12150 (scop_detection::merge_sese): Re-implement with a flood-fill
12151 algorithm that properly finds a SESE region if it exists.
12153 2018-01-18 Jakub Jelinek <jakub@redhat.com>
12156 * match.pd ((P + A) - P, P - (P + A), (P + A) - (P + B)): For
12157 pointer_diff optimizations use view_convert instead of convert.
12159 2018-01-17 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12161 * config/rs6000/rs6000.md (*call_indirect_nonlocal_sysv<mode>):
12162 Generate different code for -mno-speculate-indirect-jumps.
12163 (*call_value_indirect_nonlocal_sysv<mode>): Likewise.
12164 (*call_indirect_aix<mode>): Disable for
12165 -mno-speculate-indirect-jumps.
12166 (*call_indirect_aix<mode>_nospec): New define_insn.
12167 (*call_value_indirect_aix<mode>): Disable for
12168 -mno-speculate-indirect-jumps.
12169 (*call_value_indirect_aix<mode>_nospec): New define_insn.
12170 (*sibcall_nonlocal_sysv<mode>): Generate different code for
12171 -mno-speculate-indirect-jumps.
12172 (*sibcall_value_nonlocal_sysv<mode>): Likewise.
12174 2018-01-17 Michael Meissner <meissner@linux.vnet.ibm.com>
12176 * config/rs6000/rs6000.c (rs6000_emit_move): If we load or store a
12177 long double type, set the flags for noting the default long double
12178 type, even if we don't pass or return a long double type.
12180 2018-01-17 Jan Hubicka <hubicka@ucw.cz>
12183 * ipa-inline.c (flatten_function): Do not overwrite final inlining
12186 2018-01-17 Will Schmidt <will_schmidt@vnet.ibm.com>
12188 * config/rs6000/rs6000.c (rs6000_gimple_builtin): Add gimple folding
12189 support for merge[hl].
12190 (fold_mergehl_helper): New helper function.
12191 (tree-vector-builder.h): New #include for tree_vector_builder usage.
12192 * config/rs6000/altivec.md (altivec_vmrghw_direct): Add xxmrghw insn.
12193 (altivec_vmrglw_direct): Add xxmrglw insn.
12195 2018-01-17 Andrew Waterman <andrew@sifive.com>
12197 * config/riscv/riscv.c (riscv_conditional_register_usage): If
12198 UNITS_PER_FP_ARG is 0, set call_used_regs to 1 for all FP regs.
12200 2018-01-17 David Malcolm <dmalcolm@redhat.com>
12203 * ipa-devirt.c (add_type_duplicate): When comparing memory layout,
12204 call the lto_location_cache before reading the
12205 DECL_SOURCE_LOCATION of the types.
12207 2018-01-17 Wilco Dijkstra <wdijkstr@arm.com>
12208 Richard Sandiford <richard.sandiford@linaro.org>
12210 * config/aarch64/aarch64.md (movti_aarch64): Use Uti constraint.
12211 * config/aarch64/aarch64.c (aarch64_mov128_immediate): New function.
12212 (aarch64_legitimate_constant_p): Just support CONST_DOUBLE
12213 SF/DF/TF mode to avoid creating illegal CONST_WIDE_INT immediates.
12214 * config/aarch64/aarch64-protos.h (aarch64_mov128_immediate):
12216 * config/aarch64/constraints.md (aarch64_movti_operand):
12218 * config/aarch64/predicates.md (Uti): Add new constraint.
12220 2018-01-17 Carl Love <cel@us.ibm.com>
12222 * config/rs6000/vsx.md (define_expand xl_len_r,
12223 define_expand stxvl, define_expand *stxvl): Add match_dup argument.
12224 (define_insn): Add, match_dup 1 argument to define_insn stxvll and
12226 (define_expand, define_insn): Move the shift left from the
12227 define_insn to the define_expand for lxvl and stxvl instructions.
12228 * config/rs6000/rs6000-builtin.def (BU_P9V_64BIT_VSX_2): Change LXVL
12229 and XL_LEN_R definitions to PURE.
12231 2018-01-17 Uros Bizjak <ubizjak@gmail.com>
12233 * config/i386/i386.c (indirect_thunk_name): Declare regno
12234 as unsigned int. Compare regno with INVALID_REGNUM.
12235 (output_indirect_thunk): Ditto.
12236 (output_indirect_thunk_function): Ditto.
12237 (ix86_code_end): Declare regno as unsigned int. Use INVALID_REGNUM
12238 in the call to output_indirect_thunk_function.
12240 2018-01-17 Richard Sandiford <richard.sandiford@linaro.org>
12242 PR middle-end/83884
12243 * expr.c (expand_expr_real_1): Use the size of GET_MODE (op0)
12244 rather than the size of inner_type to determine the stack slot size
12245 when handling VIEW_CONVERT_EXPRs on strict-alignment targets.
12247 2018-01-16 Sebastian Peryt <sebastian.peryt@intel.com>
12250 * config/i386/i386.c (ix86_option_override_internal): Add PTA_RDRND
12253 2018-01-16 Michael Meissner <meissner@linux.vnet.ibm.com>
12255 * config.gcc (powerpc*-linux*-*): Add support for 64-bit little
12256 endian Linux systems to optionally enable multilibs for selecting
12257 the long double type if the user configured an explicit type.
12258 * config/rs6000/rs6000.h (TARGET_IEEEQUAD_MULTILIB): Indicate we
12259 have no long double multilibs if not defined.
12260 * config/rs6000/rs6000.c (rs6000_option_override_internal): Do not
12261 warn if the user used -mabi={ieee,ibm}longdouble and we built
12262 multilibs for long double.
12263 * config/rs6000/linux64.h (MULTILIB_DEFAULTS_IEEE): Define as the
12264 appropriate multilib option.
12265 (MULTILIB_DEFAULTS): Add MULTILIB_DEFAULTS_IEEE to the default
12267 * config/rs6000/t-ldouble-linux64le-ibm: New configuration files
12268 for building long double multilibs.
12269 * config/rs6000/t-ldouble-linux64le-ieee: Likewise.
12271 2018-01-16 John David Anglin <danglin@gcc.gnu.org>
12273 * config.gcc (hppa*-*-linux*): Change callee copies ABI to caller
12276 * config/pa.h (MALLOC_ABI_ALIGNMENT): Set 32-bit alignment default to
12278 * config/pa/pa32-linux.h (MALLOC_ABI_ALIGNMENT): Set alignment to
12281 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Cleanup type and mode
12284 * config/pa/pa.c (pa_function_arg_size): Apply CEIL to GET_MODE_SIZE
12287 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
12289 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): For an
12290 ADDR_EXPR, do not count the offset of a COMPONENT_REF twice.
12292 2018-01-16 Kelvin Nilsen <kelvin@gcc.gnu.org>
12294 * config/rs6000/rs6000-p8swap.c (rs6000_gen_stvx): Generate
12295 different rtl trees depending on TARGET_64BIT.
12296 (rs6000_gen_lvx): Likewise.
12298 2018-01-16 Eric Botcazou <ebotcazou@adacore.com>
12300 * config/visium/visium.md (nop): Tweak comment.
12301 (hazard_nop): Likewise.
12303 2018-01-16 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
12305 * config/rs6000/rs6000.c (rs6000_opt_vars): Add entry for
12306 -mspeculate-indirect-jumps.
12307 * config/rs6000/rs6000.md (*call_indirect_elfv2<mode>): Disable
12308 for -mno-speculate-indirect-jumps.
12309 (*call_indirect_elfv2<mode>_nospec): New define_insn.
12310 (*call_value_indirect_elfv2<mode>): Disable for
12311 -mno-speculate-indirect-jumps.
12312 (*call_value_indirect_elfv2<mode>_nospec): New define_insn.
12313 (indirect_jump): Emit different RTL for
12314 -mno-speculate-indirect-jumps.
12315 (*indirect_jump<mode>): Disable for
12316 -mno-speculate-indirect-jumps.
12317 (*indirect_jump<mode>_nospec): New define_insn.
12318 (tablejump): Emit different RTL for
12319 -mno-speculate-indirect-jumps.
12320 (tablejumpsi): Disable for -mno-speculate-indirect-jumps.
12321 (tablejumpsi_nospec): New define_expand.
12322 (tablejumpdi): Disable for -mno-speculate-indirect-jumps.
12323 (tablejumpdi_nospec): New define_expand.
12324 (*tablejump<mode>_internal1): Disable for
12325 -mno-speculate-indirect-jumps.
12326 (*tablejump<mode>_internal1_nospec): New define_insn.
12327 * config/rs6000/rs6000.opt (mspeculate-indirect-jumps): New
12330 2018-01-16 Artyom Skrobov tyomitch@gmail.com
12332 * caller-save.c (insert_save): Drop unnecessary parameter. All
12335 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12336 Richard Biener <rguenth@suse.de>
12339 * gimplify.c (gimplify_one_sizepos): For is_gimple_constant (expr)
12340 return early, inline manually is_gimple_sizepos. Make sure if we
12341 call gimplify_expr we don't end up with a gimple constant.
12342 * tree.c (variably_modified_type_p): Don't return true for
12343 is_gimple_constant (_t). Inline manually is_gimple_sizepos.
12344 * gimplify.h (is_gimple_sizepos): Remove.
12346 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12348 PR tree-optimization/83857
12349 * tree-vect-loop.c (vect_analyze_loop_operations): Don't call
12350 vectorizable_live_operation for pure SLP statements.
12351 (vectorizable_live_operation): Handle PHIs.
12353 2018-01-16 Richard Biener <rguenther@suse.de>
12355 PR tree-optimization/83867
12356 * tree-vect-stmts.c (vect_transform_stmt): Precompute
12357 nested_in_vect_loop_p since the scalar stmt may get invalidated.
12359 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12362 * stor-layout.c (handle_warn_if_not_align): Use byte_position and
12363 multiple_of_p instead of unchecked tree_to_uhwi and UHWI check.
12364 If off is not INTEGER_CST, issue a may not be aligned warning
12365 rather than isn't aligned. Use isn%'t rather than isn't.
12366 * fold-const.c (multiple_of_p) <case BIT_AND_EXPR>: Don't fall through
12368 <case MULT_EXPR>: Improve the case when bottom and one of the
12369 MULT_EXPR operands are INTEGER_CSTs and bottom is multiple of that
12370 operand, in that case check if the other operand is multiple of
12371 bottom divided by the INTEGER_CST operand.
12373 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12376 * config/pa/pa.h (FUNCTION_ARG_SIZE): Delete.
12377 * config/pa/pa-protos.h (pa_function_arg_size): Declare.
12378 * config/pa/som.h (ASM_DECLARE_FUNCTION_NAME): Use
12379 pa_function_arg_size instead of FUNCTION_ARG_SIZE.
12380 * config/pa/pa.c (pa_function_arg_advance): Likewise.
12381 (pa_function_arg, pa_arg_partial_bytes): Likewise.
12382 (pa_function_arg_size): New function.
12384 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12386 * fold-const.c (fold_ternary_loc): Construct the vec_perm_indices
12387 in a separate statement.
12389 2018-01-16 Richard Sandiford <richard.sandiford@linaro.org>
12391 PR tree-optimization/83847
12392 * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Don't
12393 group gathers and scatters.
12395 2018-01-16 Jakub Jelinek <jakub@redhat.com>
12397 PR rtl-optimization/86620
12398 * params.def (max-sched-ready-insns): Bump minimum value to 1.
12400 PR rtl-optimization/83213
12401 * recog.c (peep2_attempt): Copy over CROSSING_JUMP_P from peepinsn
12402 to last if both are JUMP_INSNs.
12404 PR tree-optimization/83843
12405 * gimple-ssa-store-merging.c
12406 (imm_store_chain_info::output_merged_store): Handle bit_not_p on
12407 store_immediate_info for bswap/nop orig_stores.
12409 2018-01-15 Andrew Waterman <andrew@sifive.com>
12411 * config/riscv/riscv.c (riscv_rtx_costs) <MULT>: Increase cost if
12413 <UDIV>: Increase cost if !TARGET_DIV.
12415 2018-01-15 Segher Boessenkool <segher@kernel.crashing.org>
12417 * config/rs6000/rs6000.md (define_attr "type"): Remove delayed_cr.
12418 (define_attr "cr_logical_3op"): New.
12419 (cceq_ior_compare): Adjust.
12420 (cceq_ior_compare_complement): Adjust.
12421 (*cceq_rev_compare): Adjust.
12422 * config/rs6000/rs6000.c (rs6000_adjust_cost): Adjust.
12423 (is_cracked_insn): Adjust.
12424 (insn_must_be_first_in_group): Adjust.
12425 * config/rs6000/40x.md: Adjust.
12426 * config/rs6000/440.md: Adjust.
12427 * config/rs6000/476.md: Adjust.
12428 * config/rs6000/601.md: Adjust.
12429 * config/rs6000/603.md: Adjust.
12430 * config/rs6000/6xx.md: Adjust.
12431 * config/rs6000/7450.md: Adjust.
12432 * config/rs6000/7xx.md: Adjust.
12433 * config/rs6000/8540.md: Adjust.
12434 * config/rs6000/cell.md: Adjust.
12435 * config/rs6000/e300c2c3.md: Adjust.
12436 * config/rs6000/e500mc.md: Adjust.
12437 * config/rs6000/e500mc64.md: Adjust.
12438 * config/rs6000/e5500.md: Adjust.
12439 * config/rs6000/e6500.md: Adjust.
12440 * config/rs6000/mpc.md: Adjust.
12441 * config/rs6000/power4.md: Adjust.
12442 * config/rs6000/power5.md: Adjust.
12443 * config/rs6000/power6.md: Adjust.
12444 * config/rs6000/power7.md: Adjust.
12445 * config/rs6000/power8.md: Adjust.
12446 * config/rs6000/power9.md: Adjust.
12447 * config/rs6000/rs64.md: Adjust.
12448 * config/rs6000/titan.md: Adjust.
12450 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12452 * config/i386/predicates.md (indirect_branch_operand): Rewrite
12453 ix86_indirect_branch_register logic.
12455 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12457 * config/i386/constraints.md (Bs): Update
12458 ix86_indirect_branch_register check. Don't check
12459 ix86_indirect_branch_register with GOT_memory_operand.
12461 * config/i386/predicates.md (GOT_memory_operand): Don't check
12462 ix86_indirect_branch_register here.
12463 (GOT32_symbol_operand): Likewise.
12465 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12467 * config/i386/predicates.md (constant_call_address_operand):
12468 Rewrite ix86_indirect_branch_register logic.
12469 (sibcall_insn_operand): Likewise.
12471 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12473 * config/i386/constraints.md (Bs): Replace
12474 ix86_indirect_branch_thunk_register with
12475 ix86_indirect_branch_register.
12477 * config/i386/i386.md (indirect_jump): Likewise.
12478 (tablejump): Likewise.
12479 (*sibcall_memory): Likewise.
12480 (*sibcall_value_memory): Likewise.
12481 Peepholes of indirect call and jump via memory: Likewise.
12482 * config/i386/i386.opt: Likewise.
12483 * config/i386/predicates.md (indirect_branch_operand): Likewise.
12484 (GOT_memory_operand): Likewise.
12485 (call_insn_operand): Likewise.
12486 (sibcall_insn_operand): Likewise.
12487 (GOT32_symbol_operand): Likewise.
12489 2018-01-15 Jakub Jelinek <jakub@redhat.com>
12491 PR middle-end/83837
12492 * omp-expand.c (expand_omp_atomic_pipeline): Use loaded_val
12493 type rather than type addr's type points to.
12494 (expand_omp_atomic_mutex): Likewise.
12495 (expand_omp_atomic): Likewise.
12497 2018-01-15 H.J. Lu <hongjiu.lu@intel.com>
12500 * config/i386/i386.c (output_indirect_thunk_function): Use
12501 ASM_OUTPUT_LABEL, instead of ASM_OUTPUT_DEF, for TARGET_MACHO
12502 for __x86_return_thunk.
12504 2018-01-15 Richard Biener <rguenther@suse.de>
12506 PR middle-end/83850
12507 * expmed.c (extract_bit_field_1): Fix typo.
12509 2018-01-15 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
12512 * config/arm/iterators.md (VF): New mode iterator.
12513 * config/arm/neon.md (neon_vabd<mode>_2): Use the above.
12514 Remove integer-related logic from pattern.
12515 (neon_vabd<mode>_3): Likewise.
12517 2018-01-15 Jakub Jelinek <jakub@redhat.com>
12519 PR middle-end/82694
12520 * common.opt (fstrict-overflow): No longer an alias.
12521 (fwrapv-pointer): New option.
12522 * tree.h (TYPE_OVERFLOW_WRAPS, TYPE_OVERFLOW_UNDEFINED): Define
12523 also for pointer types based on flag_wrapv_pointer.
12524 * opts.c (common_handle_option) <case OPT_fstrict_overflow>: Set
12525 opts->x_flag_wrap[pv] to !value, clear opts->x_flag_trapv if
12526 opts->x_flag_wrapv got set.
12527 * fold-const.c (fold_comparison, fold_binary_loc): Revert 2017-08-01
12528 changes, just use TYPE_OVERFLOW_UNDEFINED on pointer type instead of
12529 POINTER_TYPE_OVERFLOW_UNDEFINED.
12530 * match.pd: Likewise in address comparison pattern.
12531 * doc/invoke.texi: Document -fwrapv and -fstrict-overflow.
12533 2018-01-15 Richard Biener <rguenther@suse.de>
12536 * tree.c (free_lang_data_in_type): Always unlink TYPE_DECLs
12537 from TYPE_FIELDS. Free TYPE_BINFO if not used by devirtualization.
12538 Reset type names to their identifier if their TYPE_DECL doesn't
12539 have linkage (and thus is used for ODR and devirt).
12540 (save_debug_info_for_decl): Remove.
12541 (save_debug_info_for_type): Likewise.
12542 (add_tree_to_fld_list): Adjust.
12543 * tree-pretty-print.c (dump_generic_node): Make dumping of
12544 type names more robust.
12546 2018-01-15 Richard Biener <rguenther@suse.de>
12548 * BASE-VER: Bump to 8.0.1.
12550 2018-01-14 Martin Sebor <msebor@redhat.com>
12553 * builtins.c (check_access): Avoid warning when the no-warning bit
12556 2018-01-14 Cory Fields <cory-nospam-@coryfields.com>
12558 * tree-ssa-loop-im.c (sort_bbs_in_loop_postorder_cmp): Stabilize sort.
12559 * ira-color (allocno_hard_regs_compare): Likewise.
12561 2018-01-14 Nathan Rossi <nathan@nathanrossi.com>
12564 * config/microblaze/microblaze.c (microblaze_asm_output_ident):
12565 Use .pushsection/.popsection.
12567 2018-01-14 Martin Sebor <msebor@redhat.com>
12570 * doc/invoke.texi (-Wlass-memaccess): Document suppression by casting.
12572 2018-01-14 Jakub Jelinek <jakub@redhat.com>
12574 * config.gcc (i[34567]86-*-*): Remove one duplicate gfniintrin.h
12575 entry from extra_headers.
12576 (x86_64-*-*): Remove two duplicate gfniintrin.h entries from
12577 extra_headers, make the list bitwise identical to the i?86-*-* one.
12579 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12581 * config/i386/i386.c (ix86_set_indirect_branch_type): Disallow
12582 -mcmodel=large with -mindirect-branch=thunk,
12583 -mindirect-branch=thunk-extern, -mfunction-return=thunk and
12584 -mfunction-return=thunk-extern.
12585 * doc/invoke.texi: Document -mcmodel=large is incompatible with
12586 -mindirect-branch=thunk, -mindirect-branch=thunk-extern,
12587 -mfunction-return=thunk and -mfunction-return=thunk-extern.
12589 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12591 * config/i386/i386.c (print_reg): Print the name of the full
12592 integer register without '%'.
12593 (ix86_print_operand): Handle 'V'.
12594 * doc/extend.texi: Document 'V' modifier.
12596 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12598 * config/i386/constraints.md (Bs): Disallow memory operand for
12599 -mindirect-branch-register.
12601 * config/i386/predicates.md (indirect_branch_operand): Likewise.
12602 (GOT_memory_operand): Likewise.
12603 (call_insn_operand): Likewise.
12604 (sibcall_insn_operand): Likewise.
12605 (GOT32_symbol_operand): Likewise.
12606 * config/i386/i386.md (indirect_jump): Call convert_memory_address
12607 for -mindirect-branch-register.
12608 (tablejump): Likewise.
12609 (*sibcall_memory): Likewise.
12610 (*sibcall_value_memory): Likewise.
12611 Disallow peepholes of indirect call and jump via memory for
12612 -mindirect-branch-register.
12613 (*call_pop): Replace m with Bw.
12614 (*call_value_pop): Likewise.
12615 (*sibcall_pop_memory): Replace m with Bs.
12616 * config/i386/i386.opt (mindirect-branch-register): New option.
12617 * doc/invoke.texi: Document -mindirect-branch-register option.
12619 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12621 * config/i386/i386-protos.h (ix86_output_function_return): New.
12622 * config/i386/i386.c (ix86_set_indirect_branch_type): Also
12623 set function_return_type.
12624 (indirect_thunk_name): Add ret_p to indicate thunk for function
12626 (output_indirect_thunk_function): Pass false to
12627 indirect_thunk_name.
12628 (ix86_output_indirect_branch_via_reg): Likewise.
12629 (ix86_output_indirect_branch_via_push): Likewise.
12630 (output_indirect_thunk_function): Create alias for function
12631 return thunk if regno < 0.
12632 (ix86_output_function_return): New function.
12633 (ix86_handle_fndecl_attribute): Handle function_return.
12634 (ix86_attribute_table): Add function_return.
12635 * config/i386/i386.h (machine_function): Add
12636 function_return_type.
12637 * config/i386/i386.md (simple_return_internal): Use
12638 ix86_output_function_return.
12639 (simple_return_internal_long): Likewise.
12640 * config/i386/i386.opt (mfunction-return=): New option.
12641 (indirect_branch): Mention -mfunction-return=.
12642 * doc/extend.texi: Document function_return function attribute.
12643 * doc/invoke.texi: Document -mfunction-return= option.
12645 2018-01-14 H.J. Lu <hongjiu.lu@intel.com>
12647 * config/i386/i386-opts.h (indirect_branch): New.
12648 * config/i386/i386-protos.h (ix86_output_indirect_jmp): Likewise.
12649 * config/i386/i386.c (ix86_using_red_zone): Disallow red-zone
12650 with local indirect jump when converting indirect call and jump.
12651 (ix86_set_indirect_branch_type): New.
12652 (ix86_set_current_function): Call ix86_set_indirect_branch_type.
12653 (indirectlabelno): New.
12654 (indirect_thunk_needed): Likewise.
12655 (indirect_thunk_bnd_needed): Likewise.
12656 (indirect_thunks_used): Likewise.
12657 (indirect_thunks_bnd_used): Likewise.
12658 (INDIRECT_LABEL): Likewise.
12659 (indirect_thunk_name): Likewise.
12660 (output_indirect_thunk): Likewise.
12661 (output_indirect_thunk_function): Likewise.
12662 (ix86_output_indirect_branch_via_reg): Likewise.
12663 (ix86_output_indirect_branch_via_push): Likewise.
12664 (ix86_output_indirect_branch): Likewise.
12665 (ix86_output_indirect_jmp): Likewise.
12666 (ix86_code_end): Call output_indirect_thunk_function if needed.
12667 (ix86_output_call_insn): Call ix86_output_indirect_branch if
12669 (ix86_handle_fndecl_attribute): Handle indirect_branch.
12670 (ix86_attribute_table): Add indirect_branch.
12671 * config/i386/i386.h (machine_function): Add indirect_branch_type
12672 and has_local_indirect_jump.
12673 * config/i386/i386.md (indirect_jump): Set has_local_indirect_jump
12675 (tablejump): Likewise.
12676 (*indirect_jump): Use ix86_output_indirect_jmp.
12677 (*tablejump_1): Likewise.
12678 (simple_return_indirect_internal): Likewise.
12679 * config/i386/i386.opt (mindirect-branch=): New option.
12680 (indirect_branch): New.
12683 (thunk-inline): Likewise.
12684 (thunk-extern): Likewise.
12685 * doc/extend.texi: Document indirect_branch function attribute.
12686 * doc/invoke.texi: Document -mindirect-branch= option.
12688 2018-01-14 Jan Hubicka <hubicka@ucw.cz>
12691 * ipa-inline.c (edge_badness): Tolerate roundoff errors.
12693 2018-01-14 Richard Sandiford <richard.sandiford@linaro.org>
12695 * ipa-inline.c (want_inline_small_function_p): Return false if
12696 inlining has already failed with CIF_FINAL_ERROR.
12697 (update_caller_keys): Call want_inline_small_function_p before
12699 (update_callee_keys): Likewise.
12701 2018-01-10 Kelvin Nilsen <kelvin@gcc.gnu.org>
12703 * config/rs6000/rs6000-p8swap.c (rs6000_sum_of_two_registers_p):
12705 (rs6000_quadword_masked_address_p): Likewise.
12706 (quad_aligned_load_p): Likewise.
12707 (quad_aligned_store_p): Likewise.
12708 (const_load_sequence_p): Add comment to describe the outer-most loop.
12709 (mimic_memory_attributes_and_flags): New function.
12710 (rs6000_gen_stvx): Likewise.
12711 (replace_swapped_aligned_store): Likewise.
12712 (rs6000_gen_lvx): Likewise.
12713 (replace_swapped_aligned_load): Likewise.
12714 (replace_swapped_load_constant): Capitalize argument name in
12715 comment describing this function.
12716 (rs6000_analyze_swaps): Add a third pass to search for vector loads
12717 and stores that access quad-word aligned addresses and replace
12718 with stvx or lvx instructions when appropriate.
12719 * config/rs6000/rs6000-protos.h (rs6000_sum_of_two_registers_p):
12720 New function prototype.
12721 (rs6000_quadword_masked_address_p): Likewise.
12722 (rs6000_gen_lvx): Likewise.
12723 (rs6000_gen_stvx): Likewise.
12724 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode>): For modes
12725 VSX_D (V2DF, V2DI), modify this split to select lvx instruction
12726 when memory address is aligned.
12727 (*vsx_le_perm_load_<mode>): For modes VSX_W (V4SF, V4SI), modify
12728 this split to select lvx instruction when memory address is aligned.
12729 (*vsx_le_perm_load_v8hi): Modify this split to select lvx
12730 instruction when memory address is aligned.
12731 (*vsx_le_perm_load_v16qi): Likewise.
12732 (four unnamed splitters): Modify to select the stvx instruction
12733 when memory is aligned.
12735 2018-01-13 Jan Hubicka <hubicka@ucw.cz>
12737 * predict.c (determine_unlikely_bbs): Handle correctly BBs
12738 which appears in the queue multiple times.
12740 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12741 Alan Hayward <alan.hayward@arm.com>
12742 David Sherwood <david.sherwood@arm.com>
12744 * tree-vectorizer.h (vec_lower_bound): New structure.
12745 (_loop_vec_info): Add check_nonzero and lower_bounds.
12746 (LOOP_VINFO_CHECK_NONZERO): New macro.
12747 (LOOP_VINFO_LOWER_BOUNDS): Likewise.
12748 (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Check lower_bounds too.
12749 * tree-data-ref.h (dr_with_seg_len): Add access_size and align
12750 fields. Make seg_len the distance travelled, not including the
12752 (dr_direction_indicator): Declare.
12753 (dr_zero_step_indicator): Likewise.
12754 (dr_known_forward_stride_p): Likewise.
12755 * tree-data-ref.c: Include stringpool.h, tree-vrp.h and
12757 (runtime_alias_check_p): Allow runtime alias checks with
12759 (operator ==): Compare access_size and align.
12760 (prune_runtime_alias_test_list): Rework for new distinction between
12761 the access_size and seg_len.
12762 (create_intersect_range_checks_index): Likewise. Cope with polynomial
12764 (get_segment_min_max): New function.
12765 (create_intersect_range_checks): Use it.
12766 (dr_step_indicator): New function.
12767 (dr_direction_indicator): Likewise.
12768 (dr_zero_step_indicator): Likewise.
12769 (dr_known_forward_stride_p): Likewise.
12770 * tree-loop-distribution.c (data_ref_segment_size): Return
12771 DR_STEP * (niters - 1).
12772 (compute_alias_check_pairs): Update call to the dr_with_seg_len
12774 * tree-vect-data-refs.c (vect_check_nonzero_value): New function.
12775 (vect_preserves_scalar_order_p): New function, split out from...
12776 (vect_analyze_data_ref_dependence): ...here. Check for zero steps.
12777 (vect_vfa_segment_size): Return DR_STEP * (length_factor - 1).
12778 (vect_vfa_access_size): New function.
12779 (vect_vfa_align): Likewise.
12780 (vect_compile_time_alias): Take access_size_a and access_b arguments.
12781 (dump_lower_bound): New function.
12782 (vect_check_lower_bound): Likewise.
12783 (vect_small_gap_p): Likewise.
12784 (vectorizable_with_step_bound_p): Likewise.
12785 (vect_prune_runtime_alias_test_list): Ignore cross-iteration
12786 depencies if the vectorization factor is 1. Convert the checks
12787 for nonzero steps into checks on the bounds of DR_STEP. Try using
12788 a bunds check for variable steps if the minimum required step is
12789 relatively small. Update calls to the dr_with_seg_len
12790 constructor and to vect_compile_time_alias.
12791 * tree-vect-loop-manip.c (vect_create_cond_for_lower_bounds): New
12793 (vect_loop_versioning): Call it.
12794 * tree-vect-loop.c (vect_analyze_loop_2): Clear LOOP_VINFO_LOWER_BOUNDS
12796 (vect_estimate_min_profitable_iters): Account for any bounds checks.
12798 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12799 Alan Hayward <alan.hayward@arm.com>
12800 David Sherwood <david.sherwood@arm.com>
12802 * doc/sourcebuild.texi (vect_scatter_store): Document.
12803 * optabs.def (scatter_store_optab, mask_scatter_store_optab): New
12805 * doc/md.texi (scatter_store@var{m}, mask_scatter_store@var{m}):
12807 * genopinit.c (main): Add supports_vec_scatter_store and
12808 supports_vec_scatter_store_cached to target_optabs.
12809 * gimple.h (gimple_expr_type): Handle IFN_SCATTER_STORE and
12810 IFN_MASK_SCATTER_STORE.
12811 * internal-fn.def (SCATTER_STORE, MASK_SCATTER_STORE): New internal
12813 * internal-fn.h (internal_store_fn_p): Declare.
12814 (internal_fn_stored_value_index): Likewise.
12815 * internal-fn.c (scatter_store_direct): New macro.
12816 (expand_scatter_store_optab_fn): New function.
12817 (direct_scatter_store_optab_supported_p): New macro.
12818 (internal_store_fn_p): New function.
12819 (internal_gather_scatter_fn_p): Handle IFN_SCATTER_STORE and
12820 IFN_MASK_SCATTER_STORE.
12821 (internal_fn_mask_index): Likewise.
12822 (internal_fn_stored_value_index): New function.
12823 (internal_gather_scatter_fn_supported_p): Adjust operand numbers
12824 for scatter stores.
12825 * optabs-query.h (supports_vec_scatter_store_p): Declare.
12826 * optabs-query.c (supports_vec_scatter_store_p): New function.
12827 * tree-vectorizer.h (vect_get_store_rhs): Declare.
12828 * tree-vect-data-refs.c (vect_analyze_data_ref_access): Return
12829 true for scatter stores.
12830 (vect_gather_scatter_fn_p): Handle scatter stores too.
12831 (vect_check_gather_scatter): Consider using scatter stores if
12832 supports_vec_scatter_store_p.
12833 * tree-vect-patterns.c (vect_try_gather_scatter_pattern): Handle
12834 scatter stores too.
12835 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
12836 internal_fn_stored_value_index.
12837 (check_load_store_masking): Handle scatter stores too.
12838 (vect_get_store_rhs): Make public.
12839 (vectorizable_call): Use internal_store_fn_p.
12840 (vectorizable_store): Handle scatter store internal functions.
12841 (vect_transform_stmt): Compare GROUP_STORE_COUNT with GROUP_SIZE
12842 when deciding whether the end of the group has been reached.
12843 * config/aarch64/aarch64.md (UNSPEC_ST1_SCATTER): New unspec.
12844 * config/aarch64/aarch64-sve.md (scatter_store<mode>): New expander.
12845 (mask_scatter_store<mode>): New insns.
12847 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12848 Alan Hayward <alan.hayward@arm.com>
12849 David Sherwood <david.sherwood@arm.com>
12851 * tree-vectorizer.h (vect_gather_scatter_fn_p): Declare.
12852 * tree-vect-data-refs.c (vect_gather_scatter_fn_p): Make public.
12853 * tree-vect-stmts.c (vect_truncate_gather_scatter_offset): New
12855 (vect_use_strided_gather_scatters_p): Take a masked_p argument.
12856 Use vect_truncate_gather_scatter_offset if we can't treat the
12857 operation as a normal gather load or scatter store.
12858 (get_group_load_store_type): Take the gather_scatter_info
12859 as argument. Try using a gather load or scatter store for
12860 single-element groups.
12861 (get_load_store_type): Update calls to get_group_load_store_type
12862 and vect_use_strided_gather_scatters_p.
12864 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12865 Alan Hayward <alan.hayward@arm.com>
12866 David Sherwood <david.sherwood@arm.com>
12868 * tree-vectorizer.h (vect_create_data_ref_ptr): Take an extra
12869 optional tree argument.
12870 * tree-vect-data-refs.c (vect_check_gather_scatter): Check for
12872 (vect_create_data_ref_ptr): Take the iv_step as an optional argument,
12873 but continue to use the current value as a fallback.
12874 (bump_vector_ptr): Use operand_equal_p rather than tree_int_cst_compare
12875 to compare the updates.
12876 * tree-vect-stmts.c (vect_use_strided_gather_scatters_p): New function.
12877 (get_load_store_type): Use it when handling a strided access.
12878 (vect_get_strided_load_store_ops): New function.
12879 (vect_get_data_ptr_increment): Likewise.
12880 (vectorizable_load): Handle strided gather loads. Always pass
12881 a step to vect_create_data_ref_ptr and bump_vector_ptr.
12883 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12884 Alan Hayward <alan.hayward@arm.com>
12885 David Sherwood <david.sherwood@arm.com>
12887 * doc/md.texi (gather_load@var{m}): Document.
12888 (mask_gather_load@var{m}): Likewise.
12889 * genopinit.c (main): Add supports_vec_gather_load and
12890 supports_vec_gather_load_cached to target_optabs.
12891 * optabs-tree.c (init_tree_optimization_optabs): Use
12892 ggc_cleared_alloc to allocate target_optabs.
12893 * optabs.def (gather_load_optab, mask_gather_laod_optab): New optabs.
12894 * internal-fn.def (GATHER_LOAD, MASK_GATHER_LOAD): New internal
12896 * internal-fn.h (internal_load_fn_p): Declare.
12897 (internal_gather_scatter_fn_p): Likewise.
12898 (internal_fn_mask_index): Likewise.
12899 (internal_gather_scatter_fn_supported_p): Likewise.
12900 * internal-fn.c (gather_load_direct): New macro.
12901 (expand_gather_load_optab_fn): New function.
12902 (direct_gather_load_optab_supported_p): New macro.
12903 (direct_internal_fn_optab): New function.
12904 (internal_load_fn_p): Likewise.
12905 (internal_gather_scatter_fn_p): Likewise.
12906 (internal_fn_mask_index): Likewise.
12907 (internal_gather_scatter_fn_supported_p): Likewise.
12908 * optabs-query.c (supports_at_least_one_mode_p): New function.
12909 (supports_vec_gather_load_p): Likewise.
12910 * optabs-query.h (supports_vec_gather_load_p): Declare.
12911 * tree-vectorizer.h (gather_scatter_info): Add ifn, element_type
12912 and memory_type field.
12913 (NUM_PATTERNS): Bump to 15.
12914 * tree-vect-data-refs.c: Include internal-fn.h.
12915 (vect_gather_scatter_fn_p): New function.
12916 (vect_describe_gather_scatter_call): Likewise.
12917 (vect_check_gather_scatter): Try using internal functions for
12918 gather loads. Recognize existing calls to a gather load function.
12919 (vect_analyze_data_refs): Consider using gather loads if
12920 supports_vec_gather_load_p.
12921 * tree-vect-patterns.c (vect_get_load_store_mask): New function.
12922 (vect_get_gather_scatter_offset_type): Likewise.
12923 (vect_convert_mask_for_vectype): Likewise.
12924 (vect_add_conversion_to_patterm): Likewise.
12925 (vect_try_gather_scatter_pattern): Likewise.
12926 (vect_recog_gather_scatter_pattern): New pattern recognizer.
12927 (vect_vect_recog_func_ptrs): Add it.
12928 * tree-vect-stmts.c (exist_non_indexing_operands_for_use_p): Use
12929 internal_fn_mask_index and internal_gather_scatter_fn_p.
12930 (check_load_store_masking): Take the gather_scatter_info as an
12931 argument and handle gather loads.
12932 (vect_get_gather_scatter_ops): New function.
12933 (vectorizable_call): Check internal_load_fn_p.
12934 (vectorizable_load): Likewise. Handle gather load internal
12936 (vectorizable_store): Update call to check_load_store_masking.
12937 * config/aarch64/aarch64.md (UNSPEC_LD1_GATHER): New unspec.
12938 * config/aarch64/iterators.md (SVE_S, SVE_D): New mode iterators.
12939 * config/aarch64/predicates.md (aarch64_gather_scale_operand_w)
12940 (aarch64_gather_scale_operand_d): New predicates.
12941 * config/aarch64/aarch64-sve.md (gather_load<mode>): New expander.
12942 (mask_gather_load<mode>): New insns.
12944 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12945 Alan Hayward <alan.hayward@arm.com>
12946 David Sherwood <david.sherwood@arm.com>
12948 * optabs.def (fold_left_plus_optab): New optab.
12949 * doc/md.texi (fold_left_plus_@var{m}): Document.
12950 * internal-fn.def (IFN_FOLD_LEFT_PLUS): New internal function.
12951 * internal-fn.c (fold_left_direct): Define.
12952 (expand_fold_left_optab_fn): Likewise.
12953 (direct_fold_left_optab_supported_p): Likewise.
12954 * fold-const-call.c (fold_const_fold_left): New function.
12955 (fold_const_call): Use it to fold CFN_FOLD_LEFT_PLUS.
12956 * tree-parloops.c (valid_reduction_p): New function.
12957 (gather_scalar_reductions): Use it.
12958 * tree-vectorizer.h (FOLD_LEFT_REDUCTION): New vect_reduction_type.
12959 (vect_finish_replace_stmt): Declare.
12960 * tree-vect-loop.c (fold_left_reduction_fn): New function.
12961 (needs_fold_left_reduction_p): New function, split out from...
12962 (vect_is_simple_reduction): ...here. Accept reductions that
12963 forbid reassociation, but give them type FOLD_LEFT_REDUCTION.
12964 (vect_force_simple_reduction): Also store the reduction type in
12965 the assignment's STMT_VINFO_REDUC_TYPE.
12966 (vect_model_reduction_cost): Handle FOLD_LEFT_REDUCTION.
12967 (merge_with_identity): New function.
12968 (vect_expand_fold_left): Likewise.
12969 (vectorize_fold_left_reduction): Likewise.
12970 (vectorizable_reduction): Handle FOLD_LEFT_REDUCTION. Leave the
12971 scalar phi in place for it. Check for target support and reject
12972 cases that would reassociate the operation. Defer the transform
12973 phase to vectorize_fold_left_reduction.
12974 * config/aarch64/aarch64.md (UNSPEC_FADDA): New unspec.
12975 * config/aarch64/aarch64-sve.md (fold_left_plus_<mode>): New expander.
12976 (*fold_left_plus_<mode>, *pred_fold_left_plus_<mode>): New insns.
12978 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12980 * tree-if-conv.c (predicate_mem_writes): Remove redundant
12981 call to ifc_temp_var.
12983 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
12984 Alan Hayward <alan.hayward@arm.com>
12985 David Sherwood <david.sherwood@arm.com>
12987 * target.def (legitimize_address_displacement): Take the original
12988 offset as a poly_int.
12989 * targhooks.h (default_legitimize_address_displacement): Update
12991 * targhooks.c (default_legitimize_address_displacement): Likewise.
12992 * doc/tm.texi: Regenerate.
12993 * lra-constraints.c (base_plus_disp_to_reg): Take the displacement
12994 as an argument, moving assert of ad->disp == ad->disp_term to...
12995 (process_address_1): ...here. Update calls to base_plus_disp_to_reg.
12996 Try calling targetm.legitimize_address_displacement before expanding
12997 the address rather than afterwards, and adjust for the new interface.
12998 * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
12999 Match the new hook interface. Handle SVE addresses.
13000 * config/sh/sh.c (sh_legitimize_address_displacement): Make the
13001 new hook interface.
13003 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13005 * Makefile.in (OBJS): Add early-remat.o.
13006 * target.def (select_early_remat_modes): New hook.
13007 * doc/tm.texi.in (TARGET_SELECT_EARLY_REMAT_MODES): New hook.
13008 * doc/tm.texi: Regenerate.
13009 * targhooks.h (default_select_early_remat_modes): Declare.
13010 * targhooks.c (default_select_early_remat_modes): New function.
13011 * timevar.def (TV_EARLY_REMAT): New timevar.
13012 * passes.def (pass_early_remat): New pass.
13013 * tree-pass.h (make_pass_early_remat): Declare.
13014 * early-remat.c: New file.
13015 * config/aarch64/aarch64.c (aarch64_select_early_remat_modes): New
13017 (TARGET_SELECT_EARLY_REMAT_MODES): Define.
13019 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13020 Alan Hayward <alan.hayward@arm.com>
13021 David Sherwood <david.sherwood@arm.com>
13023 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Replace
13024 vfm1 with a bound_epilog parameter.
13025 (vect_do_peeling): Update calls accordingly, and move the prologue
13026 call earlier in the function. Treat the base bound_epilog as 0 for
13027 fully-masked loops and retain vf - 1 for other loops. Add 1 to
13028 this base when peeling for gaps.
13029 * tree-vect-loop.c (vect_analyze_loop_2): Allow peeling for gaps
13030 with fully-masked loops.
13031 (vect_estimate_min_profitable_iters): Handle the single peeled
13032 iteration in that case.
13034 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13035 Alan Hayward <alan.hayward@arm.com>
13036 David Sherwood <david.sherwood@arm.com>
13038 * tree-vect-data-refs.c (vect_analyze_group_access_1): Allow
13039 single-element interleaving even if the size is not a power of 2.
13040 * tree-vect-stmts.c (get_load_store_type): Disallow elementwise
13041 accesses for single-element interleaving if the group size is
13044 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13045 Alan Hayward <alan.hayward@arm.com>
13046 David Sherwood <david.sherwood@arm.com>
13048 * doc/md.texi (fold_extract_last_@var{m}): Document.
13049 * doc/sourcebuild.texi (vect_fold_extract_last): Likewise.
13050 * optabs.def (fold_extract_last_optab): New optab.
13051 * internal-fn.def (FOLD_EXTRACT_LAST): New internal function.
13052 * internal-fn.c (fold_extract_direct): New macro.
13053 (expand_fold_extract_optab_fn): Likewise.
13054 (direct_fold_extract_optab_supported_p): Likewise.
13055 * tree-vectorizer.h (EXTRACT_LAST_REDUCTION): New vect_reduction_type.
13056 * tree-vect-loop.c (vect_model_reduction_cost): Handle
13057 EXTRACT_LAST_REDUCTION.
13058 (get_initial_def_for_reduction): Do not create an initial vector
13059 for EXTRACT_LAST_REDUCTION reductions.
13060 (vectorizable_reduction): Leave the scalar phi in place for
13061 EXTRACT_LAST_REDUCTIONs. Try using EXTRACT_LAST_REDUCTION
13062 ahead of INTEGER_INDUC_COND_REDUCTION. Do not check for an
13063 epilogue code for EXTRACT_LAST_REDUCTION and defer the
13064 transform phase to vectorizable_condition.
13065 * tree-vect-stmts.c (vect_finish_stmt_generation_1): New function,
13067 (vect_finish_stmt_generation): ...here.
13068 (vect_finish_replace_stmt): New function.
13069 (vectorizable_condition): Handle EXTRACT_LAST_REDUCTION.
13070 * config/aarch64/aarch64-sve.md (fold_extract_last_<mode>): New
13072 * config/aarch64/aarch64.md (UNSPEC_CLASTB): New unspec.
13074 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13075 Alan Hayward <alan.hayward@arm.com>
13076 David Sherwood <david.sherwood@arm.com>
13078 * doc/md.texi (extract_last_@var{m}): Document.
13079 * optabs.def (extract_last_optab): New optab.
13080 * internal-fn.def (EXTRACT_LAST): New internal function.
13081 * internal-fn.c (cond_unary_direct): New macro.
13082 (expand_cond_unary_optab_fn): Likewise.
13083 (direct_cond_unary_optab_supported_p): Likewise.
13084 * tree-vect-loop.c (vectorizable_live_operation): Allow fully-masked
13085 loops using EXTRACT_LAST.
13086 * config/aarch64/aarch64-sve.md (aarch64_sve_lastb<mode>): Rename to...
13087 (extract_last_<mode>): ...this optab.
13088 (vec_extract<mode><Vel>): Update accordingly.
13090 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13091 Alan Hayward <alan.hayward@arm.com>
13092 David Sherwood <david.sherwood@arm.com>
13094 * target.def (empty_mask_is_expensive): New hook.
13095 * doc/tm.texi.in (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): New hook.
13096 * doc/tm.texi: Regenerate.
13097 * targhooks.h (default_empty_mask_is_expensive): Declare.
13098 * targhooks.c (default_empty_mask_is_expensive): New function.
13099 * tree-vectorizer.c (vectorize_loops): Only call optimize_mask_stores
13100 if the target says that empty masks are expensive.
13101 * config/aarch64/aarch64.c (aarch64_empty_mask_is_expensive):
13103 (TARGET_VECTORIZE_EMPTY_MASK_IS_EXPENSIVE): Redefine.
13105 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13106 Alan Hayward <alan.hayward@arm.com>
13107 David Sherwood <david.sherwood@arm.com>
13109 * tree-vectorizer.h (_loop_vec_info::mask_skip_niters): New field.
13110 (LOOP_VINFO_MASK_SKIP_NITERS): New macro.
13111 (vect_use_loop_mask_for_alignment_p): New function.
13112 (vect_prepare_for_masked_peels, vect_gen_while_not): Declare.
13113 * tree-vect-loop-manip.c (vect_set_loop_masks_directly): Add an
13114 niters_skip argument. Make sure that the first niters_skip elements
13115 of the first iteration are inactive.
13116 (vect_set_loop_condition_masked): Handle LOOP_VINFO_MASK_SKIP_NITERS.
13117 Update call to vect_set_loop_masks_directly.
13118 (get_misalign_in_elems): New function, split out from...
13119 (vect_gen_prolog_loop_niters): ...here.
13120 (vect_update_init_of_dr): Take a code argument that specifies whether
13121 the adjustment should be added or subtracted.
13122 (vect_update_init_of_drs): Likewise.
13123 (vect_prepare_for_masked_peels): New function.
13124 (vect_do_peeling): Skip prologue peeling if we're using a mask
13125 instead. Update call to vect_update_inits_of_drs.
13126 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13128 (vect_analyze_loop_2): Allow fully-masked loops with peeling for
13129 alignment. Do not include the number of peeled iterations in
13130 the minimum threshold in that case.
13131 (vectorizable_induction): Adjust the start value down by
13132 LOOP_VINFO_MASK_SKIP_NITERS iterations.
13133 (vect_transform_loop): Call vect_prepare_for_masked_peels.
13134 Take the number of skipped iterations into account when calculating
13136 * tree-vect-stmts.c (vect_gen_while_not): New function.
13138 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13139 Alan Hayward <alan.hayward@arm.com>
13140 David Sherwood <david.sherwood@arm.com>
13142 * doc/sourcebuild.texi (vect_fully_masked): Document.
13143 * params.def (PARAM_MIN_VECT_LOOP_BOUND): Change minimum and
13144 default value to 0.
13145 * tree-vect-loop.c (vect_analyze_loop_costing): New function,
13147 (vect_analyze_loop_2): ...here. Don't check the vectorization
13148 factor against the number of loop iterations if the loop is
13151 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13152 Alan Hayward <alan.hayward@arm.com>
13153 David Sherwood <david.sherwood@arm.com>
13155 * tree-ssa-loop-ivopts.c (USE_ADDRESS): Split into...
13156 (USE_REF_ADDRESS, USE_PTR_ADDRESS): ...these new use types.
13157 (dump_groups): Update accordingly.
13158 (iv_use::mem_type): New member variable.
13159 (address_p): New function.
13160 (record_use): Add a mem_type argument and initialize the new
13162 (record_group_use): Add a mem_type argument. Use address_p.
13163 Remove obsolete null checks of base_object. Update call to record_use.
13164 (find_interesting_uses_op): Update call to record_group_use.
13165 (find_interesting_uses_cond): Likewise.
13166 (find_interesting_uses_address): Likewise.
13167 (get_mem_type_for_internal_fn): New function.
13168 (find_address_like_use): Likewise.
13169 (find_interesting_uses_stmt): Try find_address_like_use before
13170 calling find_interesting_uses_op.
13171 (addr_offset_valid_p): Use the iv mem_type field as the type
13172 of the addressed memory.
13173 (add_autoinc_candidates): Likewise.
13174 (get_address_cost): Likewise.
13175 (split_small_address_groups_p): Use address_p.
13176 (split_address_groups): Likewise.
13177 (add_iv_candidate_for_use): Likewise.
13178 (autoinc_possible_for_pair): Likewise.
13179 (rewrite_groups): Likewise.
13180 (get_use_type): Check for USE_REF_ADDRESS instead of USE_ADDRESS.
13181 (determine_group_iv_cost): Update after split of USE_ADDRESS.
13182 (get_alias_ptr_type_for_ptr_address): New function.
13183 (rewrite_use_address): Rewrite address uses in calls that were
13184 identified by find_address_like_use.
13186 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13187 Alan Hayward <alan.hayward@arm.com>
13188 David Sherwood <david.sherwood@arm.com>
13190 * expr.c (expand_expr_addr_expr_1): Handle ADDR_EXPRs of
13192 * gimple-expr.h (is_gimple_addressable: Likewise.
13193 * gimple-expr.c (is_gimple_address): Likewise.
13194 * internal-fn.c (expand_call_mem_ref): New function.
13195 (expand_mask_load_optab_fn): Use it.
13196 (expand_mask_store_optab_fn): Likewise.
13198 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13199 Alan Hayward <alan.hayward@arm.com>
13200 David Sherwood <david.sherwood@arm.com>
13202 * doc/md.texi (cond_add@var{mode}, cond_sub@var{mode})
13203 (cond_and@var{mode}, cond_ior@var{mode}, cond_xor@var{mode})
13204 (cond_smin@var{mode}, cond_smax@var{mode}, cond_umin@var{mode})
13205 (cond_umax@var{mode}): Document.
13206 * optabs.def (cond_add_optab, cond_sub_optab, cond_and_optab)
13207 (cond_ior_optab, cond_xor_optab, cond_smin_optab, cond_smax_optab)
13208 (cond_umin_optab, cond_umax_optab): New optabs.
13209 * internal-fn.def (COND_ADD, COND_SUB, COND_MIN, COND_MAX, COND_AND)
13210 (COND_IOR, COND_XOR): New internal functions.
13211 * internal-fn.h (get_conditional_internal_fn): Declare.
13212 * internal-fn.c (cond_binary_direct): New macro.
13213 (expand_cond_binary_optab_fn): Likewise.
13214 (direct_cond_binary_optab_supported_p): Likewise.
13215 (get_conditional_internal_fn): New function.
13216 * tree-vect-loop.c (vectorizable_reduction): Handle fully-masked loops.
13217 Cope with reduction statements that are vectorized as calls rather
13219 * config/aarch64/aarch64-sve.md (cond_<optab><mode>): New insns.
13220 * config/aarch64/iterators.md (UNSPEC_COND_ADD, UNSPEC_COND_SUB)
13221 (UNSPEC_COND_SMAX, UNSPEC_COND_UMAX, UNSPEC_COND_SMIN)
13222 (UNSPEC_COND_UMIN, UNSPEC_COND_AND, UNSPEC_COND_ORR)
13223 (UNSPEC_COND_EOR): New unspecs.
13224 (optab): Add mappings for them.
13225 (SVE_COND_INT_OP, SVE_COND_FP_OP): New int iterators.
13226 (sve_int_op, sve_fp_op): New int attributes.
13228 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13229 Alan Hayward <alan.hayward@arm.com>
13230 David Sherwood <david.sherwood@arm.com>
13232 * optabs.def (while_ult_optab): New optab.
13233 * doc/md.texi (while_ult@var{m}@var{n}): Document.
13234 * internal-fn.def (WHILE_ULT): New internal function.
13235 * internal-fn.h (direct_internal_fn_supported_p): New override
13236 that takes two types as argument.
13237 * internal-fn.c (while_direct): New macro.
13238 (expand_while_optab_fn): New function.
13239 (convert_optab_supported_p): Likewise.
13240 (direct_while_optab_supported_p): New macro.
13241 * wide-int.h (wi::udiv_ceil): New function.
13242 * tree-vectorizer.h (rgroup_masks): New structure.
13243 (vec_loop_masks): New typedef.
13244 (_loop_vec_info): Add masks, mask_compare_type, can_fully_mask_p
13245 and fully_masked_p.
13246 (LOOP_VINFO_CAN_FULLY_MASK_P, LOOP_VINFO_FULLY_MASKED_P)
13247 (LOOP_VINFO_MASKS, LOOP_VINFO_MASK_COMPARE_TYPE): New macros.
13248 (vect_max_vf): New function.
13249 (slpeel_make_loop_iterate_ntimes): Delete.
13250 (vect_set_loop_condition, vect_get_loop_mask_type, vect_gen_while)
13251 (vect_halve_mask_nunits, vect_double_mask_nunits): Declare.
13252 (vect_record_loop_mask, vect_get_loop_mask): Likewise.
13253 * tree-vect-loop-manip.c: Include tree-ssa-loop-niter.h,
13254 internal-fn.h, stor-layout.h and optabs-query.h.
13255 (vect_set_loop_mask): New function.
13256 (add_preheader_seq): Likewise.
13257 (add_header_seq): Likewise.
13258 (interleave_supported_p): Likewise.
13259 (vect_maybe_permute_loop_masks): Likewise.
13260 (vect_set_loop_masks_directly): Likewise.
13261 (vect_set_loop_condition_masked): Likewise.
13262 (vect_set_loop_condition_unmasked): New function, split out from
13263 slpeel_make_loop_iterate_ntimes.
13264 (slpeel_make_loop_iterate_ntimes): Rename to..
13265 (vect_set_loop_condition): ...this. Use vect_set_loop_condition_masked
13266 for fully-masked loops and vect_set_loop_condition_unmasked otherwise.
13267 (vect_do_peeling): Update call accordingly.
13268 (vect_gen_vector_loop_niters): Use VF as the step for fully-masked
13270 * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
13271 mask_compare_type, can_fully_mask_p and fully_masked_p.
13272 (release_vec_loop_masks): New function.
13273 (_loop_vec_info): Use it to free the loop masks.
13274 (can_produce_all_loop_masks_p): New function.
13275 (vect_get_max_nscalars_per_iter): Likewise.
13276 (vect_verify_full_masking): Likewise.
13277 (vect_analyze_loop_2): Save LOOP_VINFO_CAN_FULLY_MASK_P around
13278 retries, and free the mask rgroups before retrying. Check loop-wide
13279 reasons for disallowing fully-masked loops. Make the final decision
13280 about whether use a fully-masked loop or not.
13281 (vect_estimate_min_profitable_iters): Do not assume that peeling
13282 for the number of iterations will be needed for fully-masked loops.
13283 (vectorizable_reduction): Disable fully-masked loops.
13284 (vectorizable_live_operation): Likewise.
13285 (vect_halve_mask_nunits): New function.
13286 (vect_double_mask_nunits): Likewise.
13287 (vect_record_loop_mask): Likewise.
13288 (vect_get_loop_mask): Likewise.
13289 (vect_transform_loop): Handle the case in which the final loop
13290 iteration might handle a partial vector. Call vect_set_loop_condition
13291 instead of slpeel_make_loop_iterate_ntimes.
13292 * tree-vect-stmts.c: Include tree-ssa-loop-niter.h and gimple-fold.h.
13293 (check_load_store_masking): New function.
13294 (prepare_load_store_mask): Likewise.
13295 (vectorizable_store): Handle fully-masked loops.
13296 (vectorizable_load): Likewise.
13297 (supportable_widening_operation): Use vect_halve_mask_nunits for
13299 (supportable_narrowing_operation): Likewise vect_double_mask_nunits.
13300 (vect_gen_while): New function.
13301 * config/aarch64/aarch64.md (umax<mode>3): New expander.
13302 (aarch64_uqdec<mode>): New insn.
13304 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13305 Alan Hayward <alan.hayward@arm.com>
13306 David Sherwood <david.sherwood@arm.com>
13308 * optabs.def (reduc_and_scal_optab, reduc_ior_scal_optab)
13309 (reduc_xor_scal_optab): New optabs.
13310 * doc/md.texi (reduc_and_scal_@var{m}, reduc_ior_scal_@var{m})
13311 (reduc_xor_scal_@var{m}): Document.
13312 * doc/sourcebuild.texi (vect_logical_reduc): Likewise.
13313 * internal-fn.def (IFN_REDUC_AND, IFN_REDUC_IOR, IFN_REDUC_XOR): New
13314 internal functions.
13315 * fold-const-call.c (fold_const_call): Handle them.
13316 * tree-vect-loop.c (reduction_fn_for_scalar_code): Return the new
13317 internal functions for BIT_AND_EXPR, BIT_IOR_EXPR and BIT_XOR_EXPR.
13318 * config/aarch64/aarch64-sve.md (reduc_<bit_reduc>_scal_<mode>):
13319 (*reduc_<bit_reduc>_scal_<mode>): New patterns.
13320 * config/aarch64/iterators.md (UNSPEC_ANDV, UNSPEC_ORV)
13321 (UNSPEC_XORV): New unspecs.
13322 (optab): Add entries for them.
13323 (BITWISEV): New int iterator.
13324 (bit_reduc_op): New int attributes.
13326 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13327 Alan Hayward <alan.hayward@arm.com>
13328 David Sherwood <david.sherwood@arm.com>
13330 * doc/md.texi (vec_shl_insert_@var{m}): New optab.
13331 * internal-fn.def (VEC_SHL_INSERT): New internal function.
13332 * optabs.def (vec_shl_insert_optab): New optab.
13333 * tree-vectorizer.h (can_duplicate_and_interleave_p): Declare.
13334 (duplicate_and_interleave): Likewise.
13335 * tree-vect-loop.c: Include internal-fn.h.
13336 (neutral_op_for_slp_reduction): New function, split out from
13337 get_initial_defs_for_reduction.
13338 (get_initial_def_for_reduction): Handle option 2 for variable-length
13339 vectors by loading the neutral value into a vector and then shifting
13340 the initial value into element 0.
13341 (get_initial_defs_for_reduction): Replace the code argument with
13342 the neutral value calculated by neutral_op_for_slp_reduction.
13343 Use gimple_build_vector for constant-length vectors.
13344 Use IFN_VEC_SHL_INSERT for variable-length vectors if all
13345 but the first group_size elements have a neutral value.
13346 Use duplicate_and_interleave otherwise.
13347 (vect_create_epilog_for_reduction): Take a neutral_op parameter.
13348 Update call to get_initial_defs_for_reduction. Handle SLP
13349 reductions for variable-length vectors by creating one vector
13350 result for each scalar result, with the elements associated
13351 with other scalar results stubbed out with the neutral value.
13352 (vectorizable_reduction): Call neutral_op_for_slp_reduction.
13353 Require IFN_VEC_SHL_INSERT for double reductions on
13354 variable-length vectors, or SLP reductions that have
13355 a neutral value. Require can_duplicate_and_interleave_p
13356 support for variable-length unchained SLP reductions if there
13357 is no neutral value, such as for MIN/MAX reductions. Also require
13358 the number of vector elements to be a multiple of the number of
13359 SLP statements when doing variable-length unchained SLP reductions.
13360 Update call to vect_create_epilog_for_reduction.
13361 * tree-vect-slp.c (can_duplicate_and_interleave_p): Make public
13362 and remove initial values.
13363 (duplicate_and_interleave): Make public.
13364 * config/aarch64/aarch64.md (UNSPEC_INSR): New unspec.
13365 * config/aarch64/aarch64-sve.md (vec_shl_insert_<mode>): New insn.
13367 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13368 Alan Hayward <alan.hayward@arm.com>
13369 David Sherwood <david.sherwood@arm.com>
13371 * tree-vect-slp.c: Include gimple-fold.h and internal-fn.h
13372 (can_duplicate_and_interleave_p): New function.
13373 (vect_get_and_check_slp_defs): Take the vector of statements
13374 rather than just the current one. Remove excess parentheses.
13375 Restriction rejectinon of vect_constant_def and vect_external_def
13376 for variable-length vectors to boolean types, or types for which
13377 can_duplicate_and_interleave_p is false.
13378 (vect_build_slp_tree_2): Update call to vect_get_and_check_slp_defs.
13379 (duplicate_and_interleave): New function.
13380 (vect_get_constant_vectors): Use gimple_build_vector for
13381 constant-length vectors and suitable variable-length constant
13382 vectors. Use duplicate_and_interleave for other variable-length
13383 vectors. Don't defer the update when inserting new statements.
13385 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13386 Alan Hayward <alan.hayward@arm.com>
13387 David Sherwood <david.sherwood@arm.com>
13389 * tree-vect-loop.c (vect_estimate_min_profitable_iters): Make sure
13390 min_profitable_iters doesn't go negative.
13392 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13393 Alan Hayward <alan.hayward@arm.com>
13394 David Sherwood <david.sherwood@arm.com>
13396 * doc/md.texi (vec_mask_load_lanes@var{m}@var{n}): Document.
13397 (vec_mask_store_lanes@var{m}@var{n}): Likewise.
13398 * optabs.def (vec_mask_load_lanes_optab): New optab.
13399 (vec_mask_store_lanes_optab): Likewise.
13400 * internal-fn.def (MASK_LOAD_LANES): New internal function.
13401 (MASK_STORE_LANES): Likewise.
13402 * internal-fn.c (mask_load_lanes_direct): New macro.
13403 (mask_store_lanes_direct): Likewise.
13404 (expand_mask_load_optab_fn): Handle masked operations.
13405 (expand_mask_load_lanes_optab_fn): New macro.
13406 (expand_mask_store_optab_fn): Handle masked operations.
13407 (expand_mask_store_lanes_optab_fn): New macro.
13408 (direct_mask_load_lanes_optab_supported_p): Likewise.
13409 (direct_mask_store_lanes_optab_supported_p): Likewise.
13410 * tree-vectorizer.h (vect_store_lanes_supported): Take a masked_p
13412 (vect_load_lanes_supported): Likewise.
13413 * tree-vect-data-refs.c (strip_conversion): New function.
13414 (can_group_stmts_p): Likewise.
13415 (vect_analyze_data_ref_accesses): Use it instead of checking
13416 for a pair of assignments.
13417 (vect_store_lanes_supported): Take a masked_p parameter.
13418 (vect_load_lanes_supported): Likewise.
13419 * tree-vect-loop.c (vect_analyze_loop_2): Update calls to
13420 vect_store_lanes_supported and vect_load_lanes_supported.
13421 * tree-vect-slp.c (vect_analyze_slp_instance): Likewise.
13422 * tree-vect-stmts.c (get_group_load_store_type): Take a masked_p
13423 parameter. Don't allow gaps for masked accesses.
13424 Use vect_get_store_rhs. Update calls to vect_store_lanes_supported
13425 and vect_load_lanes_supported.
13426 (get_load_store_type): Take a masked_p parameter and update
13427 call to get_group_load_store_type.
13428 (vectorizable_store): Update call to get_load_store_type.
13429 Handle IFN_MASK_STORE_LANES.
13430 (vectorizable_load): Update call to get_load_store_type.
13431 Handle IFN_MASK_LOAD_LANES.
13433 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13434 Alan Hayward <alan.hayward@arm.com>
13435 David Sherwood <david.sherwood@arm.com>
13437 * config/aarch64/aarch64-modes.def: Define x2, x3 and x4 vector
13439 * config/aarch64/aarch64-protos.h
13440 (aarch64_sve_struct_memory_operand_p): Declare.
13441 * config/aarch64/iterators.md (SVE_STRUCT): New mode iterator.
13442 (vector_count, insn_length, VSINGLE, vsingle): New mode attributes.
13443 (VPRED, vpred): Handle SVE structure modes.
13444 * config/aarch64/constraints.md (Utx): New constraint.
13445 * config/aarch64/predicates.md (aarch64_sve_struct_memory_operand)
13446 (aarch64_sve_struct_nonimmediate_operand): New predicates.
13447 * config/aarch64/aarch64.md (UNSPEC_LDN, UNSPEC_STN): New unspecs.
13448 * config/aarch64/aarch64-sve.md (mov<mode>, *aarch64_sve_mov<mode>_le)
13449 (*aarch64_sve_mov<mode>_be, pred_mov<mode>): New patterns for
13450 structure modes. Split into pieces after RA.
13451 (vec_load_lanes<mode><vsingle>, vec_mask_load_lanes<mode><vsingle>)
13452 (vec_store_lanes<mode><vsingle>, vec_mask_store_lanes<mode><vsingle>):
13454 * config/aarch64/aarch64.c (aarch64_classify_vector_mode): Handle
13455 SVE structure modes.
13456 (aarch64_classify_address): Likewise.
13457 (sizetochar): Move earlier in file.
13458 (aarch64_print_operand): Handle SVE register lists.
13459 (aarch64_array_mode): New function.
13460 (aarch64_sve_struct_memory_operand_p): Likewise.
13461 (TARGET_ARRAY_MODE): Redefine.
13463 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13464 Alan Hayward <alan.hayward@arm.com>
13465 David Sherwood <david.sherwood@arm.com>
13467 * target.def (array_mode): New target hook.
13468 * doc/tm.texi.in (TARGET_ARRAY_MODE): New hook.
13469 * doc/tm.texi: Regenerate.
13470 * hooks.h (hook_optmode_mode_uhwi_none): Declare.
13471 * hooks.c (hook_optmode_mode_uhwi_none): New function.
13472 * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Use
13473 targetm.array_mode.
13474 * stor-layout.c (mode_for_array): Likewise. Support polynomial
13477 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13478 Alan Hayward <alan.hayward@arm.com>
13479 David Sherwood <david.sherwood@arm.com>
13481 * fold-const.c (fold_binary_loc): Check the argument types
13482 rather than the result type when testing for a vector operation.
13484 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13486 * doc/tm.texi.in (DWARF_LAZY_REGISTER_VALUE): Document.
13487 * doc/tm.texi: Regenerate.
13489 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13490 Alan Hayward <alan.hayward@arm.com>
13491 David Sherwood <david.sherwood@arm.com>
13493 * doc/invoke.texi (-msve-vector-bits=): Document new option.
13494 (sve): Document new AArch64 extension.
13495 * doc/md.texi (w): Extend the description of the AArch64
13496 constraint to include SVE vectors.
13497 (Upl, Upa): Document new AArch64 predicate constraints.
13498 * config/aarch64/aarch64-opts.h (aarch64_sve_vector_bits_enum): New
13500 * config/aarch64/aarch64.opt (sve_vector_bits): New enum.
13501 (msve-vector-bits=): New option.
13502 * config/aarch64/aarch64-option-extensions.def (fp, simd): Disable
13503 SVE when these are disabled.
13504 (sve): New extension.
13505 * config/aarch64/aarch64-modes.def: Define SVE vector and predicate
13506 modes. Adjust their number of units based on aarch64_sve_vg.
13507 (MAX_BITSIZE_MODE_ANY_MODE): Define.
13508 * config/aarch64/aarch64-protos.h (ADDR_QUERY_ANY): New
13509 aarch64_addr_query_type.
13510 (aarch64_const_vec_all_same_in_range_p, aarch64_sve_pred_mode)
13511 (aarch64_sve_cnt_immediate_p, aarch64_sve_addvl_addpl_immediate_p)
13512 (aarch64_sve_inc_dec_immediate_p, aarch64_add_offset_temporaries)
13513 (aarch64_split_add_offset, aarch64_output_sve_cnt_immediate)
13514 (aarch64_output_sve_addvl_addpl, aarch64_output_sve_inc_dec_immediate)
13515 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): Declare.
13516 (aarch64_simd_imm_zero_p): Delete.
13517 (aarch64_check_zero_based_sve_index_immediate): Declare.
13518 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
13519 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
13520 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
13521 (aarch64_sve_float_mul_immediate_p): Likewise.
13522 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
13523 rather than an rtx.
13524 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): Declare.
13525 (aarch64_expand_mov_immediate): Take a gen_vec_duplicate callback.
13526 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move): Declare.
13527 (aarch64_expand_sve_vec_cmp_int, aarch64_expand_sve_vec_cmp_float)
13528 (aarch64_expand_sve_vcond, aarch64_expand_sve_vec_perm): Declare.
13529 (aarch64_regmode_natural_size): Likewise.
13530 * config/aarch64/aarch64.h (AARCH64_FL_SVE): New macro.
13531 (AARCH64_FL_V8_3, AARCH64_FL_RCPC, AARCH64_FL_DOTPROD): Shift
13533 (AARCH64_ISA_SVE, TARGET_SVE): New macros.
13534 (FIXED_REGISTERS, CALL_USED_REGISTERS, REGISTER_NAMES): Add entries
13535 for VG and the SVE predicate registers.
13536 (V_ALIASES): Add a "z"-prefixed alias.
13537 (FIRST_PSEUDO_REGISTER): Change to P15_REGNUM + 1.
13538 (AARCH64_DWARF_VG, AARCH64_DWARF_P0): New macros.
13539 (PR_REGNUM_P, PR_LO_REGNUM_P): Likewise.
13540 (PR_LO_REGS, PR_HI_REGS, PR_REGS): New reg_classes.
13541 (REG_CLASS_NAMES): Add entries for them.
13542 (REG_CLASS_CONTENTS): Likewise. Update ALL_REGS to include VG
13543 and the predicate registers.
13544 (aarch64_sve_vg): Declare.
13545 (BITS_PER_SVE_VECTOR, BYTES_PER_SVE_VECTOR, BYTES_PER_SVE_PRED)
13546 (SVE_BYTE_MODE, MAX_COMPILE_TIME_VEC_BYTES): New macros.
13547 (REGMODE_NATURAL_SIZE): Define.
13548 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins): Handle
13550 * config/aarch64/aarch64.c: Include cfgrtl.h.
13551 (simd_immediate_info): Add a constructor for series vectors,
13552 and an associated step field.
13553 (aarch64_sve_vg): New variable.
13554 (aarch64_dbx_register_number): Handle VG and the predicate registers.
13555 (aarch64_vect_struct_mode_p, aarch64_vector_mode_p): Delete.
13556 (VEC_ADVSIMD, VEC_SVE_DATA, VEC_SVE_PRED, VEC_STRUCT, VEC_ANY_SVE)
13557 (VEC_ANY_DATA, VEC_STRUCT): New constants.
13558 (aarch64_advsimd_struct_mode_p, aarch64_sve_pred_mode_p)
13559 (aarch64_classify_vector_mode, aarch64_vector_data_mode_p)
13560 (aarch64_sve_data_mode_p, aarch64_sve_pred_mode)
13561 (aarch64_get_mask_mode): New functions.
13562 (aarch64_hard_regno_nregs): Handle SVE data modes for FP_REGS
13563 and FP_LO_REGS. Handle PR_REGS, PR_LO_REGS and PR_HI_REGS.
13564 (aarch64_hard_regno_mode_ok): Handle VG. Also handle the SVE
13565 predicate modes and predicate registers. Explicitly restrict
13566 GPRs to modes of 16 bytes or smaller. Only allow FP registers
13567 to store a vector mode if it is recognized by
13568 aarch64_classify_vector_mode.
13569 (aarch64_regmode_natural_size): New function.
13570 (aarch64_hard_regno_caller_save_mode): Return the original mode
13572 (aarch64_sve_cnt_immediate_p, aarch64_output_sve_cnt_immediate)
13573 (aarch64_sve_addvl_addpl_immediate_p, aarch64_output_sve_addvl_addpl)
13574 (aarch64_sve_inc_dec_immediate_p, aarch64_output_sve_inc_dec_immediate)
13575 (aarch64_add_offset_1_temporaries, aarch64_offset_temporaries): New
13577 (aarch64_add_offset): Add a temp2 parameter. Assert that temp1
13578 does not overlap dest if the function is frame-related. Handle
13580 (aarch64_split_add_offset): New function.
13581 (aarch64_add_sp, aarch64_sub_sp): Add temp2 parameters and pass
13582 them aarch64_add_offset.
13583 (aarch64_allocate_and_probe_stack_space): Add a temp2 parameter
13584 and update call to aarch64_sub_sp.
13585 (aarch64_add_cfa_expression): New function.
13586 (aarch64_expand_prologue): Pass extra temporary registers to the
13587 functions above. Handle the case in which we need to emit new
13588 DW_CFA_expressions for registers that were originally saved
13589 relative to the stack pointer, but now have to be expressed
13590 relative to the frame pointer.
13591 (aarch64_output_mi_thunk): Pass extra temporary registers to the
13593 (aarch64_expand_epilogue): Likewise. Prevent inheritance of
13594 IP0 and IP1 values for SVE frames.
13595 (aarch64_expand_vec_series): New function.
13596 (aarch64_expand_sve_widened_duplicate): Likewise.
13597 (aarch64_expand_sve_const_vector): Likewise.
13598 (aarch64_expand_mov_immediate): Add a gen_vec_duplicate parameter.
13599 Handle SVE constants. Use emit_move_insn to move a force_const_mem
13600 into the register, rather than emitting a SET directly.
13601 (aarch64_emit_sve_pred_move, aarch64_expand_sve_mem_move)
13602 (aarch64_get_reg_raw_mode, offset_4bit_signed_scaled_p)
13603 (offset_6bit_unsigned_scaled_p, aarch64_offset_7bit_signed_scaled_p)
13604 (offset_9bit_signed_scaled_p): New functions.
13605 (aarch64_replicate_bitmask_imm): New function.
13606 (aarch64_bitmask_imm): Use it.
13607 (aarch64_cannot_force_const_mem): Reject expressions involving
13608 a CONST_POLY_INT. Update call to aarch64_classify_symbol.
13609 (aarch64_classify_index): Handle SVE indices, by requiring
13610 a plain register index with a scale that matches the element size.
13611 (aarch64_classify_address): Handle SVE addresses. Assert that
13612 the mode of the address is VOIDmode or an integer mode.
13613 Update call to aarch64_classify_symbol.
13614 (aarch64_classify_symbolic_expression): Update call to
13615 aarch64_classify_symbol.
13616 (aarch64_const_vec_all_in_range_p): New function.
13617 (aarch64_print_vector_float_operand): Likewise.
13618 (aarch64_print_operand): Handle 'N' and 'C'. Use "zN" rather than
13619 "vN" for FP registers with SVE modes. Handle (const ...) vectors
13620 and the FP immediates 1.0 and 0.5.
13621 (aarch64_print_address_internal): Handle SVE addresses.
13622 (aarch64_print_operand_address): Use ADDR_QUERY_ANY.
13623 (aarch64_regno_regclass): Handle predicate registers.
13624 (aarch64_secondary_reload): Handle big-endian reloads of SVE
13626 (aarch64_class_max_nregs): Handle SVE modes and predicate registers.
13627 (aarch64_rtx_costs): Check for ADDVL and ADDPL instructions.
13628 (aarch64_convert_sve_vector_bits): New function.
13629 (aarch64_override_options): Use it to handle -msve-vector-bits=.
13630 (aarch64_classify_symbol): Take the offset as a HOST_WIDE_INT
13631 rather than an rtx.
13632 (aarch64_legitimate_constant_p): Use aarch64_classify_vector_mode.
13633 Handle SVE vector and predicate modes. Accept VL-based constants
13634 that need only one temporary register, and VL offsets that require
13635 no temporary registers.
13636 (aarch64_conditional_register_usage): Mark the predicate registers
13637 as fixed if SVE isn't available.
13638 (aarch64_vector_mode_supported_p): Use aarch64_classify_vector_mode.
13639 Return true for SVE vector and predicate modes.
13640 (aarch64_simd_container_mode): Take the number of bits as a poly_int64
13641 rather than an unsigned int. Handle SVE modes.
13642 (aarch64_preferred_simd_mode): Update call accordingly. Handle
13644 (aarch64_autovectorize_vector_sizes): Add BYTES_PER_SVE_VECTOR
13646 (aarch64_sve_index_immediate_p, aarch64_sve_arith_immediate_p)
13647 (aarch64_sve_bitmask_immediate_p, aarch64_sve_dup_immediate_p)
13648 (aarch64_sve_cmp_immediate_p, aarch64_sve_float_arith_immediate_p)
13649 (aarch64_sve_float_mul_immediate_p): New functions.
13650 (aarch64_sve_valid_immediate): New function.
13651 (aarch64_simd_valid_immediate): Use it as the fallback for SVE vectors.
13652 Explicitly reject structure modes. Check for INDEX constants.
13653 Handle PTRUE and PFALSE constants.
13654 (aarch64_check_zero_based_sve_index_immediate): New function.
13655 (aarch64_simd_imm_zero_p): Delete.
13656 (aarch64_mov_operand_p): Use aarch64_simd_valid_immediate for
13657 vector modes. Accept constants in the range of CNT[BHWD].
13658 (aarch64_simd_scalar_immediate_valid_for_move): Explicitly
13659 ask for an Advanced SIMD mode.
13660 (aarch64_sve_ld1r_operand_p, aarch64_sve_ldr_operand_p): New functions.
13661 (aarch64_simd_vector_alignment): Handle SVE predicates.
13662 (aarch64_vectorize_preferred_vector_alignment): New function.
13663 (aarch64_simd_vector_alignment_reachable): Use it instead of
13665 (aarch64_shift_truncation_mask): Use aarch64_vector_data_mode_p.
13666 (aarch64_output_sve_mov_immediate, aarch64_output_ptrue): New
13668 (MAX_VECT_LEN): Delete.
13669 (expand_vec_perm_d): Add a vec_flags field.
13670 (emit_unspec2, aarch64_expand_sve_vec_perm): New functions.
13671 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
13672 (aarch64_evpc_ext): Don't apply a big-endian lane correction
13674 (aarch64_evpc_rev): Rename to...
13675 (aarch64_evpc_rev_local): ...this. Use a predicated operation for SVE.
13676 (aarch64_evpc_rev_global): New function.
13677 (aarch64_evpc_dup): Enforce a 64-byte range for SVE DUP.
13678 (aarch64_evpc_tbl): Use MAX_COMPILE_TIME_VEC_BYTES instead of
13680 (aarch64_evpc_sve_tbl): New function.
13681 (aarch64_expand_vec_perm_const_1): Update after rename of
13682 aarch64_evpc_rev. Handle SVE permutes too, trying
13683 aarch64_evpc_rev_global and using aarch64_evpc_sve_tbl rather
13684 than aarch64_evpc_tbl.
13685 (aarch64_vectorize_vec_perm_const): Initialize vec_flags.
13686 (aarch64_sve_cmp_operand_p, aarch64_unspec_cond_code)
13687 (aarch64_gen_unspec_cond, aarch64_expand_sve_vec_cmp_int)
13688 (aarch64_emit_unspec_cond, aarch64_emit_unspec_cond_or)
13689 (aarch64_emit_inverted_unspec_cond, aarch64_expand_sve_vec_cmp_float)
13690 (aarch64_expand_sve_vcond): New functions.
13691 (aarch64_modes_tieable_p): Use aarch64_vector_data_mode_p instead
13692 of aarch64_vector_mode_p.
13693 (aarch64_dwarf_poly_indeterminate_value): New function.
13694 (aarch64_compute_pressure_classes): Likewise.
13695 (aarch64_can_change_mode_class): Likewise.
13696 (TARGET_GET_RAW_RESULT_MODE, TARGET_GET_RAW_ARG_MODE): Redefine.
13697 (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): Likewise.
13698 (TARGET_VECTORIZE_GET_MASK_MODE): Likewise.
13699 (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Likewise.
13700 (TARGET_COMPUTE_PRESSURE_CLASSES): Likewise.
13701 (TARGET_CAN_CHANGE_MODE_CLASS): Likewise.
13702 * config/aarch64/constraints.md (Upa, Upl, Uav, Uat, Usv, Usi, Utr)
13703 (Uty, Dm, vsa, vsc, vsd, vsi, vsn, vsl, vsm, vsA, vsM, vsN): New
13705 (Dn, Dl, Dr): Accept const as well as const_vector.
13706 (Dz): Likewise. Compare against CONST0_RTX.
13707 * config/aarch64/iterators.md: Refer to "Advanced SIMD" instead
13708 of "vector" where appropriate.
13709 (SVE_ALL, SVE_BH, SVE_BHS, SVE_BHSI, SVE_HSDI, SVE_HSF, SVE_SD)
13710 (SVE_SDI, SVE_I, SVE_F, PRED_ALL, PRED_BHS): New mode iterators.
13711 (UNSPEC_SEL, UNSPEC_ANDF, UNSPEC_IORF, UNSPEC_XORF, UNSPEC_COND_LT)
13712 (UNSPEC_COND_LE, UNSPEC_COND_EQ, UNSPEC_COND_NE, UNSPEC_COND_GE)
13713 (UNSPEC_COND_GT, UNSPEC_COND_LO, UNSPEC_COND_LS, UNSPEC_COND_HS)
13714 (UNSPEC_COND_HI, UNSPEC_COND_UO): New unspecs.
13715 (Vetype, VEL, Vel, VWIDE, Vwide, vw, vwcore, V_INT_EQUIV)
13716 (v_int_equiv): Extend to SVE modes.
13717 (Vesize, V128, v128, Vewtype, V_FP_EQUIV, v_fp_equiv, VPRED): New
13719 (LOGICAL_OR, SVE_INT_UNARY, SVE_FP_UNARY): New code iterators.
13720 (optab): Handle popcount, smin, smax, umin, umax, abs and sqrt.
13721 (logical_nn, lr, sve_int_op, sve_fp_op): New code attributs.
13722 (LOGICALF, OPTAB_PERMUTE, UNPACK, UNPACK_UNSIGNED, SVE_COND_INT_CMP)
13723 (SVE_COND_FP_CMP): New int iterators.
13724 (perm_hilo): Handle the new unpack unspecs.
13725 (optab, logicalf_op, su, perm_optab, cmp_op, imm_con): New int
13727 * config/aarch64/predicates.md (aarch64_sve_cnt_immediate)
13728 (aarch64_sve_addvl_addpl_immediate, aarch64_split_add_offset_immediate)
13729 (aarch64_pluslong_or_poly_operand, aarch64_nonmemory_operand)
13730 (aarch64_equality_operator, aarch64_constant_vector_operand)
13731 (aarch64_sve_ld1r_operand, aarch64_sve_ldr_operand): New predicates.
13732 (aarch64_sve_nonimmediate_operand): Likewise.
13733 (aarch64_sve_general_operand): Likewise.
13734 (aarch64_sve_dup_operand, aarch64_sve_arith_immediate): Likewise.
13735 (aarch64_sve_sub_arith_immediate, aarch64_sve_inc_dec_immediate)
13736 (aarch64_sve_logical_immediate, aarch64_sve_mul_immediate): Likewise.
13737 (aarch64_sve_dup_immediate, aarch64_sve_cmp_vsc_immediate): Likewise.
13738 (aarch64_sve_cmp_vsd_immediate, aarch64_sve_index_immediate): Likewise.
13739 (aarch64_sve_float_arith_immediate): Likewise.
13740 (aarch64_sve_float_arith_with_sub_immediate): Likewise.
13741 (aarch64_sve_float_mul_immediate, aarch64_sve_arith_operand): Likewise.
13742 (aarch64_sve_add_operand, aarch64_sve_logical_operand): Likewise.
13743 (aarch64_sve_lshift_operand, aarch64_sve_rshift_operand): Likewise.
13744 (aarch64_sve_mul_operand, aarch64_sve_cmp_vsc_operand): Likewise.
13745 (aarch64_sve_cmp_vsd_operand, aarch64_sve_index_operand): Likewise.
13746 (aarch64_sve_float_arith_operand): Likewise.
13747 (aarch64_sve_float_arith_with_sub_operand): Likewise.
13748 (aarch64_sve_float_mul_operand): Likewise.
13749 (aarch64_sve_vec_perm_operand): Likewise.
13750 (aarch64_pluslong_operand): Include aarch64_sve_addvl_addpl_immediate.
13751 (aarch64_mov_operand): Accept const_poly_int and const_vector.
13752 (aarch64_simd_lshift_imm, aarch64_simd_rshift_imm): Accept const
13753 as well as const_vector.
13754 (aarch64_simd_imm_zero, aarch64_simd_imm_minus_one): Move earlier
13755 in file. Use CONST0_RTX and CONSTM1_RTX.
13756 (aarch64_simd_or_scalar_imm_zero): Likewise. Add match_codes.
13757 (aarch64_simd_reg_or_zero): Accept const as well as const_vector.
13758 Use aarch64_simd_imm_zero.
13759 * config/aarch64/aarch64-sve.md: New file.
13760 * config/aarch64/aarch64.md: Include it.
13761 (VG_REGNUM, P0_REGNUM, P7_REGNUM, P15_REGNUM): New register numbers.
13762 (UNSPEC_REV, UNSPEC_LD1_SVE, UNSPEC_ST1_SVE, UNSPEC_MERGE_PTRUE)
13763 (UNSPEC_PTEST_PTRUE, UNSPEC_UNPACKSHI, UNSPEC_UNPACKUHI)
13764 (UNSPEC_UNPACKSLO, UNSPEC_UNPACKULO, UNSPEC_PACK)
13765 (UNSPEC_FLOAT_CONVERT, UNSPEC_WHILE_LO): New unspec constants.
13766 (sve): New attribute.
13767 (enabled): Disable instructions with the sve attribute unless
13769 (movqi, movhi): Pass CONST_POLY_INT operaneds through
13770 aarch64_expand_mov_immediate.
13771 (*mov<mode>_aarch64, *movsi_aarch64, *movdi_aarch64): Handle
13772 CNT[BHSD] immediates.
13773 (movti): Split CONST_POLY_INT moves into two halves.
13774 (add<mode>3): Accept aarch64_pluslong_or_poly_operand.
13775 Split additions that need a temporary here if the destination
13776 is the stack pointer.
13777 (*add<mode>3_aarch64): Handle ADDVL and ADDPL immediates.
13778 (*add<mode>3_poly_1): New instruction.
13779 (set_clobber_cc): New expander.
13781 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13783 * simplify-rtx.c (simplify_immed_subreg): Add an inner_bytes
13784 parameter and use it instead of GET_MODE_SIZE (innermode). Use
13785 inner_bytes * BITS_PER_UNIT instead of GET_MODE_BITSIZE (innermode).
13786 Use CEIL (inner_bytes, GET_MODE_UNIT_SIZE (innermode)) instead of
13787 GET_MODE_NUNITS (innermode). Also add a first_elem parameter.
13788 Change innermode from fixed_mode_size to machine_mode.
13789 (simplify_subreg): Update call accordingly. Handle a constant-sized
13790 subreg of a variable-length CONST_VECTOR.
13792 2018-01-13 Richard Sandiford <richard.sandiford@linaro.org>
13793 Alan Hayward <alan.hayward@arm.com>
13794 David Sherwood <david.sherwood@arm.com>
13796 * tree-ssa-address.c (mem_ref_valid_without_offset_p): New function.
13797 (add_offset_to_base): New function, split out from...
13798 (create_mem_ref): ...here. When handling a scale other than 1,
13799 check first whether the address is valid without the offset.
13800 Add it into the base if so, leaving the index and scale as-is.
13802 2018-01-12 Jakub Jelinek <jakub@redhat.com>
13805 * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Call
13806 fold_for_warn before checking if arg2 is INTEGER_CST.
13808 2018-01-12 Segher Boessenkool <segher@kernel.crashing.org>
13810 * config/rs6000/predicates.md (load_multiple_operation): Delete.
13811 (store_multiple_operation): Delete.
13812 * config/rs6000/rs6000-cpus.def (601): Remove MASK_STRING.
13813 * config/rs6000/rs6000-protos.h (rs6000_output_load_multiple): Delete.
13814 * config/rs6000/rs6000-string.c (expand_block_move): Delete everything
13815 guarded by TARGET_STRING.
13816 (rs6000_output_load_multiple): Delete.
13817 * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
13818 OPTION_MASK_STRING / TARGET_STRING handling.
13819 (print_operand) <'N', 'O'>: Add comment that these are unused now.
13820 (const rs6000_opt_masks) <"string">: Change mask to 0.
13821 * config/rs6000/rs6000.h (TARGET_DEFAULT): Remove MASK_STRING.
13822 (MASK_STRING): Delete.
13823 * config/rs6000/rs6000.md (*mov<mode>_string): Delete TARGET_STRING
13825 (load_multiple): Delete.
13832 (store_multiple): Delete.
13839 (movmemsi_8reg): Delete.
13840 (corresponding unnamed define_insn): Delete.
13841 (movmemsi_6reg): Delete.
13842 (corresponding unnamed define_insn): Delete.
13843 (movmemsi_4reg): Delete.
13844 (corresponding unnamed define_insn): Delete.
13845 (movmemsi_2reg): Delete.
13846 (corresponding unnamed define_insn): Delete.
13847 (movmemsi_1reg): Delete.
13848 (corresponding unnamed define_insn): Delete.
13849 * config/rs6000/rs6000.opt (mno-string): New.
13850 (mstring): Replace by deprecation warning stub.
13851 * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mstring.
13853 2018-01-12 Jakub Jelinek <jakub@redhat.com>
13855 * regrename.c (regrename_do_replace): If replacing the same
13856 reg multiple times, try to reuse last created gen_raw_REG.
13859 * bb-reorder.c (pass_partition_blocks::gate): In lto don't partition
13860 main to workaround a bug in GDB.
13862 2018-01-12 Tom de Vries <tom@codesourcery.com>
13865 * config.gcc (nvptx*-*-*): Set use_gcc_stdint=wrap.
13867 2018-01-12 Vladimir Makarov <vmakarov@redhat.com>
13869 PR rtl-optimization/80481
13870 * ira-color.c (get_cap_member): New function.
13871 (allocnos_conflict_by_live_ranges_p): Use it.
13872 (slot_coalesced_allocno_live_ranges_intersect_p): Add assert.
13873 (setup_slot_coalesced_allocno_live_ranges): Ditto.
13875 2018-01-12 Uros Bizjak <ubizjak@gmail.com>
13878 * config/alpha/alpha.md (*saddsi_1): New insn_ans_split pattern.
13879 (*saddl_se_1): Ditto.
13880 (*ssubsi_1): Ditto.
13881 (*ssubl_se_1): Ditto.
13883 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13885 * tree-predcom.c (aff_combination_dr_offset): Use wi::to_poly_widest
13886 rather than wi::to_widest for DR_INITs.
13887 * tree-vect-data-refs.c (vect_find_same_alignment_drs): Use
13888 wi::to_poly_offset rather than wi::to_offset for DR_INIT.
13889 (vect_analyze_data_ref_accesses): Require both DR_INITs to be
13891 (vect_analyze_group_access_1): Note that here.
13893 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13895 * tree-vectorizer.c (get_vec_alignment_for_array_type): Handle
13896 polynomial type sizes.
13898 2018-01-12 Richard Sandiford <richard.sandiford@linaro.org>
13900 * gimplify.c (gimple_add_tmp_var_fn): Allow variables to have a
13901 poly_uint64 size, rather than requiring an unsigned HOST_WIDE_INT size.
13902 (gimple_add_tmp_var): Likewise.
13904 2018-01-12 Martin Liska <mliska@suse.cz>
13906 * gimple.c (gimple_alloc_counts): Use uint64_t instead of int.
13907 (gimple_alloc_sizes): Likewise.
13908 (dump_gimple_statistics): Use PRIu64 in printf format.
13909 * gimple.h: Change uint64_t to int.
13911 2018-01-12 Martin Liska <mliska@suse.cz>
13913 * tree-core.h: Use uint64_t instead of int.
13914 * tree.c (tree_node_counts): Likewise.
13915 (tree_node_sizes): Likewise.
13916 (dump_tree_statistics): Use PRIu64 in printf format.
13918 2018-01-12 Martin Liska <mliska@suse.cz>
13920 * Makefile.in: As qsort_chk is implemented in vec.c, add
13921 vec.o to linkage of gencfn-macros.
13922 * tree.c (build_new_poly_int_cst): Add CXX_MEM_STAT_INFO as it's
13923 passing the info to record_node_allocation_statistics.
13924 (test_vector_cst_patterns): Add CXX_MEM_STAT_INFO to declaration
13926 * ggc-common.c (struct ggc_usage): Add operator== and use
13927 it in operator< and compare function.
13928 * mem-stats.h (struct mem_usage): Likewise.
13929 * vec.c (struct vec_usage): Remove operator< and compare
13930 function. Can be simply inherited.
13932 2018-01-12 Martin Jambor <mjambor@suse.cz>
13935 * params.def: New parameter PARAM_AVOID_FMA_MAX_BITS.
13936 * tree-ssa-math-opts.c: Include domwalk.h.
13937 (convert_mult_to_fma_1): New function.
13938 (fma_transformation_info): New type.
13939 (fma_deferring_state): Likewise.
13940 (cancel_fma_deferring): New function.
13941 (result_of_phi): Likewise.
13942 (last_fma_candidate_feeds_initial_phi): Likewise.
13943 (convert_mult_to_fma): Added deferring logic, split actual
13944 transformation to convert_mult_to_fma_1.
13945 (math_opts_dom_walker): New type.
13946 (math_opts_dom_walker::after_dom_children): New method, body moved
13947 here from pass_optimize_widening_mul::execute, added deferring logic
13949 (pass_optimize_widening_mul::execute): Moved most of code to
13950 math_opts_dom_walker::after_dom_children.
13951 * config/i386/x86-tune.def (X86_TUNE_AVOID_128FMA_CHAINS): New.
13952 * config/i386/i386.c (ix86_option_override_internal): Added
13953 maybe_setting of PARAM_AVOID_FMA_MAX_BITS.
13955 2018-01-12 Richard Biener <rguenther@suse.de>
13958 * dwarf2out.c (gen_variable_die): Do not reset old_die for
13959 inline instance vars.
13961 2018-01-12 Oleg Endo <olegendo@gcc.gnu.org>
13964 * config/rx/rx.c (rx_is_restricted_memory_address):
13965 Handle SUBREG case.
13967 2018-01-12 Richard Biener <rguenther@suse.de>
13969 PR tree-optimization/80846
13970 * target.def (split_reduction): New target hook.
13971 * targhooks.c (default_split_reduction): New function.
13972 * targhooks.h (default_split_reduction): Declare.
13973 * tree-vect-loop.c (vect_create_epilog_for_reduction): If the
13974 target requests first reduce vectors by combining low and high
13976 * tree-vect-stmts.c (vect_gen_perm_mask_any): Adjust.
13977 (get_vectype_for_scalar_type_and_size): Export.
13978 * tree-vectorizer.h (get_vectype_for_scalar_type_and_size): Declare.
13979 * doc/tm.texi.in (TARGET_VECTORIZE_SPLIT_REDUCTION): Document.
13980 * doc/tm.texi: Regenerate.
13981 * config/i386/i386.c (ix86_split_reduction): Implement
13982 TARGET_VECTORIZE_SPLIT_REDUCTION.
13984 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
13987 * config/sparc/sparc.h (PIC_OFFSET_TABLE_REGNUM): Set to INVALID_REGNUM
13988 in PIC mode except for TARGET_VXWORKS_RTP.
13989 * config/sparc/sparc.c: Include cfgrtl.h.
13990 (TARGET_INIT_PIC_REG): Define.
13991 (TARGET_USE_PSEUDO_PIC_REG): Likewise.
13992 (sparc_pic_register_p): New predicate.
13993 (sparc_legitimate_address_p): Use it.
13994 (sparc_legitimize_pic_address): Likewise.
13995 (sparc_delegitimize_address): Likewise.
13996 (sparc_mode_dependent_address_p): Likewise.
13997 (gen_load_pcrel_sym): Remove 4th parameter.
13998 (load_got_register): Adjust call to above. Remove obsolete stuff.
13999 (sparc_expand_prologue): Do not call load_got_register here.
14000 (sparc_flat_expand_prologue): Likewise.
14001 (sparc_output_mi_thunk): Set the pic_offset_table_rtx object.
14002 (sparc_use_pseudo_pic_reg): New function.
14003 (sparc_init_pic_reg): Likewise.
14004 * config/sparc/sparc.md (vxworks_load_got): Set the GOT register.
14005 (builtin_setjmp_receiver): Enable only for TARGET_VXWORKS_RTP.
14007 2018-01-12 Christophe Lyon <christophe.lyon@linaro.org>
14009 * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
14010 Add item for branch_cost.
14012 2018-01-12 Eric Botcazou <ebotcazou@adacore.com>
14014 PR rtl-optimization/83565
14015 * rtlanal.c (nonzero_bits1): On WORD_REGISTER_OPERATIONS machines, do
14016 not extend the result to a larger mode for rotate operations.
14017 (num_sign_bit_copies1): Likewise.
14019 2018-01-12 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
14022 * config/sol2.h (STARTFILE_ARCH_SPEC): Don't use with -shared or
14024 Use values-Xc.o for -pedantic.
14025 Link with values-xpg4.o for C90, values-xpg6.o otherwise.
14027 2018-01-12 Martin Liska <mliska@suse.cz>
14030 * ipa-devirt.c (final_warning_record::grow_type_warnings):
14032 (possible_polymorphic_call_targets): Use it.
14033 (ipa_devirt): Likewise.
14035 2018-01-12 Martin Liska <mliska@suse.cz>
14037 * profile-count.h (enum profile_quality): Use 0 as invalid
14038 enum value of profile_quality.
14040 2018-01-12 Chung-Ju Wu <jasonwucj@gmail.com>
14042 * doc/invoke.texi (NDS32 Options): Add -mext-perf, -mext-perf2 and
14043 -mext-string options.
14045 2018-01-12 Richard Biener <rguenther@suse.de>
14047 * lto-streamer-out.c (DFS::DFS_write_tree_body): Process
14048 DECL_DEBUG_EXPR conditional on DECL_HAS_DEBUG_EXPR_P.
14049 * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
14051 * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
14053 2018-01-11 Michael Meissner <meissner@linux.vnet.ibm.com>
14055 * configure.ac (--with-long-double-format): Add support for the
14056 configuration option to change the default long double format on
14058 * config.gcc (powerpc*-linux*-*): Likewise.
14059 * configure: Regenerate.
14060 * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If long
14061 double is IEEE, define __KC__ and __KF__ to allow floatn.h to be
14062 used without modification.
14064 2018-01-11 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14066 * config/rs6000/rs6000-builtin.def (BU_P7_MISC_X): New #define.
14067 (SPEC_BARRIER): New instantiation of BU_P7_MISC_X.
14068 * config/rs6000/rs6000.c (rs6000_expand_builtin): Handle
14069 MISC_BUILTIN_SPEC_BARRIER.
14070 (rs6000_init_builtins): Likewise.
14071 * config/rs6000/rs6000.md (UNSPECV_SPEC_BARRIER): New UNSPECV
14073 (speculation_barrier): New define_insn.
14074 * doc/extend.texi: Document __builtin_speculation_barrier.
14076 2018-01-11 Jakub Jelinek <jakub@redhat.com>
14079 * config/i386/i386.c (ix86_expand_vector_init_one_nonzero): If one_var
14080 is 0, for V{8,16}S[IF] and V[48]D[IF]mode use gen_vec_set<mode>_0.
14081 * config/i386/sse.md (VI8_AVX_AVX512F, VI4F_256_512): New mode
14083 (ssescalarmodesuffix): Add 512-bit vectors. Use "d" or "q" for
14084 integral modes instead of "ss" and "sd".
14085 (vec_set<mode>_0): New define_insns for 256-bit and 512-bit
14086 vectors with 32-bit and 64-bit elements.
14087 (vecdupssescalarmodesuffix): New mode attribute.
14088 (vec_dup<mode>): Use it.
14090 2018-01-11 H.J. Lu <hongjiu.lu@intel.com>
14093 * config/i386/i386.c (ix86_compute_frame_layout): Align stack
14094 frame if argument is passed on stack.
14096 2018-01-11 Jakub Jelinek <jakub@redhat.com>
14099 * ree.c (combine_reaching_defs): Optimize also
14100 reg2=exp; reg1=reg2; reg2=any_extend(reg1); into
14101 reg2=any_extend(exp); reg1=reg2;, formatting fix.
14103 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
14105 PR middle-end/83189
14106 * gimple-ssa-isolate-paths.c (isolate_path): Fix profile update.
14108 2018-01-11 Jan Hubicka <hubicka@ucw.cz>
14110 PR middle-end/83718
14111 * tree-inline.c (copy_cfg_body): Adjust num&den for scaling
14112 after they are computed.
14114 2018-01-11 Bin Cheng <bin.cheng@arm.com>
14116 PR tree-optimization/83695
14117 * gimple-loop-linterchange.cc
14118 (tree_loop_interchange::interchange_loops): Call scev_reset_htab to
14119 reset cached scev information after interchange.
14120 (pass_linterchange::execute): Remove call to scev_reset_htab.
14122 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14124 * config/arm/arm_neon.h (vfmlal_lane_low_u32, vfmlal_lane_high_u32,
14125 vfmlalq_laneq_low_u32, vfmlalq_lane_low_u32, vfmlal_laneq_low_u32,
14126 vfmlalq_laneq_high_u32, vfmlalq_lane_high_u32, vfmlal_laneq_high_u32,
14127 vfmlsl_lane_low_u32, vfmlsl_lane_high_u32, vfmlslq_laneq_low_u32,
14128 vfmlslq_lane_low_u32, vfmlsl_laneq_low_u32, vfmlslq_laneq_high_u32,
14129 vfmlslq_lane_high_u32, vfmlsl_laneq_high_u32): Define.
14130 * config/arm/arm_neon_builtins.def (vfmal_lane_low,
14131 vfmal_lane_lowv4hf, vfmal_lane_lowv8hf, vfmal_lane_high,
14132 vfmal_lane_highv4hf, vfmal_lane_highv8hf, vfmsl_lane_low,
14133 vfmsl_lane_lowv4hf, vfmsl_lane_lowv8hf, vfmsl_lane_high,
14134 vfmsl_lane_highv4hf, vfmsl_lane_highv8hf): New sets of builtins.
14135 * config/arm/iterators.md (VFMLSEL2, vfmlsel2): New mode attributes.
14136 (V_lane_reg): Likewise.
14137 * config/arm/neon.md (neon_vfm<vfml_op>l_lane_<vfml_half><VCVTF:mode>):
14139 (neon_vfm<vfml_op>l_lane_<vfml_half><vfmlsel2><mode>): Likewise.
14140 (vfmal_lane_low<mode>_intrinsic,
14141 vfmal_lane_low<vfmlsel2><mode>_intrinsic,
14142 vfmal_lane_high<vfmlsel2><mode>_intrinsic,
14143 vfmal_lane_high<mode>_intrinsic, vfmsl_lane_low<mode>_intrinsic,
14144 vfmsl_lane_low<vfmlsel2><mode>_intrinsic,
14145 vfmsl_lane_high<vfmlsel2><mode>_intrinsic,
14146 vfmsl_lane_high<mode>_intrinsic): New define_insns.
14148 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14150 * config/arm/arm-cpus.in (fp16fml): New feature.
14151 (ALL_SIMD): Add fp16fml.
14152 (armv8.2-a): Add fp16fml as an option.
14153 (armv8.3-a): Likewise.
14154 (armv8.4-a): Add fp16fml as part of fp16.
14155 * config/arm/arm.h (TARGET_FP16FML): Define.
14156 * config/arm/arm-c.c (arm_cpu_builtins): Define __ARM_FEATURE_FP16_FML
14158 * config/arm/arm-modes.def (V2HF): Define.
14159 * config/arm/arm_neon.h (vfmlal_low_u32, vfmlsl_low_u32,
14160 vfmlal_high_u32, vfmlsl_high_u32, vfmlalq_low_u32,
14161 vfmlslq_low_u32, vfmlalq_high_u32, vfmlslq_high_u32): Define.
14162 * config/arm/arm_neon_builtins.def (vfmal_low, vfmal_high,
14163 vfmsl_low, vfmsl_high): New set of builtins.
14164 * config/arm/iterators.md (PLUSMINUS): New code iterator.
14165 (vfml_op): New code attribute.
14166 (VFMLHALVES): New int iterator.
14167 (VFML, VFMLSEL): New mode attributes.
14168 (V_reg): Define mapping for V2HF.
14169 (V_hi, V_lo): New mode attributes.
14170 (VF_constraint): Likewise.
14171 (vfml_half, vfml_half_selector): New int attributes.
14172 * config/arm/neon.md (neon_vfm<vfml_op>l_<vfml_half><mode>): New
14174 (vfmal_low<mode>_intrinsic, vfmsl_high<mode>_intrinsic,
14175 vfmal_high<mode>_intrinsic, vfmsl_low<mode>_intrinsic):
14177 * config/arm/t-arm-elf (v8_fps): Add fp16fml.
14178 * config/arm/t-multilib (v8_2_a_simd_variants): Add fp16fml.
14179 * config/arm/unspecs.md (UNSPEC_VFML_LO, UNSPEC_VFML_HI): New unspecs.
14180 * doc/invoke.texi (ARM Options): Document fp16fml. Update armv8.4-a
14182 * doc/sourcebuild.texi (arm_fp16fml_neon_ok, arm_fp16fml_neon):
14183 Document new effective target and option set.
14185 2018-01-11 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14187 * config/arm/arm-cpus.in (armv8_4): New feature.
14188 (ARMv8_4a): New fgroup.
14189 (armv8.4-a): New arch.
14190 * config/arm/arm-tables.opt: Regenerate.
14191 * config/arm/t-aprofile: Add matching rules for -march=armv8.4-a.
14192 * config/arm/t-arm-elf (all_v8_archs): Add armv8.4-a.
14193 * config/arm/t-multilib (v8_4_a_simd_variants): New variable.
14194 Add matching rules for -march=armv8.4-a and extensions.
14195 * doc/invoke.texi (ARM Options): Document -march=armv8.4-a.
14197 2018-01-11 Oleg Endo <olegendo@gcc.gnu.org>
14200 * config/rx/rx.md (BW): New mode attribute.
14201 (sync_lock_test_and_setsi): Add mode suffix to insn output.
14203 2018-01-11 Richard Biener <rguenther@suse.de>
14205 PR tree-optimization/83435
14206 * graphite.c (canonicalize_loop_form): Ignore fake loop exit edges.
14207 * graphite-scop-detection.c (scop_detection::get_sese): Likewise.
14208 * tree-vrp.c (add_assert_info): Drop TREE_OVERFLOW if they appear.
14210 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14211 Alan Hayward <alan.hayward@arm.com>
14212 David Sherwood <david.sherwood@arm.com>
14214 * config/aarch64/aarch64.c (aarch64_address_info): Add a const_offset
14216 (aarch64_classify_address): Initialize it. Track polynomial offsets.
14217 (aarch64_print_address_internal): Use it to check for a zero offset.
14219 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14220 Alan Hayward <alan.hayward@arm.com>
14221 David Sherwood <david.sherwood@arm.com>
14223 * config/aarch64/aarch64-modes.def (NUM_POLY_INT_COEFFS): Set to 2.
14224 * config/aarch64/aarch64-protos.h (aarch64_initial_elimination_offset):
14225 Return a poly_int64 rather than a HOST_WIDE_INT.
14226 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a poly_int64
14227 rather than a HOST_WIDE_INT.
14228 * config/aarch64/aarch64.h (aarch64_frame): Protect with
14229 HAVE_POLY_INT_H rather than HOST_WIDE_INT. Change locals_offset,
14230 hard_fp_offset, frame_size, initial_adjust, callee_offset and
14231 final_offset from HOST_WIDE_INT to poly_int64.
14232 * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args): Use
14233 to_constant when getting the number of units in an Advanced SIMD
14235 (aarch64_builtin_vectorized_function): Check for a constant number
14237 * config/aarch64/aarch64-simd.md (mov<mode>): Handle polynomial
14239 (aarch64_ld<VSTRUCT:nregs>_lane<VALLDIF:mode>): Use the nunits
14240 attribute instead of GET_MODE_NUNITS.
14241 * config/aarch64/aarch64.c (aarch64_hard_regno_nregs)
14242 (aarch64_class_max_nregs): Use the constant_lowest_bound of the
14243 GET_MODE_SIZE for fixed-size registers.
14244 (aarch64_const_vec_all_same_in_range_p): Use const_vec_duplicate_p.
14245 (aarch64_hard_regno_call_part_clobbered, aarch64_classify_index)
14246 (aarch64_mode_valid_for_sched_fusion_p, aarch64_classify_address)
14247 (aarch64_legitimize_address_displacement, aarch64_secondary_reload)
14248 (aarch64_print_operand, aarch64_print_address_internal)
14249 (aarch64_address_cost, aarch64_rtx_costs, aarch64_register_move_cost)
14250 (aarch64_short_vector_p, aapcs_vfp_sub_candidate)
14251 (aarch64_simd_attr_length_rglist, aarch64_operands_ok_for_ldpstp):
14252 Handle polynomial GET_MODE_SIZE.
14253 (aarch64_hard_regno_caller_save_mode): Likewise. Return modes
14254 wider than SImode without modification.
14255 (tls_symbolic_operand_type): Use strip_offset instead of split_const.
14256 (aarch64_pass_by_reference, aarch64_layout_arg, aarch64_pad_reg_upward)
14257 (aarch64_gimplify_va_arg_expr): Assert that we don't yet handle
14258 passing and returning SVE modes.
14259 (aarch64_function_value, aarch64_layout_arg): Use gen_int_mode
14260 rather than GEN_INT.
14261 (aarch64_emit_probe_stack_range): Take the size as a poly_int64
14262 rather than a HOST_WIDE_INT, but call sorry if it isn't constant.
14263 (aarch64_allocate_and_probe_stack_space): Likewise.
14264 (aarch64_layout_frame): Cope with polynomial offsets.
14265 (aarch64_save_callee_saves, aarch64_restore_callee_saves): Take the
14266 start_offset as a poly_int64 rather than a HOST_WIDE_INT. Track
14267 polynomial offsets.
14268 (offset_9bit_signed_unscaled_p, offset_12bit_unsigned_scaled_p)
14269 (aarch64_offset_7bit_signed_scaled_p): Take the offset as a
14270 poly_int64 rather than a HOST_WIDE_INT.
14271 (aarch64_get_separate_components, aarch64_process_components)
14272 (aarch64_expand_prologue, aarch64_expand_epilogue)
14273 (aarch64_use_return_insn_p): Handle polynomial frame offsets.
14274 (aarch64_anchor_offset): New function, split out from...
14275 (aarch64_legitimize_address): ...here.
14276 (aarch64_builtin_vectorization_cost): Handle polynomial
14277 TYPE_VECTOR_SUBPARTS.
14278 (aarch64_simd_check_vect_par_cnst_half): Handle polynomial
14280 (aarch64_simd_make_constant, aarch64_expand_vector_init): Get the
14281 number of elements from the PARALLEL rather than the mode.
14282 (aarch64_shift_truncation_mask): Use GET_MODE_UNIT_BITSIZE
14283 rather than GET_MODE_BITSIZE.
14284 (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_ext)
14285 (aarch64_evpc_rev, aarch64_evpc_dup, aarch64_evpc_zip)
14286 (aarch64_expand_vec_perm_const_1): Handle polynomial
14287 d->perm.length () and d->perm elements.
14288 (aarch64_evpc_tbl): Likewise. Use nelt rather than GET_MODE_NUNITS.
14289 Apply to_constant to d->perm elements.
14290 (aarch64_simd_valid_immediate, aarch64_vec_fpconst_pow_of_2): Handle
14291 polynomial CONST_VECTOR_NUNITS.
14292 (aarch64_move_pointer): Take amount as a poly_int64 rather
14294 (aarch64_progress_pointer): Avoid temporary variable.
14295 * config/aarch64/aarch64.md (aarch64_<crc_variant>): Use
14296 the mode attribute instead of GET_MODE.
14298 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14299 Alan Hayward <alan.hayward@arm.com>
14300 David Sherwood <david.sherwood@arm.com>
14302 * config/aarch64/aarch64.c (aarch64_force_temporary): Assert that
14303 x exists before using it.
14304 (aarch64_add_constant_internal): Rename to...
14305 (aarch64_add_offset_1): ...this. Replace regnum with separate
14306 src and dest rtxes. Handle the case in which they're different,
14307 including when the offset is zero. Replace scratchreg with an rtx.
14308 Use 2 additions if there is no spare register into which we can
14309 move a 16-bit constant.
14310 (aarch64_add_constant): Delete.
14311 (aarch64_add_offset): Replace reg with separate src and dest
14312 rtxes. Take a poly_int64 offset instead of a HOST_WIDE_INT.
14313 Use aarch64_add_offset_1.
14314 (aarch64_add_sp, aarch64_sub_sp): Take the scratch register as
14315 an rtx rather than an int. Take the delta as a poly_int64
14316 rather than a HOST_WIDE_INT. Use aarch64_add_offset.
14317 (aarch64_expand_mov_immediate): Update uses of aarch64_add_offset.
14318 (aarch64_expand_prologue): Update calls to aarch64_sub_sp,
14319 aarch64_allocate_and_probe_stack_space and aarch64_add_offset.
14320 (aarch64_expand_epilogue): Update calls to aarch64_add_offset
14321 and aarch64_add_sp.
14322 (aarch64_output_mi_thunk): Use aarch64_add_offset rather than
14323 aarch64_add_constant.
14325 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14327 * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
14328 Use scalar_float_mode.
14330 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14332 * config/aarch64/aarch64-simd.md
14333 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): Avoid GET_MODE_NUNITS.
14334 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Likewise.
14335 (aarch64_fml<f16mac1>l_lane_lowv2sf): Likewise.
14336 (aarch64_fml<f16mac1>l_lane_highv2sf): Likewise.
14337 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Likewise.
14338 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Likewise.
14339 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Likewise.
14340 (aarch64_fml<f16mac1>l_laneq_highv2sf): Likewise.
14341 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Likewise.
14342 (aarch64_fml<f16mac1>lq_lane_highv4sf): Likewise.
14344 2018-01-11 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14347 * config/arm/arm.c (arm_declare_function_name): Set arch_to_print if
14348 targ_options->x_arm_arch_string is non NULL.
14350 2018-01-11 Tamar Christina <tamar.christina@arm.com>
14352 * config/aarch64/aarch64.h
14353 (AARCH64_FL_FOR_ARCH8_4): Add AARCH64_FL_DOTPROD.
14355 2018-01-11 Sudakshina Das <sudi.das@arm.com>
14358 * expmed.c (emit_store_flag_force): Swap if const op0
14359 and change VOIDmode to mode of op0.
14361 2018-01-11 Richard Sandiford <richard.sandiford@linaro.org>
14363 PR rtl-optimization/83761
14364 * caller-save.c (replace_reg_with_saved_mem): Pass bits rather
14365 than bytes to mode_for_size.
14367 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14369 PR middle-end/83189
14370 * gfortran.fortran-torture/compile/pr83189.f90: New testcase.
14371 * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Handle zero
14374 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14376 PR middle-end/83575
14377 * cfgrtl.c (rtl_verify_edges): Only verify fixability of partition
14378 when in layout mode.
14379 (cfg_layout_finalize): Do not verify cfg before we are out of layout.
14380 * cfgcleanup.c (try_optimize_cfg): Only verify flow info when doing
14383 2018-01-10 Michael Collison <michael.collison@arm.com>
14385 * config/aarch64/aarch64-modes.def (V2HF): New VECTOR_MODE.
14386 * config/aarch64/aarch64-option-extension.def: Add
14387 AARCH64_OPT_EXTENSION of 'fp16fml'.
14388 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14389 (__ARM_FEATURE_FP16_FML): Define if TARGET_F16FML is true.
14390 * config/aarch64/predicates.md (aarch64_lane_imm3): New predicate.
14391 * config/aarch64/constraints.md (Ui7): New constraint.
14392 * config/aarch64/iterators.md (VFMLA_W): New mode iterator.
14393 (VFMLA_SEL_W): Ditto.
14396 (VFMLA16_LOW): New int iterator.
14397 (VFMLA16_HIGH): Ditto.
14398 (UNSPEC_FMLAL): New unspec.
14399 (UNSPEC_FMLSL): Ditto.
14400 (UNSPEC_FMLAL2): Ditto.
14401 (UNSPEC_FMLSL2): Ditto.
14402 (f16mac): New code attribute.
14403 * config/aarch64/aarch64-simd-builtins.def
14404 (aarch64_fmlal_lowv2sf): Ditto.
14405 (aarch64_fmlsl_lowv2sf): Ditto.
14406 (aarch64_fmlalq_lowv4sf): Ditto.
14407 (aarch64_fmlslq_lowv4sf): Ditto.
14408 (aarch64_fmlal_highv2sf): Ditto.
14409 (aarch64_fmlsl_highv2sf): Ditto.
14410 (aarch64_fmlalq_highv4sf): Ditto.
14411 (aarch64_fmlslq_highv4sf): Ditto.
14412 (aarch64_fmlal_lane_lowv2sf): Ditto.
14413 (aarch64_fmlsl_lane_lowv2sf): Ditto.
14414 (aarch64_fmlal_laneq_lowv2sf): Ditto.
14415 (aarch64_fmlsl_laneq_lowv2sf): Ditto.
14416 (aarch64_fmlalq_lane_lowv4sf): Ditto.
14417 (aarch64_fmlsl_lane_lowv4sf): Ditto.
14418 (aarch64_fmlalq_laneq_lowv4sf): Ditto.
14419 (aarch64_fmlsl_laneq_lowv4sf): Ditto.
14420 (aarch64_fmlal_lane_highv2sf): Ditto.
14421 (aarch64_fmlsl_lane_highv2sf): Ditto.
14422 (aarch64_fmlal_laneq_highv2sf): Ditto.
14423 (aarch64_fmlsl_laneq_highv2sf): Ditto.
14424 (aarch64_fmlalq_lane_highv4sf): Ditto.
14425 (aarch64_fmlsl_lane_highv4sf): Ditto.
14426 (aarch64_fmlalq_laneq_highv4sf): Ditto.
14427 (aarch64_fmlsl_laneq_highv4sf): Ditto.
14428 * config/aarch64/aarch64-simd.md:
14429 (aarch64_fml<f16mac1>l<f16quad>_low<mode>): New pattern.
14430 (aarch64_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
14431 (aarch64_simd_fml<f16mac1>l<f16quad>_low<mode>): Ditto.
14432 (aarch64_simd_fml<f16mac1>l<f16quad>_high<mode>): Ditto.
14433 (aarch64_fml<f16mac1>l_lane_lowv2sf): Ditto.
14434 (aarch64_fml<f16mac1>l_lane_highv2sf): Ditto.
14435 (aarch64_simd_fml<f16mac>l_lane_lowv2sf): Ditto.
14436 (aarch64_simd_fml<f16mac>l_lane_highv2sf): Ditto.
14437 (aarch64_fml<f16mac1>lq_laneq_lowv4sf): Ditto.
14438 (aarch64_fml<f16mac1>lq_laneq_highv4sf): Ditto.
14439 (aarch64_simd_fml<f16mac>lq_laneq_lowv4sf): Ditto.
14440 (aarch64_simd_fml<f16mac>lq_laneq_highv4sf): Ditto.
14441 (aarch64_fml<f16mac1>l_laneq_lowv2sf): Ditto.
14442 (aarch64_fml<f16mac1>l_laneq_highv2sf): Ditto.
14443 (aarch64_simd_fml<f16mac>l_laneq_lowv2sf): Ditto.
14444 (aarch64_simd_fml<f16mac>l_laneq_highv2sf): Ditto.
14445 (aarch64_fml<f16mac1>lq_lane_lowv4sf): Ditto.
14446 (aarch64_fml<f16mac1>lq_lane_highv4sf): Ditto.
14447 (aarch64_simd_fml<f16mac>lq_lane_lowv4sf): Ditto.
14448 (aarch64_simd_fml<f16mac>lq_lane_highv4sf): Ditto.
14449 * config/aarch64/arm_neon.h (vfmlal_low_u32): New intrinsic.
14450 (vfmlsl_low_u32): Ditto.
14451 (vfmlalq_low_u32): Ditto.
14452 (vfmlslq_low_u32): Ditto.
14453 (vfmlal_high_u32): Ditto.
14454 (vfmlsl_high_u32): Ditto.
14455 (vfmlalq_high_u32): Ditto.
14456 (vfmlslq_high_u32): Ditto.
14457 (vfmlal_lane_low_u32): Ditto.
14458 (vfmlsl_lane_low_u32): Ditto.
14459 (vfmlal_laneq_low_u32): Ditto.
14460 (vfmlsl_laneq_low_u32): Ditto.
14461 (vfmlalq_lane_low_u32): Ditto.
14462 (vfmlslq_lane_low_u32): Ditto.
14463 (vfmlalq_laneq_low_u32): Ditto.
14464 (vfmlslq_laneq_low_u32): Ditto.
14465 (vfmlal_lane_high_u32): Ditto.
14466 (vfmlsl_lane_high_u32): Ditto.
14467 (vfmlal_laneq_high_u32): Ditto.
14468 (vfmlsl_laneq_high_u32): Ditto.
14469 (vfmlalq_lane_high_u32): Ditto.
14470 (vfmlslq_lane_high_u32): Ditto.
14471 (vfmlalq_laneq_high_u32): Ditto.
14472 (vfmlslq_laneq_high_u32): Ditto.
14473 * config/aarch64/aarch64.h (AARCH64_FL_F16SML): New flag.
14474 (AARCH64_FL_FOR_ARCH8_4): New.
14475 (AARCH64_ISA_F16FML): New ISA flag.
14476 (TARGET_F16FML): New feature flag for fp16fml.
14477 (doc/invoke.texi): Document new fp16fml option.
14479 2018-01-10 Michael Collison <michael.collison@arm.com>
14481 * config/aarch64/aarch64-builtins.c:
14482 (aarch64_types_ternopu_imm_qualifiers, TYPES_TERNOPUI): New.
14483 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14484 (__ARM_FEATURE_SHA3): Define if TARGET_SHA3 is true.
14485 * config/aarch64/aarch64.h (AARCH64_FL_SHA3): New flags.
14486 (AARCH64_ISA_SHA3): New ISA flag.
14487 (TARGET_SHA3): New feature flag for sha3.
14488 * config/aarch64/iterators.md (sha512_op): New int attribute.
14489 (CRYPTO_SHA512): New int iterator.
14490 (UNSPEC_SHA512H): New unspec.
14491 (UNSPEC_SHA512H2): Ditto.
14492 (UNSPEC_SHA512SU0): Ditto.
14493 (UNSPEC_SHA512SU1): Ditto.
14494 * config/aarch64/aarch64-simd-builtins.def
14495 (aarch64_crypto_sha512hqv2di): New builtin.
14496 (aarch64_crypto_sha512h2qv2di): Ditto.
14497 (aarch64_crypto_sha512su0qv2di): Ditto.
14498 (aarch64_crypto_sha512su1qv2di): Ditto.
14499 (aarch64_eor3qv8hi): Ditto.
14500 (aarch64_rax1qv2di): Ditto.
14501 (aarch64_xarqv2di): Ditto.
14502 (aarch64_bcaxqv8hi): Ditto.
14503 * config/aarch64/aarch64-simd.md:
14504 (aarch64_crypto_sha512h<sha512_op>qv2di): New pattern.
14505 (aarch64_crypto_sha512su0qv2di): Ditto.
14506 (aarch64_crypto_sha512su1qv2di): Ditto.
14507 (aarch64_eor3qv8hi): Ditto.
14508 (aarch64_rax1qv2di): Ditto.
14509 (aarch64_xarqv2di): Ditto.
14510 (aarch64_bcaxqv8hi): Ditto.
14511 * config/aarch64/arm_neon.h (vsha512hq_u64): New intrinsic.
14512 (vsha512h2q_u64): Ditto.
14513 (vsha512su0q_u64): Ditto.
14514 (vsha512su1q_u64): Ditto.
14515 (veor3q_u16): Ditto.
14516 (vrax1q_u64): Ditto.
14517 (vxarq_u64): Ditto.
14518 (vbcaxq_u16): Ditto.
14519 * config/arm/types.md (crypto_sha512): New type attribute.
14520 (crypto_sha3): Ditto.
14521 (doc/invoke.texi): Document new sha3 option.
14523 2018-01-10 Michael Collison <michael.collison@arm.com>
14525 * config/aarch64/aarch64-builtins.c:
14526 (aarch64_types_quadopu_imm_qualifiers, TYPES_QUADOPUI): New.
14527 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14528 (__ARM_FEATURE_SM3): Define if TARGET_SM4 is true.
14529 (__ARM_FEATURE_SM4): Define if TARGET_SM4 is true.
14530 * config/aarch64/aarch64.h (AARCH64_FL_SM4): New flags.
14531 (AARCH64_ISA_SM4): New ISA flag.
14532 (TARGET_SM4): New feature flag for sm4.
14533 * config/aarch64/aarch64-simd-builtins.def
14534 (aarch64_sm3ss1qv4si): Ditto.
14535 (aarch64_sm3tt1aq4si): Ditto.
14536 (aarch64_sm3tt1bq4si): Ditto.
14537 (aarch64_sm3tt2aq4si): Ditto.
14538 (aarch64_sm3tt2bq4si): Ditto.
14539 (aarch64_sm3partw1qv4si): Ditto.
14540 (aarch64_sm3partw2qv4si): Ditto.
14541 (aarch64_sm4eqv4si): Ditto.
14542 (aarch64_sm4ekeyqv4si): Ditto.
14543 * config/aarch64/aarch64-simd.md:
14544 (aarch64_sm3ss1qv4si): Ditto.
14545 (aarch64_sm3tt<sm3tt_op>qv4si): Ditto.
14546 (aarch64_sm3partw<sm3part_op>qv4si): Ditto.
14547 (aarch64_sm4eqv4si): Ditto.
14548 (aarch64_sm4ekeyqv4si): Ditto.
14549 * config/aarch64/iterators.md (sm3tt_op): New int iterator.
14550 (sm3part_op): Ditto.
14551 (CRYPTO_SM3TT): Ditto.
14552 (CRYPTO_SM3PART): Ditto.
14553 (UNSPEC_SM3SS1): New unspec.
14554 (UNSPEC_SM3TT1A): Ditto.
14555 (UNSPEC_SM3TT1B): Ditto.
14556 (UNSPEC_SM3TT2A): Ditto.
14557 (UNSPEC_SM3TT2B): Ditto.
14558 (UNSPEC_SM3PARTW1): Ditto.
14559 (UNSPEC_SM3PARTW2): Ditto.
14560 (UNSPEC_SM4E): Ditto.
14561 (UNSPEC_SM4EKEY): Ditto.
14562 * config/aarch64/constraints.md (Ui2): New constraint.
14563 * config/aarch64/predicates.md (aarch64_imm2): New predicate.
14564 * config/arm/types.md (crypto_sm3): New type attribute.
14565 (crypto_sm4): Ditto.
14566 * config/aarch64/arm_neon.h (vsm3ss1q_u32): New intrinsic.
14567 (vsm3tt1aq_u32): Ditto.
14568 (vsm3tt1bq_u32): Ditto.
14569 (vsm3tt2aq_u32): Ditto.
14570 (vsm3tt2bq_u32): Ditto.
14571 (vsm3partw1q_u32): Ditto.
14572 (vsm3partw2q_u32): Ditto.
14573 (vsm4eq_u32): Ditto.
14574 (vsm4ekeyq_u32): Ditto.
14575 (doc/invoke.texi): Document new sm4 option.
14577 2018-01-10 Michael Collison <michael.collison@arm.com>
14579 * config/aarch64/aarch64-arches.def (armv8.4-a): New architecture.
14580 * config/aarch64/aarch64.h (AARCH64_ISA_V8_4): New ISA flag.
14581 (AARCH64_FL_FOR_ARCH8_4): New.
14582 (AARCH64_FL_V8_4): New flag.
14583 (doc/invoke.texi): Document new armv8.4-a option.
14585 2018-01-10 Michael Collison <michael.collison@arm.com>
14587 * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
14588 (__ARM_FEATURE_AES): Define if TARGET_AES is true.
14589 (__ARM_FEATURE_SHA2): Define if TARGET_SHA2 is true.
14590 * config/aarch64/aarch64-option-extension.def: Add
14591 AARCH64_OPT_EXTENSION of 'sha2'.
14592 (aes): Add AARCH64_OPT_EXTENSION of 'aes'.
14593 (crypto): Disable sha2 and aes if crypto disabled.
14594 (crypto): Enable aes and sha2 if enabled.
14595 (simd): Disable sha2 and aes if simd disabled.
14596 * config/aarch64/aarch64.h (AARCH64_FL_AES, AARCH64_FL_SHA2):
14598 (AARCH64_ISA_AES, AARCH64_ISA_SHA2): New ISA flags.
14599 (TARGET_SHA2): New feature flag for sha2.
14600 (TARGET_AES): New feature flag for aes.
14601 * config/aarch64/aarch64-simd.md:
14602 (aarch64_crypto_aes<aes_op>v16qi): Make pattern
14603 conditional on TARGET_AES.
14604 (aarch64_crypto_aes<aesmc_op>v16qi): Ditto.
14605 (aarch64_crypto_sha1hsi): Make pattern conditional
14607 (aarch64_crypto_sha1hv4si): Ditto.
14608 (aarch64_be_crypto_sha1hv4si): Ditto.
14609 (aarch64_crypto_sha1su1v4si): Ditto.
14610 (aarch64_crypto_sha1<sha1_op>v4si): Ditto.
14611 (aarch64_crypto_sha1su0v4si): Ditto.
14612 (aarch64_crypto_sha256h<sha256_op>v4si): Ditto.
14613 (aarch64_crypto_sha256su0v4si): Ditto.
14614 (aarch64_crypto_sha256su1v4si): Ditto.
14615 (doc/invoke.texi): Document new aes and sha2 options.
14617 2018-01-10 Martin Sebor <msebor@redhat.com>
14619 PR tree-optimization/83781
14620 * gimple-fold.c (get_range_strlen): Avoid treating arrays of pointers
14623 2018-01-11 Martin Sebor <msebor@gmail.com>
14624 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
14626 PR tree-optimization/83501
14627 PR tree-optimization/81703
14629 * tree-ssa-strlen.c (get_string_cst): Rename...
14630 (get_string_len): ...to this. Handle global constants.
14631 (handle_char_store): Adjust.
14633 2018-01-10 Kito Cheng <kito.cheng@gmail.com>
14634 Jim Wilson <jimw@sifive.com>
14636 * config/riscv/riscv-protos.h (riscv_output_return): New.
14637 * config/riscv/riscv.c (struct machine_function): New naked_p field.
14638 (riscv_attribute_table, riscv_output_return),
14639 (riscv_handle_fndecl_attribute, riscv_naked_function_p),
14640 (riscv_allocate_stack_slots_for_args, riscv_warn_func_return): New.
14641 (riscv_compute_frame_info): Only compute frame->mask if not a naked
14643 (riscv_expand_prologue): Add early return for naked function.
14644 (riscv_expand_epilogue): Likewise.
14645 (riscv_function_ok_for_sibcall): Return false for naked function.
14646 (riscv_set_current_function): New.
14647 (TARGET_SET_CURRENT_FUNCTION, TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS),
14648 (TARGET_ATTRIBUTE_TABLE, TARGET_WARN_FUNC_RETURN): New.
14649 * config/riscv/riscv.md (simple_return): Call riscv_output_return.
14650 * doc/extend.texi (RISC-V Function Attributes): New.
14652 2018-01-10 Michael Meissner <meissner@linux.vnet.ibm.com>
14654 * config/rs6000/rs6000.c (is_complex_IBM_long_double): Explicitly
14655 check for 128-bit long double before checking TCmode.
14656 * config/rs6000/rs6000.h (FLOAT128_IEEE_P): Explicitly check for
14657 128-bit long doubles before checking TFmode or TCmode.
14658 (FLOAT128_IBM_P): Likewise.
14660 2018-01-10 Martin Sebor <msebor@redhat.com>
14662 PR tree-optimization/83671
14663 * builtins.c (c_strlen): Unconditionally return zero for the empty
14665 Use -Warray-bounds for warnings.
14666 * gimple-fold.c (get_range_strlen): Handle non-constant lengths
14667 for non-constant array indices with COMPONENT_REF, arrays of
14668 arrays, and pointers to arrays.
14669 (gimple_fold_builtin_strlen): Determine and set length range for
14670 non-constant character arrays.
14672 2018-01-10 Aldy Hernandez <aldyh@redhat.com>
14674 PR middle-end/81897
14675 * tree-ssa-uninit.c (convert_control_dep_chain_into_preds): Skip
14678 2018-01-10 Eric Botcazou <ebotcazou@adacore.com>
14680 * dwarf2out.c (dwarf2out_var_location): Do not pass NULL to fprintf.
14682 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
14685 * config/rs6000/rs6000.c (print_operand) <'y'>: Use
14686 VECTOR_MEM_ALTIVEC_OR_VSX_P.
14687 * config/rs6000/vsx.md (*vsx_le_perm_load_<mode> for VSX_D): Use
14688 indexed_or_indirect_operand predicate.
14689 (*vsx_le_perm_load_<mode> for VSX_W): Likewise.
14690 (*vsx_le_perm_load_v8hi): Likewise.
14691 (*vsx_le_perm_load_v16qi): Likewise.
14692 (*vsx_le_perm_store_<mode> for VSX_D): Likewise.
14693 (*vsx_le_perm_store_<mode> for VSX_W): Likewise.
14694 (*vsx_le_perm_store_v8hi): Likewise.
14695 (*vsx_le_perm_store_v16qi): Likewise.
14696 (eight unnamed splitters): Likewise.
14698 2018-01-10 Peter Bergner <bergner@vnet.ibm.com>
14700 * config/rs6000/x86intrin.h: Change #warning to #error. Update message.
14701 * config/rs6000/emmintrin.h: Likewise.
14702 * config/rs6000/mmintrin.h: Likewise.
14703 * config/rs6000/xmmintrin.h: Likewise.
14705 2018-01-10 David Malcolm <dmalcolm@redhat.com>
14708 * tree-core.h: Document EXPR_LOCATION_WRAPPER_P's usage of
14710 * tree.c (tree_nop_conversion): Return true for location wrapper
14712 (maybe_wrap_with_location): New function.
14713 (selftest::check_strip_nops): New function.
14714 (selftest::test_location_wrappers): New function.
14715 (selftest::tree_c_tests): Call it.
14716 * tree.h (STRIP_ANY_LOCATION_WRAPPER): New macro.
14717 (maybe_wrap_with_location): New decl.
14718 (EXPR_LOCATION_WRAPPER_P): New macro.
14719 (location_wrapper_p): New inline function.
14720 (tree_strip_any_location_wrapper): New inline function.
14722 2018-01-10 H.J. Lu <hongjiu.lu@intel.com>
14725 * config/i386/i386.c (ix86_compute_frame_layout): Always adjust
14726 stack_realign_offset for the largest alignment of stack slot
14728 (ix86_find_max_used_stack_alignment): New function.
14729 (ix86_finalize_stack_frame_flags): Use it. Set
14730 max_used_stack_alignment if we don't realign stack.
14731 * config/i386/i386.h (machine_function): Add
14732 max_used_stack_alignment.
14734 2018-01-10 Christophe Lyon <christophe.lyon@linaro.org>
14736 * config/arm/arm.opt (-mbranch-cost): New option.
14737 * config/arm/arm.h (BRANCH_COST): Take arm_branch_cost into
14740 2018-01-10 Segher Boessenkool <segher@kernel.crashing.org>
14743 * config/rs6000/rs6000.md (load_toc_v4_PIC_2, load_toc_v4_PIC_3b,
14744 load_toc_v4_PIC_3c): Wrap const term in CONST RTL.
14746 2018-01-10 Richard Biener <rguenther@suse.de>
14749 * dwarf2out.c (gen_subprogram_die): Hoist old_die && declaration
14750 early out so it also covers the case where we have a non-NULL
14753 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
14755 PR tree-optimization/83753
14756 * tree-vect-stmts.c (get_group_load_store_type): Use VMAT_CONTIGUOUS
14757 for non-strided grouped accesses if the number of elements is 1.
14759 2018-01-10 Jan Hubicka <hubicka@ucw.cz>
14762 * i386.c (ix86_vectorize_builtin_gather): Check TARGET_USE_GATHER.
14763 * i386.h (TARGET_USE_GATHER): Define.
14764 * x86-tune.def (X86_TUNE_USE_GATHER): New.
14766 2018-01-10 Martin Liska <mliska@suse.cz>
14769 * basic-block.h (CLEANUP_NO_PARTITIONING): New define.
14770 * bb-reorder.c (pass_reorder_blocks::execute): Do not clean up
14772 * cfgcleanup.c (try_optimize_cfg): Fix up partitioning if
14773 CLEANUP_NO_PARTITIONING is not set.
14775 2018-01-10 Richard Sandiford <richard.sandiford@linaro.org>
14777 * doc/rtl.texi: Remove documentation of (const ...) wrappers
14778 for vectors, as a partial revert of r254296.
14779 * rtl.h (const_vec_p): Delete.
14780 (const_vec_duplicate_p): Don't test for vector CONSTs.
14781 (unwrap_const_vec_duplicate, const_vec_series_p): Likewise.
14782 * expmed.c (make_tree): Likewise.
14785 * common.md (E, F): Use CONSTANT_P instead of checking for
14787 * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
14788 checking for CONST_VECTOR.
14790 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14792 PR middle-end/83575
14793 * predict.c (force_edge_cold): Handle in more sane way edges
14794 with no prediction.
14796 2018-01-09 Carl Love <cel@us.ibm.com>
14798 * config/rs6002/altivec.md (p8_vmrgow): Add support for V2DI, V2DF,
14800 (p8_vmrgew): Add support for V2DI, V2DF, V4SF types.
14801 * config/rs6000/rs6000-builtin.def: Add definitions for FLOAT2_V2DF,
14802 VMRGEW_V2DI, VMRGEW_V2DF, VMRGEW_V4SF, VMRGOW_V4SI, VMRGOW_V4SF,
14803 VMRGOW_V2DI, VMRGOW_V2DF. Remove definition for VMRGOW.
14804 * config/rs6000/rs6000-c.c (VSX_BUILTIN_VEC_FLOAT2,
14805 P8V_BUILTIN_VEC_VMRGEW, P8V_BUILTIN_VEC_VMRGOW): Add definitions.
14806 * config/rs6000/rs6000-protos.h: Add extern defition for
14807 rs6000_generate_float2_double_code.
14808 * config/rs6000/rs6000.c (rs6000_generate_float2_double_code): Add
14810 * config/rs6000/vsx.md (vsx_xvcdpsp): Add define_insn.
14811 (float2_v2df): Add define_expand.
14813 2018-01-09 Uros Bizjak <ubizjak@gmail.com>
14816 * combine.c (force_int_to_mode) <case ASHIFT>: Use mode instead of
14817 op_mode in the force_to_mode call.
14819 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
14821 * config/aarch64/aarch64.c (aarch64_evpc_trn): Use d.perm.series_p
14822 instead of checking each element individually.
14823 (aarch64_evpc_uzp): Likewise.
14824 (aarch64_evpc_zip): Likewise.
14825 (aarch64_evpc_ext): Likewise.
14826 (aarch64_evpc_rev): Likewise.
14827 (aarch64_evpc_dup): Test the encoding for a single duplicated element,
14828 instead of checking each element individually. Return true without
14830 (aarch64_vectorize_vec_perm_const): Use all_from_input_p to test
14831 whether all selected elements come from the same input, instead of
14832 checking each element individually. Remove calls to gen_rtx_REG,
14833 start_sequence and end_sequence and instead assert that no rtl is
14836 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
14838 * config/aarch64/aarch64.c (aarch64_legitimate_constant_p): Fix
14839 order of HIGH and CONST checks.
14841 2018-01-09 Richard Sandiford <richard.sandiford@linaro.org>
14843 * tree-vect-stmts.c (permute_vec_elements): Create a fresh variable
14844 if the destination isn't an SSA_NAME.
14846 2018-01-09 Richard Biener <rguenther@suse.de>
14848 PR tree-optimization/83668
14849 * graphite.c (canonicalize_loop_closed_ssa): Add edge argument,
14851 (canonicalize_loop_form): ... here, renamed from ...
14852 (canonicalize_loop_closed_ssa_form): ... this and amended to
14853 swap successor edges for loop exit blocks to make us use
14854 the RPO order we need for initial schedule generation.
14856 2018-01-09 Joseph Myers <joseph@codesourcery.com>
14858 PR tree-optimization/64811
14859 * match.pd: When optimizing comparisons with Inf, avoid
14860 introducing or losing exceptions from comparisons with NaN.
14862 2018-01-09 Martin Liska <mliska@suse.cz>
14865 * asan.c (shadow_mem_size): Add gcc_assert.
14867 2018-01-09 Georg-Johann Lay <avr@gjlay.de>
14869 Don't save registers in main().
14872 * doc/invoke.texi (AVR Options) [-mmain-is-OS_task]: Document it.
14873 * config/avr/avr.opt (-mmain-is-OS_task): New target option.
14874 * config/avr/avr.c (avr_set_current_function): Don't error if
14875 naked, OS_task or OS_main are specified at the same time.
14876 (avr_function_ok_for_sibcall): Don't disable sibcalls for OS_task,
14878 (avr_insert_attributes) [-mmain-is-OS_task] <main>: Add OS_task
14880 * common/config/avr/avr-common.c (avr_option_optimization_table):
14881 Switch on -mmain-is-OS_task for optimizing compilations.
14883 2018-01-09 Richard Biener <rguenther@suse.de>
14885 PR tree-optimization/83572
14886 * graphite.c: Include cfganal.h.
14887 (graphite_transform_loops): Connect infinite loops to exit
14888 and remove fake edges at the end.
14890 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14892 * ipa-inline.c (edge_badness): Revert accidental checkin.
14894 2018-01-09 Jan Hubicka <hubicka@ucw.cz>
14897 * ipa-comdats.c (set_comdat_group): Only set comdat group of real
14898 symbols; not inline clones.
14900 2018-01-09 Jakub Jelinek <jakub@redhat.com>
14903 * modulo-sched.c (schedule_reg_moves): Punt if we'd need to move
14904 hard registers. Formatting fixes.
14906 PR preprocessor/83722
14907 * gcc.c (try_generate_repro): Pass
14908 &temp_stderr_files[RETRY_ICE_ATTEMPTS - 1] rather than
14909 &temp_stdout_files[RETRY_ICE_ATTEMPTS - 1] as last argument to
14912 2018-01-08 Monk Chiang <sh.chiang04@gmail.com>
14913 Kito Cheng <kito.cheng@gmail.com>
14915 * config/riscv/riscv.c (machine_function::is_leaf): Remove field.
14916 (riscv_leaf_function_p): Delete.
14917 (riscv_function_ok_for_sibcall): Return false when TARGET_SAVE_RESTORE.
14919 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14921 * config/rs6000/rs6000-string.c (do_load_for_compare_from_addr): New
14923 (do_ifelse): New function.
14924 (do_isel): New function.
14925 (do_sub3): New function.
14926 (do_add3): New function.
14927 (do_load_mask_compare): New function.
14928 (do_overlap_load_compare): New function.
14929 (expand_compare_loop): New function.
14930 (expand_block_compare): Call expand_compare_loop() when appropriate.
14931 * config/rs6000/rs6000.opt (-mblock-compare-inline-limit): Change
14932 option description.
14933 (-mblock-compare-inline-loop-limit): New option.
14935 2018-01-08 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
14938 * config/rs6000/altivec.md (*altivec_vpermr_<mode>_internal):
14939 Reverse order of second and third operands in first alternative.
14940 * config/rs6000/rs6000.c (rs6000_expand_vector_set): Reverse order
14941 of first and second elements in UNSPEC_VPERMR vector.
14942 (altivec_expand_vec_perm_le): Likewise.
14944 2018-01-08 Jeff Law <law@redhat.com>
14946 PR rtl-optimizatin/81308
14947 * tree-switch-conversion.c (cfg_altered): New file scoped static.
14948 (process_switch): If group_case_labels makes a change, then set
14950 (pass_convert_switch::execute): If a switch is converted, then
14951 set cfg_altered. Return TODO_cfg_cleanup if cfg_altered is true.
14953 PR rtl-optimization/81308
14954 * recog.c (split_all_insns): Conditionally cleanup the CFG after
14957 2018-01-08 Vidya Praveen <vidyapraveen@arm.com>
14959 PR target/83663 - Revert r255946
14960 * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
14961 generation for cases where splatting a value is not useful.
14962 * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
14963 across a vec_duplicate and a paradoxical subreg forming a vector
14964 mode to a vec_concat.
14966 2018-01-08 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
14968 * config/arm/t-aprofile (MULTILIB_MATCHES): Add mapping rules for
14969 -march=armv8.3-a variants.
14970 * config/arm/t-multilib: Likewise.
14971 * config/arm/t-arm-elf: Likewise. Handle dotprod extension.
14973 2018-01-08 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
14975 * config/rs6000/rs6000.md (cceq_ior_compare): Remove * so I can use it
14977 (cceq_ior_compare_complement): Give it a name so I can use it, and
14978 change boolean_or_operator predicate to boolean_operator so it can
14979 be used to generate a crand.
14980 (eqne): New code iterator.
14981 (bd/bd_neg): New code_attrs.
14982 (<bd>_<mode>): New name for ctr<mode>_internal[12] now combined into
14983 a single define_insn.
14984 (<bd>tf_<mode>): A new insn pattern for the conditional form branch
14985 decrement (bdnzt/bdnzf/bdzt/bdzf).
14986 * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Updated
14987 with the new names of the branch decrement patterns, and added the
14988 names of the branch decrement conditional patterns.
14990 2018-01-08 Richard Biener <rguenther@suse.de>
14992 PR tree-optimization/83563
14993 * graphite.c (canonicalize_loop_closed_ssa_form): Reset the SCEV
14996 2018-01-08 Richard Biener <rguenther@suse.de>
14998 PR middle-end/83713
14999 * convert.c (do_narrow): Properly guard TYPE_OVERFLOW_WRAPS checks.
15001 2018-01-08 Richard Biener <rguenther@suse.de>
15003 PR tree-optimization/83685
15004 * tree-ssa-pre.c (create_expression_by_pieces): Do not insert
15005 references to abnormals.
15007 2018-01-08 Richard Biener <rguenther@suse.de>
15010 * dwarf2out.c (output_indirect_strings): Handle empty
15011 skeleton_debug_str_hash.
15012 (dwarf2out_early_finish): Index strings for -gsplit-dwarf.
15014 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
15016 * config/arc/arc.c (TARGET_TRAMPOLINE_ADJUST_ADDRESS): Delete.
15017 (emit_store_direct): Likewise.
15018 (arc_trampoline_adjust_address): Likewise.
15019 (arc_asm_trampoline_template): New function.
15020 (arc_initialize_trampoline): Use asm_trampoline_template.
15021 (TARGET_ASM_TRAMPOLINE_TEMPLATE): Define.
15022 * config/arc/arc.h (TRAMPOLINE_SIZE): Adjust to 16.
15023 * config/arc/arc.md (flush_icache): Delete pattern.
15025 2018-01-08 Claudiu Zissulescu <claziss@synopsys.com>
15027 * config/arc/arc-c.def (__ARC_UNALIGNED__): New define.
15028 * config/arc/arc.h (STRICT_ALIGNMENT): Control this macro using
15031 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
15034 * config/epiphany/epiphany.h (make_pass_mode_switch_use): Guard
15035 by not USED_FOR_TARGET.
15036 (make_pass_resolve_sw_modes): Likewise.
15038 2018-01-08 Sebastian Huber <sebastian.huber@embedded-brains.de>
15040 * config/nios2/nios2.h (nios2_section_threshold): Guard by not
15043 2018-01-08 Richard Biener <rguenther@suse.de>
15045 PR middle-end/83580
15046 * tree-data-ref.c (split_constant_offset): Remove STRIP_NOPS.
15048 2018-01-08 Richard Biener <rguenther@suse.de>
15050 PR middle-end/83517
15051 * match.pd ((t * 2) / 2) -> t): Add missing :c.
15053 2018-01-06 Aldy Hernandez <aldyh@redhat.com>
15055 PR middle-end/81897
15056 * tree-ssa-uninit.c (compute_control_dep_chain): Do not bail on
15057 basic blocks with a small number of successors.
15058 (convert_control_dep_chain_into_preds): Improve handling of
15060 (dump_predicates): Split apart into...
15061 (dump_pred_chain): ...here...
15062 (dump_pred_info): ...and here.
15063 (can_one_predicate_be_invalidated_p): Add debugging printfs.
15064 (can_chain_union_be_invalidated_p): Improve check for invalidation
15066 (uninit_uses_cannot_happen): Avoid unnecessary if
15067 convert_control_dep_chain_into_preds yielded nothing.
15069 2018-01-06 Martin Sebor <msebor@redhat.com>
15071 PR tree-optimization/83640
15072 * gimple-ssa-warn-restrict.c (builtin_access::builtin_access): Avoid
15073 subtracting negative offset from size.
15074 (builtin_access::overlap): Adjust offset bounds of the access to fall
15075 within the size of the object if possible.
15077 2018-01-06 Richard Sandiford <richard.sandiford@linaro.org>
15079 PR rtl-optimization/83699
15080 * expmed.c (extract_bit_field_1): Restrict the vector usage of
15081 extract_bit_field_as_subreg to cases in which the extracted
15082 value is also a vector.
15084 * lra-constraints.c (process_alt_operands): Test for the equivalence
15085 substitutions when detecting a possible reload cycle.
15087 2018-01-06 Jakub Jelinek <jakub@redhat.com>
15090 * toplev.c (process_options): Don't enable debug_nonbind_markers_p
15091 by default if flag_selective_schedling{,2}. Formatting fixes.
15093 PR rtl-optimization/83682
15094 * rtl.h (const_vec_duplicate_p): Only return true for VEC_DUPLICATE
15095 if it has non-VECTOR_MODE element mode.
15096 (vec_duplicate_p): Likewise.
15098 PR middle-end/83694
15099 * cfgexpand.c (expand_debug_expr): Punt if mode1 is VOIDmode
15100 and bitsize might be greater than MAX_BITSIZE_MODE_ANY_INT.
15102 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15105 * config/i386/i386-builtin.def
15106 (__builtin_ia32_vgf2p8affineinvqb_v64qi,
15107 __builtin_ia32_vgf2p8affineqb_v64qi, __builtin_ia32_vgf2p8mulb_v64qi):
15108 Require also OPTION_MASK_ISA_AVX512F in addition to
15109 OPTION_MASK_ISA_GFNI.
15110 (__builtin_ia32_vgf2p8affineinvqb_v16qi_mask,
15111 __builtin_ia32_vgf2p8affineqb_v16qi_mask): Require
15112 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_SSE in addition
15113 to OPTION_MASK_ISA_GFNI.
15114 (__builtin_ia32_vgf2p8mulb_v32qi_mask): Require
15115 OPTION_MASK_ISA_AVX512VL in addition to OPTION_MASK_ISA_GFNI and
15116 OPTION_MASK_ISA_AVX512BW.
15117 (__builtin_ia32_vgf2p8mulb_v16qi_mask): Require
15118 OPTION_MASK_ISA_AVX512VL instead of OPTION_MASK_ISA_AVX512BW in
15119 addition to OPTION_MASK_ISA_GFNI.
15120 (__builtin_ia32_vgf2p8affineinvqb_v16qi,
15121 __builtin_ia32_vgf2p8affineqb_v16qi, __builtin_ia32_vgf2p8mulb_v16qi):
15122 Require OPTION_MASK_ISA_SSE2 instead of OPTION_MASK_ISA_SSE in addition
15123 to OPTION_MASK_ISA_GFNI.
15124 * config/i386/i386.c (def_builtin): Change to builtin isa/isa2 being
15125 a requirement for all ISAs rather than any of them with a few
15127 (ix86_add_new_builtins): Clear OPTION_MASK_ISA_64BIT from isa before
15129 (ix86_expand_builtin): Require all ISAs from builtin's isa and isa2
15130 bitmasks to be enabled with 3 exceptions, instead of requiring any
15131 enabled ISA with lots of exceptions.
15132 * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
15133 vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>):
15134 Change avx512bw in isa attribute to avx512f.
15135 * config/i386/sgxintrin.h: Add license boilerplate.
15136 * config/i386/vaesintrin.h: Likewise. Fix macro spelling __AVX512F
15137 to __AVX512F__ and __AVX512VL to __AVX512VL__.
15138 (_mm256_aesdec_epi128, _mm256_aesdeclast_epi128, _mm256_aesenc_epi128,
15139 _mm256_aesenclast_epi128): Enable temporarily avx if __AVX__ is not
15141 * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8,
15142 _mm_gf2p8affineinv_epi64_epi8, _mm_gf2p8affine_epi64_epi8): Enable
15143 temporarily sse2 rather than sse if not enabled already.
15146 * config/i386/sse.md (VI248_VLBW): Rename to ...
15147 (VI248_AVX512VL): ... this. Don't guard V32HI with TARGET_AVX512BW.
15148 (vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
15149 vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
15150 vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
15151 vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): Use VI248_AVX512VL
15152 mode iterator instead of VI248_VLBW.
15154 2018-01-05 Jan Hubicka <hubicka@ucw.cz>
15156 * ipa-fnsummary.c (record_modified_bb_info): Add OP.
15157 (record_modified): Skip clobbers; add debug output.
15158 (param_change_prob): Use sreal frequencies.
15160 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15162 * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
15163 punt for user-aligned variables.
15165 2018-01-05 Richard Sandiford <richard.sandiford@linaro.org>
15167 * tree-chrec.c (chrec_contains_symbols): Return true for
15170 2018-01-05 Sudakshina Das <sudi.das@arm.com>
15173 * simplify-rtx.c (simplify_relational_operation_1): Add simplifications
15174 of (x|y) == x for BICS pattern.
15176 2018-01-05 Jakub Jelinek <jakub@redhat.com>
15178 PR tree-optimization/83605
15179 * gimple-ssa-strength-reduction.c: Include tree-eh.h.
15180 (find_candidates_dom_walker::before_dom_children): Ignore stmts that
15183 2018-01-05 Sebastian Huber <sebastian.huber@embedded-brains.de>
15185 * config.gcc (epiphany-*-elf*): Add (epiphany-*-rtems*) configuration.
15186 * config/epiphany/rtems.h: New file.
15188 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15189 Uros Bizjak <ubizjak@gmail.com>
15192 * config/i386/i386.md (*<rotate_insn>hi3_1 splitter): Use
15193 QIreg_operand instead of register_operand predicate.
15194 * config/i386/i386.c (ix86_rop_should_change_byte_p,
15195 set_rop_modrm_reg_bits, ix86_mitigate_rop): Use -mmitigate-rop in
15196 comments instead of -fmitigate[-_]rop.
15198 2018-01-04 Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
15201 * cgraphunit.c (symbol_table::compile): Switch to text_section
15202 before calling assembly_start debug hook.
15203 * run-rtl-passes.c (run_rtl_passes): Likewise.
15206 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15208 * tree-vrp.c (extract_range_from_binary_expr_1): Check
15209 range_int_cst_p rather than !symbolic_range_p before calling
15210 extract_range_from_multiplicative_op_1.
15212 2018-01-04 Jeff Law <law@redhat.com>
15214 * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Remove
15215 redundant test in assertion.
15217 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15219 * doc/rtl.texi: Document machine_mode wrapper classes.
15221 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15223 * fold-const.c (fold_ternary_loc): Check tree_fits_uhwi_p before
15224 using tree_to_uhwi.
15226 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15228 * tree-ssa-forwprop.c (is_combined_permutation_identity): Allow
15229 the VEC_PERM_EXPR fold to fail.
15231 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15234 * bb-reorder.c (insert_section_boundary_note): Set has_bb_partition
15235 to switched_sections.
15237 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15240 * config/arm/arm.c (arm_vectorize_vec_perm_const): Fix inverted
15241 test for d.testing.
15243 2018-01-04 Peter Bergner <bergner@vnet.ibm.com>
15246 * config/rs6000/rs6000.c (rs6000_discover_homogeneous_aggregate): Do not
15247 allow arguments in FP registers if TARGET_HARD_FLOAT is false.
15249 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15252 * cfgexpand.c (expand_debug_expr) <case BIT_FIELD_REF>: Punt if mode
15253 is BLKmode and bitpos not zero or mode change is needed.
15255 2018-01-04 Richard Sandiford <richard.sandiford@linaro.org>
15258 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): Require
15261 2018-01-04 Uros Bizjak <ubizjak@gmail.com>
15264 * config/alpha/alpha.md (*sadd<modesuffix>): Use ASHIFT
15265 instead of MULT rtx. Update all corresponding splitters.
15266 (*saddl_se): Ditto.
15267 (*ssub<modesuffix>): Ditto.
15268 (*ssubl_se): Ditto.
15269 (*cmp_sadd_di): Update split patterns.
15270 (*cmp_sadd_si): Ditto.
15271 (*cmp_sadd_sidi): Ditto.
15272 (*cmp_ssub_di): Ditto.
15273 (*cmp_ssub_si): Ditto.
15274 (*cmp_ssub_sidi): Ditto.
15275 * config/alpha/predicates.md (const23_operand): New predicate.
15276 * config/alpha/alpha.c (alpha_rtx_costs) [PLUS, MINUS]:
15277 Look for ASHIFT, not MULT inner operand.
15278 (alpha_split_conditional_move): Update for *sadd<modesuffix> change.
15280 2018-01-04 Martin Liska <mliska@suse.cz>
15282 PR gcov-profile/83669
15283 * gcov.c (output_intermediate_file): Add version to intermediate
15285 * doc/gcov.texi: Document new field 'version' in intermediate
15286 file format. Fix location of '-k' option of gcov command.
15288 2018-01-04 Martin Liska <mliska@suse.cz>
15291 * ipa-icf.c (sem_function::merge): Do not cross comdat boundary.
15293 2018-01-04 Jakub Jelinek <jakub@redhat.com>
15295 * gimple-ssa-sprintf.c (parse_directive): Cast second dir.len to uhwi.
15297 2018-01-03 Martin Sebor <msebor@redhat.com>
15299 PR tree-optimization/83655
15300 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call): Avoid
15301 checking calls with invalid arguments.
15303 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15305 * tree-vect-stmts.c (vect_get_store_rhs): New function.
15306 (vectorizable_mask_load_store): Delete.
15307 (vectorizable_call): Return false for masked loads and stores.
15308 (vectorizable_store): Handle IFN_MASK_STORE. Use vect_get_store_rhs
15309 instead of gimple_assign_rhs1.
15310 (vectorizable_load): Handle IFN_MASK_LOAD.
15311 (vect_transform_stmt): Don't set is_store for call_vec_info_type.
15313 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15315 * tree-vect-stmts.c (vect_build_gather_load_calls): New function,
15317 (vectorizable_mask_load_store): ...here.
15318 (vectorizable_load): ...and here.
15320 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15322 * tree-vect-stmts.c (vect_build_all_ones_mask)
15323 (vect_build_zero_merge_argument): New functions, split out from...
15324 (vectorizable_load): ...here.
15326 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15328 * tree-vect-stmts.c (vect_check_store_rhs): New function,
15330 (vectorizable_mask_load_store): ...here.
15331 (vectorizable_store): ...and here.
15333 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15335 * tree-vect-stmts.c (vect_check_load_store_mask): New function,
15337 (vectorizable_mask_load_store): ...here.
15339 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15341 * tree-vectorizer.h (vec_load_store_type): Moved from tree-vec-stmts.c
15342 (vect_model_store_cost): Take a vec_load_store_type instead of a
15344 * tree-vect-stmts.c (vec_load_store_type): Move to tree-vectorizer.h.
15345 (vect_model_store_cost): Take a vec_load_store_type instead of a
15347 (vectorizable_mask_load_store): Update accordingly.
15348 (vectorizable_store): Likewise.
15349 * tree-vect-slp.c (vect_analyze_slp_cost_1): Update accordingly.
15351 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15353 * tree-vect-loop.c (vect_transform_loop): Stub out scalar
15354 IFN_MASK_LOAD calls here rather than...
15355 * tree-vect-stmts.c (vectorizable_mask_load_store): ...here.
15357 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15358 Alan Hayward <alan.hayward@arm.com>
15359 David Sherwood <david.sherwood@arm.com>
15361 * expmed.c (extract_bit_field_1): For vector extracts,
15362 fall back to extract_bit_field_as_subreg if vec_extract
15365 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15366 Alan Hayward <alan.hayward@arm.com>
15367 David Sherwood <david.sherwood@arm.com>
15369 * lra-spills.c (pseudo_reg_slot_compare): Sort slots by whether
15370 they are variable or constant sized.
15371 (assign_stack_slot_num_and_sort_pseudos): Don't reuse variable-sized
15372 slots for constant-sized data.
15374 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15375 Alan Hayward <alan.hayward@arm.com>
15376 David Sherwood <david.sherwood@arm.com>
15378 * tree-vect-patterns.c (vect_recog_mask_conversion_pattern): When
15379 handling COND_EXPRs with boolean comparisons, try to find a better
15380 basis for the mask type than the boolean itself.
15382 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15384 * doc/rtl.texi (MAX_BITSIZE_MODE_ANY_MODE): Describe how the default
15385 is calculated and how it can be overridden.
15386 * genmodes.c (max_bitsize_mode_any_mode): New variable.
15387 (create_modes): Initialize it from MAX_BITSIZE_MODE_ANY_MODE,
15389 (emit_max_int): Use it to set the output MAX_BITSIZE_MODE_ANY_MODE,
15392 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15393 Alan Hayward <alan.hayward@arm.com>
15394 David Sherwood <david.sherwood@arm.com>
15396 * config/aarch64/aarch64-protos.h (aarch64_output_simd_mov_immediate):
15397 Remove the mode argument.
15398 (aarch64_simd_valid_immediate): Remove the mode and inverse
15400 * config/aarch64/iterators.md (bitsize): New iterator.
15401 * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>, and<mode>3)
15402 (ior<mode>3): Update calls to aarch64_output_simd_mov_immediate.
15403 * config/aarch64/constraints.md (Do, Db, Dn): Update calls to
15404 aarch64_simd_valid_immediate.
15405 * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Likewise.
15406 (aarch64_reg_or_bic_imm): Likewise.
15407 * config/aarch64/aarch64.c (simd_immediate_info): Replace mvn
15408 with an insn_type enum and msl with a modifier_type enum.
15409 Replace element_width with a scalar_mode. Change the shift
15410 to unsigned int. Add constructors for scalar_float_mode and
15411 scalar_int_mode elements.
15412 (aarch64_vect_float_const_representable_p): Delete.
15413 (aarch64_can_const_movi_rtx_p)
15414 (aarch64_simd_scalar_immediate_valid_for_move)
15415 (aarch64_simd_make_constant): Update call to
15416 aarch64_simd_valid_immediate.
15417 (aarch64_advsimd_valid_immediate_hs): New function.
15418 (aarch64_advsimd_valid_immediate): Likewise.
15419 (aarch64_simd_valid_immediate): Remove mode and inverse
15420 arguments. Rewrite to use the above. Use const_vec_duplicate_p
15421 to detect duplicated constants and use aarch64_float_const_zero_rtx_p
15422 and aarch64_float_const_representable_p on the result.
15423 (aarch64_output_simd_mov_immediate): Remove mode argument.
15424 Update call to aarch64_simd_valid_immediate and use of
15425 simd_immediate_info.
15426 (aarch64_output_scalar_simd_mov_immediate): Update call
15429 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15430 Alan Hayward <alan.hayward@arm.com>
15431 David Sherwood <david.sherwood@arm.com>
15433 * machmode.h (mode_precision): Prefix with CONST_MODE_PRECISION.
15434 (mode_nunits): Likewise CONST_MODE_NUNITS.
15435 * machmode.def (ADJUST_NUNITS): Document.
15436 * genmodes.c (mode_data::need_nunits_adj): New field.
15437 (blank_mode): Update accordingly.
15438 (adj_nunits): New variable.
15439 (print_maybe_const_decl): Replace CATEGORY with a NEEDS_ADJ
15441 (emit_mode_size_inline): Set need_bytesize_adj for all modes
15442 listed in adj_nunits.
15443 (emit_mode_nunits_inline): Set need_nunits_adj for all modes
15444 listed in adj_nunits. Don't emit case statements for such modes.
15445 (emit_insn_modes_h): Emit definitions of CONST_MODE_NUNITS
15446 and CONST_MODE_PRECISION. Make CONST_MODE_SIZE expand to
15447 nothing if adj_nunits is nonnull.
15448 (emit_mode_precision, emit_mode_nunits): Use print_maybe_const_decl.
15449 (emit_mode_unit_size, emit_mode_base_align, emit_mode_ibit)
15450 (emit_mode_fbit): Update use of print_maybe_const_decl.
15451 (emit_move_size): Likewise. Treat the array as non-const
15453 (emit_mode_adjustments): Handle adj_nunits.
15455 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15457 * machmode.def (VECTOR_MODES_WITH_PREFIX): Document.
15458 * genmodes.c (VECTOR_MODES_WITH_PREFIX): New macro.
15459 (VECTOR_MODES): Use it.
15460 (make_vector_modes): Take the prefix as an argument.
15462 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15463 Alan Hayward <alan.hayward@arm.com>
15464 David Sherwood <david.sherwood@arm.com>
15466 * mode-classes.def (MODE_VECTOR_BOOL): New mode class.
15467 * machmode.h (INTEGRAL_MODE_P, VECTOR_MODE_P): Return true
15468 for MODE_VECTOR_BOOL.
15469 * machmode.def (VECTOR_BOOL_MODE): Document.
15470 * genmodes.c (VECTOR_BOOL_MODE): New macro.
15471 (make_vector_bool_mode): New function.
15472 (complete_mode, emit_mode_wider, emit_mode_adjustments): Handle
15474 * lto-streamer-in.c (lto_input_mode_table): Likewise.
15475 * rtx-vector-builder.c (rtx_vector_builder::find_cached_value):
15477 * stor-layout.c (int_mode_for_mode): Likewise.
15478 * tree.c (build_vector_type_for_mode): Likewise.
15479 * varasm.c (output_constant_pool_2): Likewise.
15480 * emit-rtl.c (init_emit_once): Make sure that CONST1_RTX (BImode) and
15481 CONSTM1_RTX (BImode) are the same thing. Initialize const_tiny_rtx
15482 for MODE_VECTOR_BOOL.
15483 * expr.c (expand_expr_real_1): Use VECTOR_MODE_P instead of a list
15484 of mode class checks.
15485 * tree-vect-generic.c (expand_vector_operation): Use VECTOR_MODE_P
15486 instead of a list of mode class checks.
15487 (expand_vector_scalar_condition): Likewise.
15488 (type_for_widest_vector_mode): Handle BImode as an inner mode.
15490 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15491 Alan Hayward <alan.hayward@arm.com>
15492 David Sherwood <david.sherwood@arm.com>
15494 * machmode.h (mode_size): Change from unsigned short to
15496 (mode_to_bytes): Return a poly_uint16 rather than an unsigned short.
15497 (GET_MODE_SIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
15498 or if measurement_type is not polynomial.
15499 (fixed_size_mode::includes_p): Check for constant-sized modes.
15500 * genmodes.c (emit_mode_size_inline): Make mode_size_inline
15501 return a poly_uint16 rather than an unsigned short.
15502 (emit_mode_size): Change the type of mode_size from unsigned short
15503 to poly_uint16_pod. Use ZERO_COEFFS for the initializer.
15504 (emit_mode_adjustments): Cope with polynomial vector sizes.
15505 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15507 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15509 * auto-inc-dec.c (try_merge): Treat GET_MODE_SIZE as polynomial.
15510 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Likewise.
15511 * caller-save.c (setup_save_areas): Likewise.
15512 (replace_reg_with_saved_mem): Likewise.
15513 * calls.c (emit_library_call_value_1): Likewise.
15514 * combine-stack-adj.c (combine_stack_adjustments_for_block): Likewise.
15515 * combine.c (simplify_set, make_extraction, simplify_shift_const_1)
15516 (gen_lowpart_for_combine): Likewise.
15517 * convert.c (convert_to_integer_1): Likewise.
15518 * cse.c (equiv_constant, cse_insn): Likewise.
15519 * cselib.c (autoinc_split, cselib_hash_rtx): Likewise.
15520 (cselib_subst_to_values): Likewise.
15521 * dce.c (word_dce_process_block): Likewise.
15522 * df-problems.c (df_word_lr_mark_ref): Likewise.
15523 * dwarf2cfi.c (init_one_dwarf_reg_size): Likewise.
15524 * dwarf2out.c (multiple_reg_loc_descriptor, mem_loc_descriptor)
15525 (concat_loc_descriptor, concatn_loc_descriptor, loc_descriptor)
15526 (rtl_for_decl_location): Likewise.
15527 * emit-rtl.c (gen_highpart, widen_memory_access): Likewise.
15528 * expmed.c (extract_bit_field_1, extract_integral_bit_field): Likewise.
15529 * expr.c (emit_group_load_1, clear_storage_hints): Likewise.
15530 (emit_move_complex, emit_move_multi_word, emit_push_insn): Likewise.
15531 (expand_expr_real_1): Likewise.
15532 * function.c (assign_parm_setup_block_p, assign_parm_setup_block)
15533 (pad_below): Likewise.
15534 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
15535 * gimple-ssa-store-merging.c (rhs_valid_for_store_merging_p): Likewise.
15536 * ira.c (get_subreg_tracking_sizes): Likewise.
15537 * ira-build.c (ira_create_allocno_objects): Likewise.
15538 * ira-color.c (coalesced_pseudo_reg_slot_compare): Likewise.
15539 (ira_sort_regnos_for_alter_reg): Likewise.
15540 * ira-costs.c (record_operand_costs): Likewise.
15541 * lower-subreg.c (interesting_mode_p, simplify_gen_subreg_concatn)
15542 (resolve_simple_move): Likewise.
15543 * lra-constraints.c (get_reload_reg, operands_match_p): Likewise.
15544 (process_addr_reg, simplify_operand_subreg, curr_insn_transform)
15545 (lra_constraints): Likewise.
15546 (CONST_POOL_OK_P): Reject variable-sized modes.
15547 * lra-spills.c (slot, assign_mem_slot, pseudo_reg_slot_compare)
15548 (add_pseudo_to_slot, lra_spill): Likewise.
15549 * omp-low.c (omp_clause_aligned_alignment): Likewise.
15550 * optabs-query.c (get_best_extraction_insn): Likewise.
15551 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
15552 * optabs.c (expand_vec_perm_var, expand_vec_cond_expr): Likewise.
15553 (expand_mult_highpart, valid_multiword_target_p): Likewise.
15554 * recog.c (offsettable_address_addr_space_p): Likewise.
15555 * regcprop.c (maybe_mode_change): Likewise.
15556 * reginfo.c (choose_hard_reg_mode, record_subregs_of_mode): Likewise.
15557 * regrename.c (build_def_use): Likewise.
15558 * regstat.c (dump_reg_info): Likewise.
15559 * reload.c (complex_word_subreg_p, push_reload, find_dummy_reload)
15560 (find_reloads, find_reloads_subreg_address): Likewise.
15561 * reload1.c (eliminate_regs_1): Likewise.
15562 * rtlanal.c (for_each_inc_dec_find_inc_dec, rtx_cost): Likewise.
15563 * simplify-rtx.c (avoid_constant_pool_reference): Likewise.
15564 (simplify_binary_operation_1, simplify_subreg): Likewise.
15565 * targhooks.c (default_function_arg_padding): Likewise.
15566 (default_hard_regno_nregs, default_class_max_nregs): Likewise.
15567 * tree-cfg.c (verify_gimple_assign_binary): Likewise.
15568 (verify_gimple_assign_ternary): Likewise.
15569 * tree-inline.c (estimate_move_cost): Likewise.
15570 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15571 * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Likewise.
15572 (get_address_cost_ainc): Likewise.
15573 * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
15574 (vect_supportable_dr_alignment): Likewise.
15575 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15576 (vectorizable_reduction): Likewise.
15577 * tree-vect-stmts.c (vectorizable_assignment, vectorizable_shift)
15578 (vectorizable_operation, vectorizable_load): Likewise.
15579 * tree.c (build_same_sized_truth_vector_type): Likewise.
15580 * valtrack.c (cleanup_auto_inc_dec): Likewise.
15581 * var-tracking.c (emit_note_insn_var_location): Likewise.
15582 * config/arc/arc.h (ASM_OUTPUT_CASE_END): Use as_a <scalar_int_mode>.
15583 (ADDR_VEC_ALIGN): Likewise.
15585 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15586 Alan Hayward <alan.hayward@arm.com>
15587 David Sherwood <david.sherwood@arm.com>
15589 * machmode.h (mode_to_bits): Return a poly_uint16 rather than an
15591 (GET_MODE_BITSIZE): Return a constant if ONLY_FIXED_SIZE_MODES,
15592 or if measurement_type is polynomial.
15593 * calls.c (shift_return_value): Treat GET_MODE_BITSIZE as polynomial.
15594 * combine.c (make_extraction): Likewise.
15595 * dse.c (find_shift_sequence): Likewise.
15596 * dwarf2out.c (mem_loc_descriptor): Likewise.
15597 * expmed.c (store_integral_bit_field, extract_bit_field_1): Likewise.
15598 (extract_bit_field, extract_low_bits): Likewise.
15599 * expr.c (convert_move, convert_modes, emit_move_insn_1): Likewise.
15600 (optimize_bitfield_assignment_op, expand_assignment): Likewise.
15601 (store_expr_with_bounds, store_field, expand_expr_real_1): Likewise.
15602 * fold-const.c (optimize_bit_field_compare, merge_ranges): Likewise.
15603 * gimple-fold.c (optimize_atomic_compare_exchange_p): Likewise.
15604 * reload.c (find_reloads): Likewise.
15605 * reload1.c (alter_reg): Likewise.
15606 * stor-layout.c (bitwise_mode_for_mode, compute_record_mode): Likewise.
15607 * targhooks.c (default_secondary_memory_needed_mode): Likewise.
15608 * tree-if-conv.c (predicate_mem_writes): Likewise.
15609 * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
15610 * tree-vect-patterns.c (adjust_bool_pattern): Likewise.
15611 * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
15612 * valtrack.c (dead_debug_insert_temp): Likewise.
15613 * varasm.c (mergeable_constant_section): Likewise.
15614 * config/sh/sh.h (LOCAL_ALIGNMENT): Use as_a <fixed_size_mode>.
15616 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15617 Alan Hayward <alan.hayward@arm.com>
15618 David Sherwood <david.sherwood@arm.com>
15620 * expr.c (expand_assignment): Cope with polynomial mode sizes
15621 when assigning to a CONCAT.
15623 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15624 Alan Hayward <alan.hayward@arm.com>
15625 David Sherwood <david.sherwood@arm.com>
15627 * machmode.h (mode_precision): Change from unsigned short to
15629 (mode_to_precision): Return a poly_uint16 rather than an unsigned
15631 (GET_MODE_PRECISION): Return a constant if ONLY_FIXED_SIZE_MODES,
15632 or if measurement_type is not polynomial.
15633 (HWI_COMPUTABLE_MODE_P): Turn into a function. Optimize the case
15634 in which the mode is already known to be a scalar_int_mode.
15635 * genmodes.c (emit_mode_precision): Change the type of mode_precision
15636 from unsigned short to poly_uint16_pod. Use ZERO_COEFFS for the
15638 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15639 for GET_MODE_PRECISION.
15640 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15641 for GET_MODE_PRECISION.
15642 * combine.c (update_rsp_from_reg_equal): Treat GET_MODE_PRECISION
15644 (try_combine, find_split_point, combine_simplify_rtx): Likewise.
15645 (expand_field_assignment, make_extraction): Likewise.
15646 (make_compound_operation_int, record_dead_and_set_regs_1): Likewise.
15647 (get_last_value): Likewise.
15648 * convert.c (convert_to_integer_1): Likewise.
15649 * cse.c (cse_insn): Likewise.
15650 * expr.c (expand_expr_real_1): Likewise.
15651 * lra-constraints.c (simplify_operand_subreg): Likewise.
15652 * optabs-query.c (can_atomic_load_p): Likewise.
15653 * optabs.c (expand_atomic_load): Likewise.
15654 (expand_atomic_store): Likewise.
15655 * ree.c (combine_reaching_defs): Likewise.
15656 * rtl.h (partial_subreg_p, paradoxical_subreg_p): Likewise.
15657 * rtlanal.c (nonzero_bits1, lsb_bitfield_op_p): Likewise.
15658 * tree.h (type_has_mode_precision_p): Likewise.
15659 * ubsan.c (instrument_si_overflow): Likewise.
15661 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15662 Alan Hayward <alan.hayward@arm.com>
15663 David Sherwood <david.sherwood@arm.com>
15665 * tree.h (TYPE_VECTOR_SUBPARTS): Turn into a function and handle
15666 polynomial numbers of units.
15667 (SET_TYPE_VECTOR_SUBPARTS): Likewise.
15668 (valid_vector_subparts_p): New function.
15669 (build_vector_type): Remove temporary shim and take the number
15670 of units as a poly_uint64 rather than an int.
15671 (build_opaque_vector_type): Take the number of units as a
15672 poly_uint64 rather than an int.
15673 * tree.c (build_vector_from_ctor): Handle polynomial
15674 TYPE_VECTOR_SUBPARTS.
15675 (type_hash_canon_hash, type_cache_hasher::equal): Likewise.
15676 (uniform_vector_p, vector_type_mode, build_vector): Likewise.
15677 (build_vector_from_val): If the number of units is variable,
15678 use build_vec_duplicate_cst for constant operands and
15679 VEC_DUPLICATE_EXPR otherwise.
15680 (make_vector_type): Remove temporary is_constant ().
15681 (build_vector_type, build_opaque_vector_type): Take the number of
15682 units as a poly_uint64 rather than an int.
15683 (check_vector_cst): Handle polynomial TYPE_VECTOR_SUBPARTS and
15685 * cfgexpand.c (expand_debug_expr): Likewise.
15686 * expr.c (count_type_elements, categorize_ctor_elements_1): Likewise.
15687 (store_constructor, expand_expr_real_1): Likewise.
15688 (const_scalar_mask_from_tree): Likewise.
15689 * fold-const-call.c (fold_const_reduction): Likewise.
15690 * fold-const.c (const_binop, const_unop, fold_convert_const): Likewise.
15691 (operand_equal_p, fold_vec_perm, fold_ternary_loc): Likewise.
15692 (native_encode_vector, vec_cst_ctor_to_array): Likewise.
15693 (fold_relational_const): Likewise.
15694 (native_interpret_vector): Likewise. Change the size from an
15695 int to an unsigned int.
15696 * gimple-fold.c (gimple_fold_stmt_to_constant_1): Handle polynomial
15697 TYPE_VECTOR_SUBPARTS.
15698 (gimple_fold_indirect_ref, gimple_build_vector): Likewise.
15699 (gimple_build_vector_from_val): Use VEC_DUPLICATE_EXPR when
15700 duplicating a non-constant operand into a variable-length vector.
15701 * hsa-brig.c (hsa_op_immed::emit_to_buffer): Handle polynomial
15702 TYPE_VECTOR_SUBPARTS and VECTOR_CST_NELTS.
15703 * ipa-icf.c (sem_variable::equals): Likewise.
15704 * match.pd: Likewise.
15705 * omp-simd-clone.c (simd_clone_subparts): Likewise.
15706 * print-tree.c (print_node): Likewise.
15707 * stor-layout.c (layout_type): Likewise.
15708 * targhooks.c (default_builtin_vectorization_cost): Likewise.
15709 * tree-cfg.c (verify_gimple_comparison): Likewise.
15710 (verify_gimple_assign_binary): Likewise.
15711 (verify_gimple_assign_ternary): Likewise.
15712 (verify_gimple_assign_single): Likewise.
15713 * tree-pretty-print.c (dump_generic_node): Likewise.
15714 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15715 (simplify_bitfield_ref, is_combined_permutation_identity): Likewise.
15716 * tree-vect-data-refs.c (vect_permute_store_chain): Likewise.
15717 (vect_grouped_load_supported, vect_permute_load_chain): Likewise.
15718 (vect_shift_permute_load_chain): Likewise.
15719 * tree-vect-generic.c (nunits_for_known_piecewise_op): Likewise.
15720 (expand_vector_condition, optimize_vector_constructor): Likewise.
15721 (lower_vec_perm, get_compute_type): Likewise.
15722 * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
15723 (get_initial_defs_for_reduction, vect_transform_loop): Likewise.
15724 * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
15725 (vect_recog_mask_conversion_pattern): Likewise.
15726 * tree-vect-slp.c (vect_supported_load_permutation_p): Likewise.
15727 (vect_get_constant_vectors, vect_transform_slp_perm_load): Likewise.
15728 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
15729 (get_group_load_store_type, vectorizable_mask_load_store): Likewise.
15730 (vectorizable_bswap, simd_clone_subparts, vectorizable_assignment)
15731 (vectorizable_shift, vectorizable_operation, vectorizable_store)
15732 (vectorizable_load, vect_is_simple_cond, vectorizable_comparison)
15733 (supportable_widening_operation): Likewise.
15734 (supportable_narrowing_operation): Likewise.
15735 * tree-vector-builder.c (tree_vector_builder::binary_encoded_nelts):
15737 * varasm.c (output_constant): Likewise.
15739 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15740 Alan Hayward <alan.hayward@arm.com>
15741 David Sherwood <david.sherwood@arm.com>
15743 * tree-vect-data-refs.c (vect_permute_store_chain): Reorganize
15744 so that both the length == 3 and length != 3 cases set up their
15745 own permute vectors. Add comments explaining why we know the
15746 number of elements is constant.
15747 (vect_permute_load_chain): Likewise.
15749 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15750 Alan Hayward <alan.hayward@arm.com>
15751 David Sherwood <david.sherwood@arm.com>
15753 * machmode.h (mode_nunits): Change from unsigned char to
15755 (ONLY_FIXED_SIZE_MODES): New macro.
15756 (pod_mode::measurement_type, scalar_int_mode::measurement_type)
15757 (scalar_float_mode::measurement_type, scalar_mode::measurement_type)
15758 (complex_mode::measurement_type, fixed_size_mode::measurement_type):
15760 (mode_to_nunits): Return a poly_uint16 rather than an unsigned short.
15761 (GET_MODE_NUNITS): Return a constant if ONLY_FIXED_SIZE_MODES,
15762 or if measurement_type is not polynomial.
15763 * genmodes.c (ZERO_COEFFS): New macro.
15764 (emit_mode_nunits_inline): Make mode_nunits_inline return a
15766 (emit_mode_nunits): Change the type of mode_nunits to poly_uint16_pod.
15767 Use ZERO_COEFFS when emitting initializers.
15768 * data-streamer.h (bp_pack_poly_value): New function.
15769 (bp_unpack_poly_value): Likewise.
15770 * lto-streamer-in.c (lto_input_mode_table): Use bp_unpack_poly_value
15771 for GET_MODE_NUNITS.
15772 * lto-streamer-out.c (lto_write_mode_table): Use bp_pack_poly_value
15773 for GET_MODE_NUNITS.
15774 * tree.c (make_vector_type): Remove temporary shim and make
15775 the real function take the number of units as a poly_uint64
15776 rather than an int.
15777 (build_vector_type_for_mode): Handle polynomial nunits.
15778 * dwarf2out.c (loc_descriptor, add_const_value_attribute): Likewise.
15779 * emit-rtl.c (const_vec_series_p_1): Likewise.
15780 (gen_rtx_CONST_VECTOR): Likewise.
15781 * fold-const.c (test_vec_duplicate_folding): Likewise.
15782 * genrecog.c (validate_pattern): Likewise.
15783 * optabs-query.c (can_vec_perm_var_p, can_mult_highpart_p): Likewise.
15784 * optabs-tree.c (expand_vec_cond_expr_p): Likewise.
15785 * optabs.c (expand_vector_broadcast, expand_binop_directly): Likewise.
15786 (shift_amt_for_vec_perm_mask, expand_vec_perm_var): Likewise.
15787 (expand_vec_cond_expr, expand_mult_highpart): Likewise.
15788 * rtlanal.c (subreg_get_info): Likewise.
15789 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15790 (vect_grouped_load_supported): Likewise.
15791 * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
15792 * tree-vect-loop.c (have_whole_vector_shift): Likewise.
15793 * simplify-rtx.c (simplify_unary_operation_1): Likewise.
15794 (simplify_const_unary_operation, simplify_binary_operation_1)
15795 (simplify_const_binary_operation, simplify_ternary_operation)
15796 (test_vector_ops_duplicate, test_vector_ops): Likewise.
15797 (simplify_immed_subreg): Use GET_MODE_NUNITS on a fixed_size_mode
15798 instead of CONST_VECTOR_NUNITS.
15799 * varasm.c (output_constant_pool_2): Likewise.
15800 * rtx-vector-builder.c (rtx_vector_builder::build): Only include the
15801 explicit-encoded elements in the XVEC for variable-length vectors.
15803 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15805 * lra-constraints.c (curr_insn_transform): Use partial_subreg_p.
15807 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15808 Alan Hayward <alan.hayward@arm.com>
15809 David Sherwood <david.sherwood@arm.com>
15811 * coretypes.h (fixed_size_mode): Declare.
15812 (fixed_size_mode_pod): New typedef.
15813 * builtins.h (target_builtins::x_apply_args_mode)
15814 (target_builtins::x_apply_result_mode): Change type to
15815 fixed_size_mode_pod.
15816 * builtins.c (apply_args_size, apply_result_size, result_vector)
15817 (expand_builtin_apply_args_1, expand_builtin_apply)
15818 (expand_builtin_return): Update accordingly.
15820 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15822 * cse.c (hash_rtx_cb): Hash only the encoded elements.
15823 * cselib.c (cselib_hash_rtx): Likewise.
15824 * expmed.c (make_tree): Build VECTOR_CSTs directly from the
15825 CONST_VECTOR encoding.
15827 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15828 Jeff Law <law@redhat.com>
15831 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): For
15832 noreturn probe, use gen_pop instead of ix86_emit_restore_reg_using_pop,
15833 only set RTX_FRAME_RELATED_P on both the push and pop if cfa_reg is sp
15834 and add REG_CFA_ADJUST_CFA notes in that case to both insns.
15837 * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Do not
15838 explicitly probe *sp in a noreturn function if there were any callee
15839 register saves or frame pointer is needed.
15841 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15844 * cfgexpand.c (expand_debug_expr): Return NULL if mode is
15845 BLKmode for ternary, binary or unary expressions.
15848 * var-tracking.c (delete_vta_debug_insn): New inline function.
15849 (delete_vta_debug_insns): Add USE_CFG argument, if true, walk just
15850 insns from get_insns () to NULL instead of each bb separately.
15851 Use delete_vta_debug_insn. No longer static.
15852 (vt_debug_insns_local, variable_tracking_main_1): Adjust
15853 delete_vta_debug_insns callers.
15854 * rtl.h (delete_vta_debug_insns): Declare.
15855 * final.c (rest_of_handle_final): Call delete_vta_debug_insns
15856 instead of variable_tracking_main.
15858 2018-01-03 Martin Sebor <msebor@redhat.com>
15860 PR tree-optimization/83603
15861 * calls.c (maybe_warn_nonstring_arg): Avoid accessing function
15862 arguments past the endof the argument list in functions declared
15863 without a prototype.
15864 * gimple-ssa-warn-restrict.c (wrestrict_dom_walker::check_call):
15865 Avoid checking when arguments are null.
15867 2018-01-03 Martin Sebor <msebor@redhat.com>
15870 * doc/extend.texi (attribute const): Fix a typo.
15871 * ipa-pure-const.c ((warn_function_const, warn_function_pure): Avoid
15872 issuing -Wsuggest-attribute for void functions.
15874 2018-01-03 Martin Sebor <msebor@redhat.com>
15876 * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref): Use
15877 offset_int::from instead of wide_int::to_shwi.
15878 (maybe_diag_overlap): Remove assertion.
15879 Use HOST_WIDE_INT_PRINT_DEC instead of %lli.
15880 * gimple-ssa-sprintf.c (format_directive): Same.
15881 (parse_directive): Same.
15882 (sprintf_dom_walker::compute_format_length): Same.
15883 (try_substitute_return_value): Same.
15885 2018-01-03 Jeff Law <law@redhat.com>
15887 PR middle-end/83654
15888 * explow.c (anti_adjust_stack_and_probe_stack_clash): Test a
15889 non-constant residual for zero at runtime and avoid probing in
15890 that case. Reorganize code for trailing problem to mirror handling
15893 2018-01-03 Prathamesh Kulkarni <prathamesh.kulkarni@linaro.org>
15895 PR tree-optimization/83501
15896 * tree-ssa-strlen.c (get_string_cst): New.
15897 (handle_char_store): Call get_string_cst.
15899 2018-01-03 Martin Liska <mliska@suse.cz>
15901 PR tree-optimization/83593
15902 * tree-ssa-strlen.c: Include tree-cfg.h.
15903 (strlen_check_and_optimize_stmt): Add new argument cleanup_eh.
15904 (strlen_dom_walker): Add new member variable m_cleanup_cfg.
15905 (strlen_dom_walker::strlen_dom_walker): Initialize m_cleanup_cfg
15907 (strlen_dom_walker::before_dom_children): Call
15908 gimple_purge_dead_eh_edges. Dump tranformation with details
15910 (strlen_dom_walker::before_dom_children): Update call by adding
15911 new argument cleanup_eh.
15912 (pass_strlen::execute): Return TODO_cleanup_cfg if needed.
15914 2018-01-03 Martin Liska <mliska@suse.cz>
15917 * cif-code.def (VARIADIC_THUNK): New enum value.
15918 * ipa-fnsummary.c (compute_fn_summary): Do not inline variadic
15921 2018-01-03 Jan Beulich <jbeulich@suse.com>
15923 * sse.md (mov<mode>_internal): Tighten condition for when to use
15924 vmovdqu<ssescalarsize> for TI and OI modes.
15926 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15928 Update copyright years.
15930 2018-01-03 Martin Liska <mliska@suse.cz>
15933 * ipa-visibility.c (function_and_variable_visibility): Skip
15934 functions with noipa attribure.
15936 2018-01-03 Jakub Jelinek <jakub@redhat.com>
15938 * gcc.c (process_command): Update copyright notice dates.
15939 * gcov-dump.c (print_version): Ditto.
15940 * gcov.c (print_version): Ditto.
15941 * gcov-tool.c (print_version): Ditto.
15942 * gengtype.c (create_file): Ditto.
15943 * doc/cpp.texi: Bump @copying's copyright year.
15944 * doc/cppinternals.texi: Ditto.
15945 * doc/gcc.texi: Ditto.
15946 * doc/gccint.texi: Ditto.
15947 * doc/gcov.texi: Ditto.
15948 * doc/install.texi: Ditto.
15949 * doc/invoke.texi: Ditto.
15951 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15953 * vector-builder.h (vector_builder::m_full_nelts): Change from
15954 unsigned int to poly_uint64.
15955 (vector_builder::full_nelts): Update prototype accordingly.
15956 (vector_builder::new_vector): Likewise.
15957 (vector_builder::encoded_full_vector_p): Handle polynomial full_nelts.
15958 (vector_builder::operator ==): Likewise.
15959 (vector_builder::finalize): Likewise.
15960 * int-vector-builder.h (int_vector_builder::int_vector_builder):
15961 Take the number of elements as a poly_uint64 rather than an
15963 * vec-perm-indices.h (vec_perm_indices::m_nelts_per_input): Change
15964 from unsigned int to poly_uint64.
15965 (vec_perm_indices::vec_perm_indices): Update prototype accordingly.
15966 (vec_perm_indices::new_vector): Likewise.
15967 (vec_perm_indices::length): Likewise.
15968 (vec_perm_indices::nelts_per_input): Likewise.
15969 (vec_perm_indices::input_nelts): Likewise.
15970 * vec-perm-indices.c (vec_perm_indices::new_vector): Take the
15971 number of elements per input as a poly_uint64 rather than an
15972 unsigned int. Use the original encoding for variable-length
15973 vectors, rather than clamping each individual element.
15974 For the second and subsequent elements in each pattern,
15975 clamp the step and base before clamping their sum.
15976 (vec_perm_indices::series_p): Handle polynomial element counts.
15977 (vec_perm_indices::all_in_range_p): Likewise.
15978 (vec_perm_indices_to_tree): Likewise.
15979 (vec_perm_indices_to_rtx): Likewise.
15980 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
15981 * tree-vector-builder.c (tree_vector_builder::new_unary_operation)
15982 (tree_vector_builder::new_binary_operation): Handle polynomial
15983 element counts. Return false if we need to know the number
15984 of elements at compile time.
15985 * fold-const.c (fold_vec_perm): Punt if the number of elements
15986 isn't known at compile time.
15988 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
15990 * vec-perm-indices.h (vec_perm_builder): Change element type
15991 from HOST_WIDE_INT to poly_int64.
15992 (vec_perm_indices::element_type): Update accordingly.
15993 (vec_perm_indices::clamp): Handle polynomial element_types.
15994 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
15995 (vec_perm_indices::all_in_range_p): Likewise.
15996 (tree_to_vec_perm_builder): Check for poly_int64 trees rather
15998 * vector-builder.h (vector_builder::stepped_sequence_p): Handle
15999 polynomial vec_perm_indices element types.
16000 * int-vector-builder.h (int_vector_builder::equal_p): Likewise.
16001 * fold-const.c (fold_vec_perm): Likewise.
16002 * optabs.c (shift_amt_for_vec_perm_mask): Likewise.
16003 * tree-vect-generic.c (lower_vec_perm): Likewise.
16004 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
16005 * config/aarch64/aarch64.c (aarch64_evpc_tbl): Cast d->perm
16006 element type to HOST_WIDE_INT.
16008 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16009 Alan Hayward <alan.hayward@arm.com>
16010 David Sherwood <david.sherwood@arm.com>
16012 * alias.c (addr_side_effect_eval): Take the size as a poly_int64
16013 rather than an int. Use plus_constant.
16014 (memrefs_conflict_p): Take the sizes as poly_int64s rather than ints.
16015 Take the offset "c" as a poly_int64 rather than a HOST_WIDE_INT.
16017 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16018 Alan Hayward <alan.hayward@arm.com>
16019 David Sherwood <david.sherwood@arm.com>
16021 * calls.c (emit_call_1, expand_call): Change struct_value_size from
16022 a HOST_WIDE_INT to a poly_int64.
16024 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16025 Alan Hayward <alan.hayward@arm.com>
16026 David Sherwood <david.sherwood@arm.com>
16028 * calls.c (load_register_parameters): Cope with polynomial
16029 mode sizes. Require a constant size for BLKmode parameters
16030 that aren't described by a PARALLEL. If BLOCK_REG_PADDING
16031 forces a parameter to be padded at the lsb end in order to
16032 fill a complete number of words, require the parameter size
16033 to be ordered wrt UNITS_PER_WORD.
16035 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16036 Alan Hayward <alan.hayward@arm.com>
16037 David Sherwood <david.sherwood@arm.com>
16039 * reload1.c (spill_stack_slot_width): Change element type
16040 from unsigned int to poly_uint64_pod.
16041 (alter_reg): Treat mode sizes as polynomial.
16043 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16044 Alan Hayward <alan.hayward@arm.com>
16045 David Sherwood <david.sherwood@arm.com>
16047 * reload.c (complex_word_subreg_p): New function.
16048 (reload_inner_reg_of_subreg, push_reload): Use it.
16050 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16051 Alan Hayward <alan.hayward@arm.com>
16052 David Sherwood <david.sherwood@arm.com>
16054 * lra-constraints.c (process_alt_operands): Reject matched
16055 operands whose sizes aren't ordered.
16056 (match_reload): Refer to this check here.
16058 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16059 Alan Hayward <alan.hayward@arm.com>
16060 David Sherwood <david.sherwood@arm.com>
16062 * builtins.c (expand_ifn_atomic_compare_exchange_into_call): Assert
16063 that the mode size is in the set {1, 2, 4, 8, 16}.
16065 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16066 Alan Hayward <alan.hayward@arm.com>
16067 David Sherwood <david.sherwood@arm.com>
16069 * var-tracking.c (adjust_mems): Treat mode sizes as polynomial.
16070 Use plus_constant instead of gen_rtx_PLUS.
16072 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16073 Alan Hayward <alan.hayward@arm.com>
16074 David Sherwood <david.sherwood@arm.com>
16076 * config/cr16/cr16-protos.h (cr16_push_rounding): Declare.
16077 * config/cr16/cr16.h (PUSH_ROUNDING): Move implementation to...
16078 * config/cr16/cr16.c (cr16_push_rounding): ...this new function.
16079 * config/h8300/h8300-protos.h (h8300_push_rounding): Declare.
16080 * config/h8300/h8300.h (PUSH_ROUNDING): Move implementation to...
16081 * config/h8300/h8300.c (h8300_push_rounding): ...this new function.
16082 * config/i386/i386-protos.h (ix86_push_rounding): Declare.
16083 * config/i386/i386.h (PUSH_ROUNDING): Move implementation to...
16084 * config/i386/i386.c (ix86_push_rounding): ...this new function.
16085 * config/m32c/m32c-protos.h (m32c_push_rounding): Take and return
16087 * config/m32c/m32c.c (m32c_push_rounding): Likewise.
16088 * config/m68k/m68k-protos.h (m68k_push_rounding): Declare.
16089 * config/m68k/m68k.h (PUSH_ROUNDING): Move implementation to...
16090 * config/m68k/m68k.c (m68k_push_rounding): ...this new function.
16091 * config/pdp11/pdp11-protos.h (pdp11_push_rounding): Declare.
16092 * config/pdp11/pdp11.h (PUSH_ROUNDING): Move implementation to...
16093 * config/pdp11/pdp11.c (pdp11_push_rounding): ...this new function.
16094 * config/stormy16/stormy16-protos.h (xstormy16_push_rounding): Declare.
16095 * config/stormy16/stormy16.h (PUSH_ROUNDING): Move implementation to...
16096 * config/stormy16/stormy16.c (xstormy16_push_rounding): ...this new
16098 * expr.c (emit_move_resolve_push): Treat the input and result
16099 of PUSH_ROUNDING as a poly_int64.
16100 (emit_move_complex_push, emit_single_push_insn_1): Likewise.
16101 (emit_push_insn): Likewise.
16102 * lra-eliminations.c (mark_not_eliminable): Likewise.
16103 * recog.c (push_operand): Likewise.
16104 * reload1.c (elimination_effects): Likewise.
16105 * rtlanal.c (nonzero_bits1): Likewise.
16106 * calls.c (store_one_arg): Likewise. Require the padding to be
16107 known at compile time.
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 * expr.c (emit_single_push_insn_1): Treat mode sizes as polynomial.
16114 Use plus_constant instead of gen_rtx_PLUS.
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 * auto-inc-dec.c (set_inc_state): Take the mode size as a poly_int64
16121 rather than an int.
16123 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16124 Alan Hayward <alan.hayward@arm.com>
16125 David Sherwood <david.sherwood@arm.com>
16127 * expr.c (expand_expr_real_1): Use tree_to_poly_uint64
16128 instead of int_size_in_bytes when handling VIEW_CONVERT_EXPRs
16129 via stack temporaries. Treat the mode size as polynomial too.
16131 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16132 Alan Hayward <alan.hayward@arm.com>
16133 David Sherwood <david.sherwood@arm.com>
16135 * expr.c (expand_expr_real_2): When handling conversions involving
16136 unions, apply tree_to_poly_uint64 to the TYPE_SIZE rather than
16137 multiplying int_size_in_bytes by BITS_PER_UNIT. Treat GET_MODE_BISIZE
16138 as a poly_uint64 too.
16140 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16141 Alan Hayward <alan.hayward@arm.com>
16142 David Sherwood <david.sherwood@arm.com>
16144 * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
16146 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16147 Alan Hayward <alan.hayward@arm.com>
16148 David Sherwood <david.sherwood@arm.com>
16150 * combine.c (can_change_dest_mode): Handle polynomial
16151 REGMODE_NATURAL_SIZE.
16152 * expmed.c (store_bit_field_1): Likewise.
16153 * expr.c (store_constructor): Likewise.
16154 * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
16155 and polynomial REGMODE_NATURAL_SIZE.
16156 (gen_lowpart_common): Likewise.
16157 * reginfo.c (record_subregs_of_mode): Likewise.
16158 * rtlanal.c (read_modify_subreg_p): Likewise.
16160 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16161 Alan Hayward <alan.hayward@arm.com>
16162 David Sherwood <david.sherwood@arm.com>
16164 * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
16165 numbers of elements.
16167 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16168 Alan Hayward <alan.hayward@arm.com>
16169 David Sherwood <david.sherwood@arm.com>
16171 * match.pd: Cope with polynomial numbers of vector elements.
16173 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16174 Alan Hayward <alan.hayward@arm.com>
16175 David Sherwood <david.sherwood@arm.com>
16177 * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
16178 in a POINTER_PLUS_EXPR.
16180 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16181 Alan Hayward <alan.hayward@arm.com>
16182 David Sherwood <david.sherwood@arm.com>
16184 * omp-simd-clone.c (simd_clone_subparts): New function.
16185 (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
16186 (ipa_simd_modify_function_body): Likewise.
16188 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16189 Alan Hayward <alan.hayward@arm.com>
16190 David Sherwood <david.sherwood@arm.com>
16192 * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
16193 (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
16194 (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
16195 (expand_vector_condition, vector_element): Likewise.
16196 (subparts_gt): New function.
16197 (get_compute_type): Use subparts_gt.
16198 (count_type_subparts): Delete.
16199 (expand_vector_operations_1): Use subparts_gt instead of
16200 count_type_subparts.
16202 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16203 Alan Hayward <alan.hayward@arm.com>
16204 David Sherwood <david.sherwood@arm.com>
16206 * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
16207 (vect_compile_time_alias): ...this new function. Do the calculation
16208 on poly_ints rather than trees.
16209 (vect_prune_runtime_alias_test_list): Update call accordingly.
16211 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16212 Alan Hayward <alan.hayward@arm.com>
16213 David Sherwood <david.sherwood@arm.com>
16215 * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
16217 (vect_schedule_slp_instance): Likewise.
16219 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16220 Alan Hayward <alan.hayward@arm.com>
16221 David Sherwood <david.sherwood@arm.com>
16223 * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
16224 constant and extern definitions for variable-length vectors.
16225 (vect_get_constant_vectors): Note that the number of units
16226 is known to be constant.
16228 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16229 Alan Hayward <alan.hayward@arm.com>
16230 David Sherwood <david.sherwood@arm.com>
16232 * tree-vect-stmts.c (vectorizable_conversion): Treat the number
16233 of units as polynomial. Choose between WIDE and NARROW based
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-vect-stmts.c (simd_clone_subparts): New function.
16241 (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
16243 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16244 Alan Hayward <alan.hayward@arm.com>
16245 David Sherwood <david.sherwood@arm.com>
16247 * tree-vect-stmts.c (vectorizable_call): Treat the number of
16248 vectors as polynomial. Use build_index_vector for
16249 IFN_GOMP_SIMD_LANE.
16251 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16252 Alan Hayward <alan.hayward@arm.com>
16253 David Sherwood <david.sherwood@arm.com>
16255 * tree-vect-stmts.c (get_load_store_type): Treat the number of
16256 units as polynomial. Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
16257 for variable-length vectors.
16258 (vectorizable_mask_load_store): Treat the number of units as
16259 polynomial, asserting that it is constant if the condition has
16260 already been enforced.
16261 (vectorizable_store, vectorizable_load): Likewise.
16263 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16264 Alan Hayward <alan.hayward@arm.com>
16265 David Sherwood <david.sherwood@arm.com>
16267 * tree-vect-loop.c (vectorizable_live_operation): Treat the number
16268 of units as polynomial. Punt if we can't tell at compile time
16269 which vector contains the final result.
16271 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16272 Alan Hayward <alan.hayward@arm.com>
16273 David Sherwood <david.sherwood@arm.com>
16275 * tree-vect-loop.c (vectorizable_induction): Treat the number
16276 of units as polynomial. Punt on SLP inductions. Use an integer
16277 VEC_SERIES_EXPR for variable-length integer reductions. Use a
16278 cast of such a series for variable-length floating-point
16281 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16282 Alan Hayward <alan.hayward@arm.com>
16283 David Sherwood <david.sherwood@arm.com>
16285 * tree.h (build_index_vector): Declare.
16286 * tree.c (build_index_vector): New function.
16287 * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
16288 of units as polynomial, forcibly converting it to a constant if
16289 vectorizable_reduction has already enforced the condition.
16290 (vect_create_epilog_for_reduction): Likewise. Use build_index_vector
16291 to create a {1,2,3,...} vector.
16292 (vectorizable_reduction): Treat the number of units as polynomial.
16293 Choose vectype_in based on the largest scalar element size rather
16294 than the smallest number of units. Enforce the restrictions
16297 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16298 Alan Hayward <alan.hayward@arm.com>
16299 David Sherwood <david.sherwood@arm.com>
16301 * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
16302 number of units as polynomial.
16304 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16305 Alan Hayward <alan.hayward@arm.com>
16306 David Sherwood <david.sherwood@arm.com>
16308 * target.h (vector_sizes, auto_vector_sizes): New typedefs.
16309 * target.def (autovectorize_vector_sizes): Return the vector sizes
16310 by pointer, using vector_sizes rather than a bitmask.
16311 * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
16312 * targhooks.c (default_autovectorize_vector_sizes): Likewise.
16313 * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
16315 * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
16316 * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
16317 * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
16318 * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
16319 * omp-general.c (omp_max_vf): Likewise.
16320 * omp-low.c (omp_clause_aligned_alignment): Likewise.
16321 * optabs-query.c (can_vec_mask_load_store_p): Likewise.
16322 * tree-vect-loop.c (vect_analyze_loop): Likewise.
16323 * tree-vect-slp.c (vect_slp_bb): Likewise.
16324 * doc/tm.texi: Regenerate.
16325 * tree-vectorizer.h (current_vector_size): Change from an unsigned int
16327 * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
16328 the vector size as a poly_uint64 rather than an unsigned int.
16329 (current_vector_size): Change from an unsigned int to a poly_uint64.
16330 (get_vectype_for_scalar_type): Update accordingly.
16331 * tree.h (build_truth_vector_type): Take the size and number of
16332 units as a poly_uint64 rather than an unsigned int.
16333 (build_vector_type): Add a temporary overload that takes
16334 the number of units as a poly_uint64 rather than an unsigned int.
16335 * tree.c (make_vector_type): Likewise.
16336 (build_truth_vector_type): Take the number of units as a poly_uint64
16337 rather than an unsigned int.
16339 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16340 Alan Hayward <alan.hayward@arm.com>
16341 David Sherwood <david.sherwood@arm.com>
16343 * target.def (get_mask_mode): Take the number of units and length
16344 as poly_uint64s rather than unsigned ints.
16345 * targhooks.h (default_get_mask_mode): Update accordingly.
16346 * targhooks.c (default_get_mask_mode): Likewise.
16347 * config/i386/i386.c (ix86_get_mask_mode): Likewise.
16348 * doc/tm.texi: Regenerate.
16350 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16351 Alan Hayward <alan.hayward@arm.com>
16352 David Sherwood <david.sherwood@arm.com>
16354 * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
16355 * omp-general.c (omp_max_vf): Likewise.
16356 * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
16357 (expand_omp_simd): Handle polynomial safelen.
16358 * omp-low.c (omplow_simd_context): Add a default constructor.
16359 (omplow_simd_context::max_vf): Change from int to poly_uint64.
16360 (lower_rec_simd_input_clauses): Update accordingly.
16361 (lower_rec_input_clauses): Likewise.
16363 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16364 Alan Hayward <alan.hayward@arm.com>
16365 David Sherwood <david.sherwood@arm.com>
16367 * tree-vectorizer.h (vect_nunits_for_cost): New function.
16368 * tree-vect-loop.c (vect_model_reduction_cost): Use it.
16369 * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
16370 (vect_analyze_slp_cost): Likewise.
16371 * tree-vect-stmts.c (vect_model_store_cost): Likewise.
16372 (vect_model_load_cost): Likewise.
16374 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16375 Alan Hayward <alan.hayward@arm.com>
16376 David Sherwood <david.sherwood@arm.com>
16378 * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
16379 (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
16380 from an unsigned int * to a poly_uint64_pod *.
16381 (calculate_unrolling_factor): New function.
16382 (vect_analyze_slp_instance): Use it. Track polynomial max_nunits.
16384 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16385 Alan Hayward <alan.hayward@arm.com>
16386 David Sherwood <david.sherwood@arm.com>
16388 * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
16389 from an unsigned int to a poly_uint64.
16390 (_loop_vec_info::slp_unrolling_factor): Likewise.
16391 (_loop_vec_info::vectorization_factor): Change from an int
16393 (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
16394 (vect_get_num_vectors): New function.
16395 (vect_update_max_nunits, vect_vf_for_cost): Likewise.
16396 (vect_get_num_copies): Use vect_get_num_vectors.
16397 (vect_analyze_data_ref_dependences): Change max_vf from an int *
16398 to an unsigned int *.
16399 (vect_analyze_data_refs): Change min_vf from an int * to a
16401 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
16402 than an unsigned HOST_WIDE_INT.
16403 * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
16404 (vect_analyze_data_ref_dependence): Change max_vf from an int *
16405 to an unsigned int *.
16406 (vect_analyze_data_ref_dependences): Likewise.
16407 (vect_compute_data_ref_alignment): Handle polynomial vf.
16408 (vect_enhance_data_refs_alignment): Likewise.
16409 (vect_prune_runtime_alias_test_list): Likewise.
16410 (vect_shift_permute_load_chain): Likewise.
16411 (vect_supportable_dr_alignment): Likewise.
16412 (dependence_distance_ge_vf): Take the vectorization factor as a
16413 poly_uint64 rather than an unsigned HOST_WIDE_INT.
16414 (vect_analyze_data_refs): Change min_vf from an int * to a
16416 * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
16417 vfm1 as a poly_uint64 rather than an int. Make the same change
16418 for the returned bound_scalar.
16419 (vect_gen_vector_loop_niters): Handle polynomial vf.
16420 (vect_do_peeling): Likewise. Update call to
16421 vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
16422 (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
16424 * tree-vect-loop.c (vect_determine_vectorization_factor)
16425 (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
16426 (vect_get_known_peeling_cost): Likewise.
16427 (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
16428 (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
16429 (vect_transform_loop): Likewise. Use the lowest possible VF when
16430 updating the upper bounds of the loop.
16431 (vect_min_worthwhile_factor): Make static. Return an unsigned int
16432 rather than an int.
16433 * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
16434 polynomial unroll factors.
16435 (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
16436 (vect_make_slp_decision): Likewise.
16437 (vect_supported_load_permutation_p): Likewise, and polynomial
16439 (vect_analyze_slp_cost): Handle polynomial vf.
16440 (vect_slp_analyze_node_operations): Likewise.
16441 (vect_slp_analyze_bb_1): Likewise.
16442 (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
16443 than an unsigned HOST_WIDE_INT.
16444 * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
16445 (vectorizable_load): Handle polynomial vf.
16446 * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
16448 (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
16450 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16451 Alan Hayward <alan.hayward@arm.com>
16452 David Sherwood <david.sherwood@arm.com>
16454 * match.pd: Handle bit operations involving three constants
16455 and try to fold one pair.
16457 2018-01-03 Richard Sandiford <richard.sandiford@linaro.org>
16459 * tree-vect-loop-manip.c: Include gimple-fold.h.
16460 (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
16461 niters_maybe_zero parameters. Handle other cases besides a step of 1.
16462 (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
16463 Add a path that uses a step of VF instead of 1, but disable it
16465 (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
16466 and niters_no_overflow parameters. Update calls to
16467 slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
16468 Create a new SSA name if the latter choses to use a ste other
16469 than zero, and return it via niters_vector_mult_vf_var.
16470 * tree-vect-loop.c (vect_transform_loop): Update calls to
16471 vect_do_peeling, vect_gen_vector_loop_niters and
16472 slpeel_make_loop_iterate_ntimes.
16473 * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
16474 (vect_gen_vector_loop_niters): Update declarations after above changes.
16476 2018-01-02 Michael Meissner <meissner@linux.vnet.ibm.com>
16478 * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
16479 128-bit round to integer instructions.
16480 (ceil<mode>2): Likewise.
16481 (btrunc<mode>2): Likewise.
16482 (round<mode>2): Likewise.
16484 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16486 * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
16487 unaligned VSX load/store on P8/P9.
16488 (expand_block_clear): Allow the use of unaligned VSX
16489 load/store on P8/P9.
16491 2018-01-02 Bill Schmidt <wschmidt@linux.vnet.ibm.com>
16493 * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
16495 (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
16496 swap associated with both a load and a store.
16498 2018-01-02 Andrew Waterman <andrew@sifive.com>
16500 * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
16501 * config/riscv/riscv.md (clear_cache): Use it.
16503 2018-01-02 Artyom Skrobov <tyomitch@gmail.com>
16505 * web.c: Remove out-of-date comment.
16507 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16509 * expr.c (fixup_args_size_notes): Check that any existing
16510 REG_ARGS_SIZE notes are correct, and don't try to re-add them.
16511 (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
16512 (emit_single_push_insn): ...here.
16514 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16516 * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
16517 (const_vector_encoded_nelts): New function.
16518 (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
16519 (const_vector_int_elt, const_vector_elt): Declare.
16520 * emit-rtl.c (const_vector_int_elt_1): New function.
16521 (const_vector_elt): Likewise.
16522 * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
16523 of CONST_VECTOR_ELT.
16525 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16527 * expr.c: Include rtx-vector-builder.h.
16528 (const_vector_mask_from_tree): Use rtx_vector_builder and operate
16529 directly on the tree encoding.
16530 (const_vector_from_tree): Likewise.
16531 * optabs.c: Include rtx-vector-builder.h.
16532 (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
16533 sequence of "u" values.
16534 * vec-perm-indices.c: Include rtx-vector-builder.h.
16535 (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
16536 directly on the vec_perm_indices encoding.
16538 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16540 * doc/rtl.texi (const_vector): Describe new encoding scheme.
16541 * Makefile.in (OBJS): Add rtx-vector-builder.o.
16542 * rtx-vector-builder.h: New file.
16543 * rtx-vector-builder.c: Likewise.
16544 * rtl.h (rtx_def::u2): Add a const_vector field.
16545 (CONST_VECTOR_NPATTERNS): New macro.
16546 (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
16547 (CONST_VECTOR_DUPLICATE_P): Likewise.
16548 (CONST_VECTOR_STEPPED_P): Likewise.
16549 (CONST_VECTOR_ENCODED_ELT): Likewise.
16550 (const_vec_duplicate_p): Check for a duplicated vector encoding.
16551 (unwrap_const_vec_duplicate): Likewise.
16552 (const_vec_series_p): Check for a non-duplicated vector encoding.
16553 Say that the function only returns true for integer vectors.
16554 * emit-rtl.c: Include rtx-vector-builder.h.
16555 (gen_const_vec_duplicate_1): Delete.
16556 (gen_const_vector): Call gen_const_vec_duplicate instead of
16557 gen_const_vec_duplicate_1.
16558 (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
16559 (gen_const_vec_duplicate): Use rtx_vector_builder.
16560 (gen_const_vec_series): Likewise.
16561 (gen_rtx_CONST_VECTOR): Likewise.
16562 * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
16563 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
16564 Build a new vector rather than modifying a CONST_VECTOR in-place.
16565 (handle_special_swappables): Update call accordingly.
16566 * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
16567 (swap_const_vector_halves): Take an rtx pointer rather than rtx.
16568 Build a new vector rather than modifying a CONST_VECTOR in-place.
16569 (handle_special_swappables): Update call accordingly.
16571 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16573 * simplify-rtx.c (simplify_const_binary_operation): Use
16574 CONST_VECTOR_ELT instead of XVECEXP.
16576 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16578 * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
16579 the selector elements to be different from the data elements
16580 if the selector is a VECTOR_CST.
16581 * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
16582 ssizetype for the selector.
16584 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16586 * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
16587 before testing each element individually.
16588 * tree-vect-generic.c (lower_vec_perm): Likewise.
16590 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16592 * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
16593 * selftest-run-tests.c (selftest::run_tests): Call it.
16594 * vector-builder.h (vector_builder::operator ==): New function.
16595 (vector_builder::operator !=): Likewise.
16596 * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
16597 (vec_perm_indices::all_from_input_p): New function.
16598 * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
16599 (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
16600 * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
16601 instead of reading the VECTOR_CST directly. Detect whether both
16602 vector inputs are the same before constructing the vec_perm_indices,
16603 and update the number of inputs argument accordingly. Use the
16604 utility functions added above. Only construct sel2 if we need to.
16606 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16608 * optabs.c (expand_vec_perm_var): Use an explicit encoding for
16609 the broadcast of the low byte.
16610 (expand_mult_highpart): Use an explicit encoding for the permutes.
16611 * optabs-query.c (can_mult_highpart_p): Likewise.
16612 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
16613 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16614 (vectorizable_bswap): Likewise.
16615 * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
16616 explicit encoding for the power-of-2 permutes.
16617 (vect_permute_store_chain): Likewise.
16618 (vect_grouped_load_supported): Likewise.
16619 (vect_permute_load_chain): Likewise.
16621 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16623 * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
16624 * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
16625 * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
16626 * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
16627 * tree-vect-stmts.c (vectorizable_bswap): Likewise.
16628 (vect_gen_perm_mask_any): Likewise.
16630 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16632 * int-vector-builder.h: New file.
16633 * vec-perm-indices.h: Include int-vector-builder.h.
16634 (vec_perm_indices): Redefine as an int_vector_builder.
16635 (auto_vec_perm_indices): Delete.
16636 (vec_perm_builder): Redefine as a stand-alone class.
16637 (vec_perm_indices::vec_perm_indices): New function.
16638 (vec_perm_indices::clamp): Likewise.
16639 * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
16640 (vec_perm_indices::new_vector): New function.
16641 (vec_perm_indices::new_expanded_vector): Update for new
16642 vec_perm_indices class.
16643 (vec_perm_indices::rotate_inputs): New function.
16644 (vec_perm_indices::all_in_range_p): Operate directly on the
16645 encoded form, without computing elided elements.
16646 (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
16647 encoding. Update for new vec_perm_indices class.
16648 * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
16649 the given vec_perm_builder.
16650 (expand_vec_perm_var): Update vec_perm_builder constructor.
16651 (expand_mult_highpart): Use vec_perm_builder instead of
16652 auto_vec_perm_indices.
16653 * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
16654 vec_perm_indices instead of auto_vec_perm_indices. Use a single
16655 or double series encoding as appropriate.
16656 * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
16657 vec_perm_indices instead of auto_vec_perm_indices.
16658 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16659 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16660 (vect_permute_store_chain): Likewise.
16661 (vect_grouped_load_supported): Likewise.
16662 (vect_permute_load_chain): Likewise.
16663 (vect_shift_permute_load_chain): Likewise.
16664 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
16665 (vect_transform_slp_perm_load): Likewise.
16666 (vect_schedule_slp_instance): Likewise.
16667 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16668 (vectorizable_mask_load_store): Likewise.
16669 (vectorizable_bswap): Likewise.
16670 (vectorizable_store): Likewise.
16671 (vectorizable_load): Likewise.
16672 * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
16673 vec_perm_indices instead of auto_vec_perm_indices. Use
16674 tree_to_vec_perm_builder to read the vector from a tree.
16675 * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
16676 vec_perm_builder instead of a vec_perm_indices.
16677 (have_whole_vector_shift): Use vec_perm_builder and
16678 vec_perm_indices instead of auto_vec_perm_indices. Leave the
16679 truncation to calc_vec_perm_mask_for_shift.
16680 (vect_create_epilog_for_reduction): Likewise.
16681 * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
16682 from auto_vec_perm_indices to vec_perm_indices.
16683 (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
16684 instead of changing individual elements.
16685 (aarch64_vectorize_vec_perm_const): Use new_vector to install
16686 the vector in d.perm.
16687 * config/arm/arm.c (expand_vec_perm_d::perm): Change
16688 from auto_vec_perm_indices to vec_perm_indices.
16689 (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
16690 instead of changing individual elements.
16691 (arm_vectorize_vec_perm_const): Use new_vector to install
16692 the vector in d.perm.
16693 * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
16694 Update vec_perm_builder constructor.
16695 (rs6000_expand_interleave): Likewise.
16696 * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
16697 (rs6000_expand_interleave): Likewise.
16699 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16701 * optabs-query.c (can_vec_perm_var_p): Check whether lowering
16702 to qimode could truncate the indices.
16703 * optabs.c (expand_vec_perm_var): Likewise.
16705 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16707 * Makefile.in (OBJS): Add vec-perm-indices.o.
16708 * vec-perm-indices.h: New file.
16709 * vec-perm-indices.c: Likewise.
16710 * target.h (vec_perm_indices): Replace with a forward class
16712 (auto_vec_perm_indices): Move to vec-perm-indices.h.
16713 * optabs.h: Include vec-perm-indices.h.
16714 (expand_vec_perm): Delete.
16715 (selector_fits_mode_p, expand_vec_perm_var): Declare.
16716 (expand_vec_perm_const): Declare.
16717 * target.def (vec_perm_const_ok): Replace with...
16718 (vec_perm_const): ...this new hook.
16719 * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
16720 (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
16721 * doc/tm.texi: Regenerate.
16722 * optabs.def (vec_perm_const): Delete.
16723 * doc/md.texi (vec_perm_const): Likewise.
16724 (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
16725 * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
16726 expand_vec_perm for constant permutation vectors. Assert that
16727 the mode of variable permutation vectors is the integer equivalent
16728 of the mode that is being permuted.
16729 * optabs-query.h (selector_fits_mode_p): Declare.
16730 * optabs-query.c: Include vec-perm-indices.h.
16731 (selector_fits_mode_p): New function.
16732 (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
16733 is defined, instead of checking whether the vec_perm_const_optab
16734 exists. Use targetm.vectorize.vec_perm_const instead of
16735 targetm.vectorize.vec_perm_const_ok. Check whether the indices
16736 fit in the vector mode before using a variable permute.
16737 * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
16738 vec_perm_indices instead of an rtx.
16739 (expand_vec_perm): Replace with...
16740 (expand_vec_perm_const): ...this new function. Take the selector
16741 as a vec_perm_indices rather than an rtx. Also take the mode of
16742 the selector. Update call to shift_amt_for_vec_perm_mask.
16743 Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
16744 Use vec_perm_indices::new_expanded_vector to expand the original
16745 selector into bytes. Check whether the indices fit in the vector
16746 mode before using a variable permute.
16747 (expand_vec_perm_var): Make global.
16748 (expand_mult_highpart): Use expand_vec_perm_const.
16749 * fold-const.c: Includes vec-perm-indices.h.
16750 * tree-ssa-forwprop.c: Likewise.
16751 * tree-vect-data-refs.c: Likewise.
16752 * tree-vect-generic.c: Likewise.
16753 * tree-vect-loop.c: Likewise.
16754 * tree-vect-slp.c: Likewise.
16755 * tree-vect-stmts.c: Likewise.
16756 * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
16758 * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
16759 * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
16760 (aarch64_vectorize_vec_perm_const_ok): Fuse into...
16761 (aarch64_vectorize_vec_perm_const): ...this new function.
16762 (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16763 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16764 * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
16765 * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
16766 * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16767 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16768 (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
16770 (arm_vectorize_vec_perm_const): ...this new function. Explicitly
16771 check for NEON modes.
16772 * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
16773 * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
16774 * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
16775 (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
16777 (ix86_vectorize_vec_perm_const): ...this new function. Incorporate
16778 the old VEC_PERM_CONST conditions.
16779 * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
16780 * config/ia64/vect.md (vec_perm_const<mode>): Delete.
16781 * config/ia64/ia64.c (ia64_expand_vec_perm_const)
16782 (ia64_vectorize_vec_perm_const_ok): Merge into...
16783 (ia64_vectorize_vec_perm_const): ...this new function.
16784 * config/mips/loongson.md (vec_perm_const<mode>): Delete.
16785 * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
16786 * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
16787 * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
16788 * config/mips/mips.c (mips_expand_vec_perm_const)
16789 (mips_vectorize_vec_perm_const_ok): Merge into...
16790 (mips_vectorize_vec_perm_const): ...this new function.
16791 * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
16792 * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
16793 * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
16794 * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
16795 * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
16796 (rs6000_expand_vec_perm_const): Delete.
16797 * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
16799 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16800 (altivec_expand_vec_perm_const_le): Take each operand individually.
16801 Operate on constant selectors rather than rtxes.
16802 (altivec_expand_vec_perm_const): Likewise. Update call to
16803 altivec_expand_vec_perm_const_le.
16804 (rs6000_expand_vec_perm_const): Delete.
16805 (rs6000_vectorize_vec_perm_const_ok): Delete.
16806 (rs6000_vectorize_vec_perm_const): New function.
16807 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
16808 an element count and rtx array.
16809 (rs6000_expand_extract_even): Update call accordingly.
16810 (rs6000_expand_interleave): Likewise.
16811 * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
16812 * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
16813 * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
16814 * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
16815 (rs6000_expand_vec_perm_const): Delete.
16816 * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
16817 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16818 (altivec_expand_vec_perm_const_le): Take each operand individually.
16819 Operate on constant selectors rather than rtxes.
16820 (altivec_expand_vec_perm_const): Likewise. Update call to
16821 altivec_expand_vec_perm_const_le.
16822 (rs6000_expand_vec_perm_const): Delete.
16823 (rs6000_vectorize_vec_perm_const_ok): Delete.
16824 (rs6000_vectorize_vec_perm_const): New function. Remove stray
16825 reference to the SPE evmerge intructions.
16826 (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
16827 an element count and rtx array.
16828 (rs6000_expand_extract_even): Update call accordingly.
16829 (rs6000_expand_interleave): Likewise.
16830 * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
16831 * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
16833 (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
16835 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16837 * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
16838 vector mode and that that mode matches the mode of the data
16840 (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
16841 out into expand_vec_perm_var. Do all CONST_VECTOR handling here,
16842 directly using expand_vec_perm_1 when forcing selectors into
16844 (expand_vec_perm_var): New function, split out from expand_vec_perm.
16846 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16848 * optabs-query.h (can_vec_perm_p): Delete.
16849 (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
16850 * optabs-query.c (can_vec_perm_p): Split into...
16851 (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
16852 (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
16853 particular selector is valid.
16854 * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
16855 * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
16856 (vect_grouped_load_supported): Likewise.
16857 (vect_shift_permute_load_chain): Likewise.
16858 * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
16859 (vect_transform_slp_perm_load): Likewise.
16860 * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
16861 (vectorizable_bswap): Likewise.
16862 (vect_gen_perm_mask_checked): Likewise.
16863 * fold-const.c (fold_ternary_loc): Likewise. Don't take
16864 implementations of variable permutation vectors into account
16865 when deciding which selector to use.
16866 * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
16867 vec_perm_const_optab is supported; instead use can_vec_perm_const_p
16868 with a false third argument.
16869 * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
16870 to test whether the constant selector is valid and can_vec_perm_var_p
16871 to test whether a variable selector is valid.
16873 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16875 * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
16876 * optabs-query.c (can_vec_perm_p): Likewise.
16877 * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
16878 instead of vec_perm_indices.
16879 * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
16880 (vect_gen_perm_mask_checked): Likewise,
16881 * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
16882 (vect_gen_perm_mask_checked): Likewise,
16884 2018-01-02 Richard Sandiford <richard.sandiford@linaro.org>
16886 * optabs-query.h (qimode_for_vec_perm): Declare.
16887 * optabs-query.c (can_vec_perm_p): Split out qimode search to...
16888 (qimode_for_vec_perm): ...this new function.
16889 * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
16891 2018-01-02 Aaron Sawdey <acsawdey@linux.vnet.ibm.com>
16893 * rtlanal.c (canonicalize_condition): Return 0 if final rtx
16894 does not have a conditional at the top.
16896 2018-01-02 Richard Biener <rguenther@suse.de>
16898 * ipa-inline.c (big_speedup_p): Fix expression.
16900 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
16903 * config/i386/x86-tune-costs.h: Increase cost of integer load costs
16906 2018-01-02 Jan Hubicka <hubicka@ucw.cz>
16910 * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
16911 cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
16912 and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
16913 cond_taken_branch_cost 3->4.
16915 2018-01-01 Jakub Jelinek <jakub@redhat.com>
16917 PR tree-optimization/83581
16918 * tree-loop-distribution.c (pass_loop_distribution::execute): Return
16919 TODO_cleanup_cfg if any changes have been made.
16921 PR middle-end/83608
16922 * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
16923 convert_modes if target mode has the right side, but different mode
16926 PR middle-end/83609
16927 * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
16928 last argument when extracting from CONCAT. If either from_real or
16929 from_imag is NULL, use expansion through memory. If result is not
16930 a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
16931 the parts directly to inner mode, if even that fails, use expansion
16934 PR middle-end/83623
16935 * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
16936 check for bswap in mode rather than HImode and use that in expand_unop
16939 Copyright (C) 2018 Free Software Foundation, Inc.
16941 Copying and distribution of this file, with or without modification,
16942 are permitted in any medium without royalty provided the copyright
16943 notice and this notice are preserved.